WO2008145039A1 - Methods, systems and devices for generating upsample filter and downsample filter and for performing encoding - Google Patents
Methods, systems and devices for generating upsample filter and downsample filter and for performing encoding Download PDFInfo
- Publication number
- WO2008145039A1 WO2008145039A1 PCT/CN2008/070902 CN2008070902W WO2008145039A1 WO 2008145039 A1 WO2008145039 A1 WO 2008145039A1 CN 2008070902 W CN2008070902 W CN 2008070902W WO 2008145039 A1 WO2008145039 A1 WO 2008145039A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- filter
- signal
- downsampling
- upsampling
- video
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/59—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/117—Filters, e.g. for pre-processing or post-processing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/63—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets
Definitions
- the present invention relates to coding techniques, and more particularly to a method, system and apparatus for generating up and down sampling filters and implementing coding. Background of the invention
- the video sequence is downsampled and upsampled into two samples.
- the original video sequence is downsampled to obtain a low resolution video signal.
- the corresponding macroblock obtained by the downsampling process is then subjected to upsampling processing to obtain a high resolution video, and the high resolution video is used for enhancement layer prediction or residual image prediction.
- the upsampling filter and downsampling filter designed by using the prior art scheme are: [1 0 -5 0 20 32 20 0 -5 0 1]/32 and [2 0 -4 -3 5 19 26 19 5 -3 -4 0 2]/64,
- the upper and lower sampling filters used for inter-layer residual prediction are: [1 1]/2 and [2 0 -4 -3 5 19 26 19 5 - 3 -4 0 2]/64. It can be seen that the prior art is designed for video coding. When the upsampling filter and downsampling filter are used, the correlation between the upsampling filter and the downsampling filter is not considered, that is, the designed upsampling filter and downsampling filter are independent.
- the up-and-down sampling operations using the above-described upsampling filter and downsampling filter are independent of each other and lack of correlation. Therefore, the video signals obtained by using these filters in the encoding process will have distortion, fast effects, etc., and bring a poor visual effect to the viewer.
- Embodiments of the present invention provide a method of generating an up-and-down sampling filter, by which an upsampling filter and a downsampling filter with better filtering performance can be generated.
- Embodiments of the present invention provide a method for implementing spatial scalable video coding by using a downsampling filter, by which the quality of coding can be improved.
- Embodiments of the present invention provide a method for implementing spatial scalable video coding by using an upsampling filter, by which the quality of coding can be improved.
- Embodiments of the present invention provide a device for implementing spatial scalable video coding by using a downsampling filter, and the quality of the coding can be improved by using the device.
- Embodiments of the present invention provide a device for implementing spatial scalable video coding by using an upsampling filter, and the quality of the coding can be improved by using the device.
- Embodiments of the present invention provide a system for implementing spatial scalable video coding by using a downsampling filter and an upsampling filter, and the quality of the coding can be improved by using the system.
- Embodiments of the present invention provide an apparatus for generating an up-and-down sampling filter, by which an upsampling filter and a downsampling filter having better filtering performance can be generated.
- Embodiments of the present invention provide a method of generating an up-and-down sampling filter, the method comprising:
- Embodiments of the present invention provide a method for implementing spatial scalable video coding by using a downsampling filter, where the method includes:
- the downsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal.
- the maximum value corresponds to the wavelet filter parameters generated by the downsampling filter.
- the present invention provides a method for implementing spatial scalable video coding using an upsampling filter, the method comprising:
- the upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal.
- the maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
- An embodiment of the present invention provides a device for implementing spatial scalable video coding by using a downsampling filter, where the device includes:
- a downsampling filter configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence
- a time domain decomposition unit configured to receive the downsampled video sequence from the downsampling filter, and decompose the downsampled video sequence into an I frame, a P frame, and a B frame;
- a motion coding unit configured to receive the P frame and the B frame from a time domain decomposition unit, and perform motion coding on the P frame and the B frame to obtain a motion prediction encoded signal
- a downsampling intra prediction and intra coding unit configured to receive the I frame from the time domain decomposition unit, perform intra prediction and intra coding on the I frame, to obtain a downsampled intra prediction coding signal;
- the downsampling filter is configured to perform downsampling processing and upsampling the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal.
- the wavelet filter parameters corresponding to the maximum energy, the generated downsampling filter.
- Embodiments of the present invention provide an apparatus for implementing spatial scalable video coding by using an upsampling filter, the apparatus comprising:
- An upsampling filter configured to receive a downsampled intra prediction encoded signal, and upsample the downsampled intra prediction encoded signal to obtain an upsampled video sequence
- Upsampling intra prediction and intra coding units for receiving from the upsampling filter An upsampled video sequence, performing intra prediction and intra coding on the upsampled video sequence to obtain an upsampled intra prediction encoded signal;
- the upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal.
- the maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
- Embodiments of the present invention provide a system for implementing spatial scalable video coding by using a downsampling filter and an upsampling filter, the system comprising:
- a downsampling filter configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence
- a downsampling coding unit configured to receive the downsampled video sequence from the downsampling filter, perform prediction and encoding on the downsampled video sequence, and generate a motion prediction coded signal and a downsampled intra prediction coded signal;
- An upsampling filter configured to receive the downsampled intra prediction encoded signal from the downsampling coding unit, and perform upsampling processing on the downsampled intra prediction encoded signal to obtain an upsampled video sequence;
- Upsampling intra prediction and intra coding unit configured to receive the upsampled video sequence from the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain upsampled intra prediction Coded signal
- the downsampling filter and the upsampling filter are down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal, according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, The wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal, the generated downsampling filter and the upsampling filter.
- Embodiments of the present invention provide an apparatus for generating an up-and-down sampling filter, the apparatus comprising: For downsampling and upsampling a video signal conforming to the Markov model to obtain a unit for reconstructing the signal;
- the wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal are calculated according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, and the units of the upsampling filter and the downsampling filter are generated.
- FIG. 1 is a schematic structural diagram of a JSVM of two-level video coding in an application scenario according to an embodiment of the present invention
- FIG. 2 is a schematic flow chart of a method for generating an upsampling filter and a downsampling filter according to an embodiment of the present invention
- FIG. 2a is a schematic flowchart of implementing video signal reconstruction according to an embodiment of the present invention
- FIG. 3 is a schematic flowchart of a method for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. 2;
- 3a is a schematic structural diagram of all pixels in a video signal frame
- FIG. 4 is a schematic flow chart of a method for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2;
- FIG. 5 is an effect diagram of processing an image using a filter generated by the present invention
- FIG. 6a is a graph of layer 0 intra-frame coding of a moving object (Mobile);
- FIG. 6b is a layer 0 frame for a foreman (foreman)
- Figure 6c is a graph of layer 0 intra-frame coding of a football;
- Figure 6d is a graph of layer 0 intra-frame coding of a bus;
- Figure 6al is a layer for Mobile The graphics obtained by encoding within 1 frame;
- Figure 6b 1 is a graph of layer 1 intra-frame coding of foreman
- Figure 6cl is a graph of layer 1 intraframe coding of the football
- Figure 6dl is a graph of layer 1 intra-frame coding of the bus
- FIG. 7 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. 2;
- FIG. 8 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2;
- FIG. 9 is a schematic structural diagram of a system for realizing spatial scalable video coding by using a downsampling filter and an upsampling filter generated by the embodiment shown in FIG. Mode for carrying out the invention
- a signal is sampled by a binary binary (2D) sampling technique.
- the general downsampling method is as follows: First, the high resolution macroblock is filtered by a downsampling filter; then, the pixels of the even row even columns are directly discarded to obtain a low resolution macroblock.
- the specific operation of the upsampling is as follows: the low-resolution macroblock obtained by the downsampling process is interlaced or interlaced to obtain a high-resolution macroblock with the resolution of both the horizontal and vertical directions being twice the original resolution. .
- the high-resolution macroblocks are convoluted by the upsampling filter in the horizontal direction and the vertical direction, respectively, and combined to obtain a high-resolution image.
- the pixels corresponding to the pixels of the low-resolution macroblock are directly filled on the high-resolution macroblock by using the characteristics of the up-sampling filter, and the remaining pixels of the high-resolution macroblock are processed by the interpolation method.
- FIG. 1 is a schematic structural diagram of a JSVM of two-level video coding in an application scenario according to an embodiment of the present invention.
- the model includes: a video unit 100, a 2D spatial domain downsampling unit 110, The core encoder 120, an encoder 130 compatible with the H.264 main architecture, a 2D spatial domain upsampling unit 140, a multiplexing unit 150, and a bitstream output unit 160.
- the core encoder 120 includes: a time domain decomposition unit 121, a motion coding unit 122, and an intra prediction/coding unit 123.
- the encoder 130 compatible with the H.264 main architecture includes: a time domain decomposition unit 131 and a motion coding unit 132. And intra prediction/encoding unit 133.
- the video unit 100 inputs the video signal to the 2D spatial domain downsampling unit 110, and the 2D spatial domain downsampling unit 110 performs a binary downsampling process on the video signal to generate a low resolution video signal, and outputs the generated low resolution video signal to the video signal.
- the time domain decomposition unit 121 and the time domain decomposition unit 131 respectively perform time domain grading on the received low resolution video signals, and output the gradation video signals to the motion coding unit 122 and the motion coding unit 132, respectively, and intra prediction/ The encoding unit 123 and the intra prediction/encoding unit 133.
- the motion encoding unit 122 and the motion encoding unit 132 perform motion encoding processing on the received video signal subjected to the time domain hierarchical processing, and output it to the multiplexing unit 150.
- the intra prediction/encoding unit 133 performs intra prediction and intra coding on the received video signal subjected to the time domain hierarchical processing, and transmits the video signal to the multiplexing unit 150, and the video signal is filled by the 2D spatial domain upsampling unit 140. And the interpolation process is sent to the intra prediction/encoding unit 123.
- the intra prediction/encoding unit 123 receives the signal transmitted by the 2D spatial domain upsampling unit 140, performs intra prediction on the signal, and transmits the predicted video signal to the multiplexing unit 150.
- the multiplexing unit 150 receives the video signals transmitted by the motion coding unit 122, the motion coding unit 132, the intra prediction/coding unit 123, and the intra prediction/encoding unit 133, and multiplexes the received video signals and transmits them to the bit stream. Output unit 160.
- Intra prediction includes: inter-layer intra texture prediction and inter-layer residual prediction.
- inter-layer intra-textural prediction means that the macroblock of the low-resolution layer is not upsampled by using motion information.
- the process of predicting a macroblock to a high resolution layer; inter-layer residual prediction refers to a process of up-sampling residual information of a low-resolution layer to obtain high-resolution layer residual information prediction.
- the motion information of the high-resolution layer may be predicted by using information of adjacent macroblocks or other layer-related macroblocks, or the high-resolution layer may be layered by upsampling processing of lower layer information.
- Inter-frame texture prediction and inter-layer residual prediction may be predicted by using information of adjacent macroblocks or other layer-related macroblocks, or the high-resolution layer may be layered by upsampling processing of lower layer information.
- an upsampling filter and a downsampling filter first, a video signal conforming to the Markov model needs to be downsampled and upsampled to obtain a reconstructed signal, and then the energy of the reconstructed signal is calculated; Then, according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, the wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal are calculated, and an upsampling filter and a downsampling filter are generated.
- FIG. 2 is a flow chart showing a method of generating an upsampling filter and a downsampling filter according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps:
- Step 21 Down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal.
- Step 22 Calculate the energy of the reconstructed signal.
- Step 23 According to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, calculate the wavelet filter parameters corresponding to the maximum value of the reconstructed signal, and generate an upsampling filter and a downsampling filter.
- Step 24 Integring the upsampling filter and the downsampling filter to obtain an integer upsampling filter and an integer downsampling filter.
- Step 25 Perform length reduction on the integer upsampling filter and the integer downsampling filter to obtain an optimized integer upsampling filter and an optimized integer downsampling filter.
- steps 24 and 25 above are optional steps.
- step 21 the video signal conforming to the Markov model is subjected to downsampling processing and The sampling process, the method for obtaining the reconstructed signal includes:
- / ⁇ ,..., ⁇ 2 ⁇ are two finite linear phase impulse response (FIR) low-pass filters, both of which are causal filters.
- 95% of the video signals are video signals conforming to the Markov model, and the process of downsampling and upsampling the video signals to obtain the reconstructed signals is represented in the form of a flow diagram.
- FIG. 2a is a schematic flowchart of implementing video signal reconstruction according to an embodiment of the present invention. As shown in Figure 2a,
- the video signal S is subjected to filtering processing using ⁇ , and then subjected to downsampling processing to obtain a signal S 0 . Then on the signal S.
- the upsampling process is performed, and the S(S) processed by the upsampling is used by the ( 2 ) filter.
- a filtering process is performed to obtain a reconstructed signal Si.
- the method for calculating the energy of the reconstructed signal according to the reconstructed signal is: In this step, calculating the energy of the reconstructed signal by using the reconstructed signal into the energy model, specifically:
- Si is a reconstructed signal obtained by down-sampling and up-sampling the video signal S;
- S 2 is a video signal obtained by correlating the video signal S with the SI; and
- S 3 is for the S Line down the resulting video signal. It is the energy of the video signal after downsampling and upsampling. Is the correlation energy of the video signal before the downsampling and upsampling process and the reconstructed signal. It is the energy of the video signal after the downsampling process.
- Both ⁇ and ⁇ are positive integers greater than zero, and ⁇ is a positive integer greater than or equal to zero.
- ⁇ 4 max ⁇ E[(l 12)e(S l )] + AE[e(S 2 )] + 0E[e(S 3 )] ⁇
- the selected energy model is: ⁇ maxi ⁇ d)] ⁇ ? ⁇ )] ⁇
- the energy of the video signal after downsampling and upsampling is:
- the parameterized bi-orthogonal wavelet theory is a parameterized formula of a bi-orthogonal wavelet filter with parameters.
- the wavelet filter parameters corresponding to the maximum value of the reconstructed signal are calculated, and the methods for generating the upsampling filter and the downsampling filter include:
- step 22 the energy formula of the reconstructed signal in this embodiment is:
- parameterized formula of the parameterized bi-orthogonal wavelet filter may include, but is not limited to: a single parameter double orthogonal wavelet parameterization formula, or a two parameter double orthogonal wavelet parameterization formula, or multiple parameters. Biorthogonal wavelet parameterization formula, or single parameter and multi-parameter bi-orthogonal wavelet parameterization formula.
- Method and utilization list for obtaining up-sampling filter and down-sampling filter by using two-parameter bi-orthogonal wavelet parameterization formula, multi-parameter bi-orthogonal wavelet parameterization formula, and single-parameter and multi-parameter bi-orthogonal wavelet parameterization formula The parameter bi-orthogonal d, the wave parameterization formula is the same as the method of obtaining the up-sampling filter and the down-sampling filter, and is also within the protection scope of the present invention.
- the wavelet filter parameters in the floating point form in Table 1 are written in the form of a matrix, that is, an upsampling filter and a downsampling filter can be generated.
- step 24 the upsampling filter and the downsampling filter generated in step 23 may be integerized, and the method of generating the integer upsampling filter and the integer downsampling filter is as follows:
- the wavelet filter is used for binary multiplication, and then the approximate result of the binary multiplication is obtained.
- the integer upsampling filter and the integer downsampling filter that can be generated are:
- step 25 the integer upsampling filter and integer downsampling generated in step 24 may be performed.
- the filter is length-reduced to generate an optimized integer upsampling filter and an optimized integer downsampling filter:
- the length of the filter directly affects the speed of its operation and the reconstruction quality of the video signal. Based on the trade-off between the operation speed and the reconstruction performance of the video signal, the effects of the integer up-sampling filter and the integer down-sampling filter are compared. Small values are omitted and added to the center value to generate an optimized integer upsampling filter and an optimized integer downsampling filter.
- the up-sampling filter and the downsampling of the chrominance component or the luminance component of the video signal may be utilized by using other energy models in step 23 and a parameterized bi-orthogonal wavelet filter parameterization formula.
- the filter is designed and optimized.
- step 24 can be used to perform integer processing on the floating point filter generated in step 23 to generate an integer filter.
- step 25 may be further performed to perform length optimization processing on the integer filter generated in step 24 to generate an optimized integer upsampling filter and an optimized integer downsampling filter.
- this embodiment mainly introduces the design of the upsampling filter and the downsampling filter in the spatial scalable coding process based on H.264. Compared with the prior art, it has the following advantages: 3 ⁇ 4 mouth:
- the bi-orthogonal wavelet theory with parameters is used in the design of the upsampling filter and the downsampling filter, so that the designed upsampling filter and downsampling filter are related.
- the upsampling filter and the downsampling filter generated in this embodiment are energy models.
- the upper and lower sampling filters generated in the case of maximizing the solution are optimal for the energy model in ensuring the same filter length.
- the generated integer filters can be clipped to different lengths to generate an optimized integer filter, so that the trade-off between the operation speed and the video signal reconstruction performance is the most realistic.
- Optimized integer upsampling filters and optimized integer downsampling filters are required.
- the embodiment shown in Figure 2 shows how to generate an upsampling filter and a downsampling filter, which can be used for spatially scalable video encoding operations.
- the spatially scalable video signal encoding process is divided into: an encoding process using a downsampling filter and an encoding process using an upsampling filter. The following are introduced in specific embodiments:
- FIG. 3 is a schematic flow chart of a method for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. As shown in FIG. 3, the method includes the following steps:
- Step 31 Down-sampling the video signal as the encoded input by using a downsampling filter to obtain a downsampled video sequence.
- Step 32 Perform time domain decomposition on the video signal sequence to obtain an internal (I) frame, a prediction (P) frame, and a bidirectional prediction (B) frame.
- I internal
- P prediction
- B bidirectional prediction
- Step 33 Perform intra prediction and intra coding on the I frame to obtain a downsampled intra prediction coding signal, and perform motion coding on the P frame and the B frame to obtain a motion prediction coded signal.
- step 31 the video signal as the encoded input is subjected to the downsampling filter.
- the sampling operation, the specific operation of obtaining the downsampled video sequence is:
- FIG. 3a is a schematic structural diagram of all pixels in a video signal frame.
- Figure 3a includes: a row including pixel E, pixel e, pixel F, pixel f, pixel G, pixel g, pixel H, pixel h, and pixel I; including pixel pixel K, pixel L, pixel k, pixel M, and a row of pixels N; a column including pixel eight, pixel a, pixel C, pixel b, pixel G, pixel c, pixel L, pixel d, and pixel P; and includes pixel B, pixel D, pixel H, pixel n, pixel M and the column of pixel Q.
- step 32 the video signal sequence is subjected to time domain decomposition to obtain an operation of an I frame, a P frame, and a B frame
- step 33 the I frame is subjected to intra prediction and intra coding to obtain a downsampled intra prediction encoded signal.
- the frame and the B frame are motion-encoded to obtain a motion prediction coded signal operation, which can be the same as the corresponding operation in the prior art.
- FIG. 4 is a flow chart showing a method for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2. As shown in Figure 4, the method includes the following steps:
- Step 41 Upsampling the downsampled intra prediction encoded signal by using an upsampling filter to obtain an upsampled video sequence.
- the signal is an intra-frame prediction and intra-frame coding of the I frame in step 33 to obtain a downsampled intra-predicted coded signal.
- Step 42 Perform intra prediction and intra coding on the upsampled video sequence to obtain an upsampled intra prediction encoded signal.
- step 41 to step 42 are as follows:
- step 41 the upsampling encoded signal is upsampled using an upsampling filter to obtain an upsampled video sequence.
- the specific method is: using an upsampling filter to perform inter-frame processing on the downsampled intraframe encoded signal.
- the 2D spatial domain upsampling process of inner texture prediction and inter-layer residual prediction yields an upsampled video sequence.
- the gray squares in Fig. 3a represent the pixels of the macroblock in the downsampled intra prediction encoded signal, all of which represent the pixels of the macroblock in the upsampled video sequence.
- Use upsampling filter The waver performs intra-layer intratextural prediction and inter-layer residual prediction on the downsampled intra prediction encoded signal by 2D spatial domain upsampling, and the specific method for obtaining the upsampled video sequence is:
- pixels which are evenly spaced apart from g in the horizontal direction and the vertical direction are processed in the same manner.
- step 42 the upsampled intra-predictive coded signal is formed by performing intra prediction and transform/entropy coding on the enhancement layer using the upsampled video sequence obtained in step 41 and the original texture information.
- the downsampled intra prediction encoded signal, the motion prediction encoded signal and the upsampled intra prediction encoded signal obtained by all encoding processes are multiplexed to obtain a reconstructed video signal.
- the upsampling filter and downsampling filter generated in the embodiment shown in Figure 2 are optimal for the energy model and are interrelated.
- the upsampling filter and the downsampling filter generated by the embodiment shown in FIG. 2 are respectively used in the encoding process of the video signals shown in FIG. 3 and FIG. 4, and can be reduced. Less image energy loss, improve the quality of encoding video signals, minimize energy loss and maximize video recovery quality, and improve the reconstruction quality of video signals, optimize the reconstruction of video signals at high resolution .
- the upsampling filter and downsampling filter designed using the method shown in Figure 2 process the image to significantly improve the resolution of high resolution.
- the quantitative and qualitative test results obtained by processing the image are described below in the specific examples.
- Table 2 shows the results obtained by quantitatively testing the images.
- the columns represent the names of the eight standard images, and the rows represent the filters for testing the images, followed by SVC/H.264, h5/h7, h7/h9, H7/H9, and HH7/HH13 filters. ,
- test data is expressed in units of peak signal-to-noise ratio (PSNR), where a larger value of PSNR indicates that the test result is more ideal.
- PSNR peak signal-to-noise ratio
- Fig. 5 is a view showing an effect obtained by processing an image using a filter generated by the present invention. From the graphics, you can get a very clear view.
- Fig. 6a is a diagram of layer 0 intraframe coding of a moving object (Mobile).
- layer 0 represents the core layer
- the abscissa represents the resolution, and the value ranges from 500bps to 3000bps
- the ordinate represents PSNR, and the larger the PSNR, the more ideal the test result is.
- Fig. 6a are the curves obtained by optimizing the luminance components in the inter-layer intratextural prediction using the downsampling filter and the upsampling filter. From top to bottom: — Filter— 3—7 Filter Set, Default JS VM Filter Set, our—Filter—5—7 Filter Set, and JVT-U 147—Filter— 5—7 Curves from the Filter Set, Default JSVM Filter The curves of the bank and JVT-U 147—Filter—5-7 filter bank are indicated by dashed lines.
- Our_Filter—3-7 filter bank and our_Filter—5-7 filter bank are filter banks designed using the method of the present invention, Default JSVM filter bank and JVT-U 147—Filter— 5-7
- the filter bank is an existing filter bank.
- the values of PSNR obtained using the Our-Filter-3-7 filter bank and our-Filter-7-7 filter bank are greater than those using Default JSVM filtering.
- the value of the PSNR obtained by the filter set and the JVT-U147-Filter- 5-7 filter bank so it can be concluded that the filter generated by the technical solution of the present invention is more than the existing JS VM 6.3.1 filter.
- Figure 6b is a graph of layer 0 intraframe coding of the foreman.
- Figure 6c is a graph of layer 0 intra-frame coding of a football.
- Figure 6d is a graph of layer 0 intraframe coding of a bus.
- Figure 6al is a graph obtained by layer 1 intra-frame coding of Mobile.
- the pattern shown in Fig. 6al is different from Fig. 6a only in that Fig. 6al is a graph obtained by intraframe coding layer 1.
- the curve obtained by the test, the default JSVM filter bank and the curve of the JVT-U 147-Filter-5-7 filter bank are indicated by dashed lines.
- Figure 6b 1 is a graph of layer 1 intra-frame coding of foreman.
- Figure 6c 1 is a graph of layer 1 intraframe coding of football.
- Figure 6dl is a graph of layer 1 intra-frame coding of the bus.
- the JSVM6.3.1 filter can significantly improve the reconstruction of high-resolution images when processing images.
- FIG. 7 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. As shown in FIG. 7, the apparatus includes: a downsampling filter, a time domain decomposition unit, a motion coding unit, and downsampling intra prediction and intra coding units.
- the downsampling filter is configured to receive a video signal as an encoded input, and perform downsampling processing on the video signal to obtain a downsampled video sequence, and send the downsampled video sequence Send to the time domain decomposition unit.
- the time domain decomposition unit receives the downsampled video sequence sent by the downsampling filter, decomposes the downsampled video sequence into I frames, P frames, and B frames, and sends the I frames to the downsampled intra prediction and intra coding units, The P frame and the B frame are sent to the motion coding unit.
- the motion coding unit receives the P frame and the B frame transmitted by the time domain decomposition unit, and performs motion coding on the P frame and the B frame to obtain a motion prediction coded signal.
- the downsampled intra prediction and intra coding unit receives the I frame, and performs intra prediction and intra coding on the I frame to obtain a downsampled predictive coded signal.
- the downsampling filter used in this embodiment is to perform downsampling processing and upsampling processing on the video signal conforming to the Markov model to obtain a reconstructed signal, and then according to the bi-orthogonal wavelet theory with parameters and reconstruction.
- the energy of the signal, the wavelet filter parameter corresponding to the maximum energy of the reconstructed signal is calculated, and the generated downsampling filter is used.
- the bi-orthogonal wavelet theory with parameters is a bi-orthogonal wavelet filter with parameters. Parameterization formula.
- the video signal conforming to the Markov model used to generate the downsampling filter and the video signal as the coded input may be the same signal or different signals.
- the video signal input as a code can be all video signals, and about 95% of all of these video signals are video signals conforming to the Markov model.
- FIG. 8 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. As shown in Figure 8, the apparatus includes: an upsampling filter and upsampled intra prediction and intra coding units.
- the upsampling filter is configured to receive the downsampled intra prediction coding signal outputted by the downsampled intra prediction and the intra coding unit in FIG. 7, and perform upsampling processing on the received downsampled intra prediction coding signal, An upsampled video sequence is obtained, which is sent to the upsampled intra prediction and intra coding units.
- the upsampling intra prediction and intra coding unit is configured to receive an upsampled video sequence sent by the upsampling filter, perform intra prediction and intraframe coding on the received upsampled video sequence, to obtain an upsampled intra prediction encoded signal.
- the upsampling filter used in this embodiment is to perform downsampling processing and upsampling processing on the video signal to obtain a reconstructed signal, and then calculate according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal.
- the video signal conforming to the Markov model used to generate the upsampling filter and the video signal used as the code input may be the same signal or different signals, which are generally not the same.
- FIG. 9 is a schematic structural diagram of a system for realizing spatial scalable video coding by using a downsampling filter and an upsampling filter generated by the embodiment shown in FIG.
- the system includes: a downsampling filter, a downsampling coding unit, an upsampling filter, upsampling intra prediction and intra coding units, and a multiplexing unit.
- the downsampling coding unit includes: a time domain decomposition unit, a motion coding unit, and a downsampling intra prediction and intra coding unit.
- the downsampling filter is configured to receive a video signal as an encoded input, perform downsampling processing on the video signal, obtain a downsampled video sequence, and send the downsampled video sequence to a time domain decomposition unit of the downsampling coding unit.
- a time domain decomposition unit configured to receive a downsampled video sequence sent by the downsampling filter, decompose the received downsampled video sequence into I frames, P frames, and B frames, and send the I frames to the downsampled intra prediction and frame.
- the inner coding unit transmits the P frame and the B frame to the motion coding unit.
- the motion coding unit is configured to receive the P frame and the B frame, perform motion coding on the received P frame and the B frame to obtain a motion coded signal, and send the motion prediction coded signal to the multiplexing unit.
- Down-sampling intra-pre and intra-coded units for receiving I-frames, for incoming I-frames
- the intra-frame prediction and the intra-frame coding obtain a downsampled intra prediction and coding signal, and send the downsampled intra prediction coding signal to the upsampling filter and the multiplexing unit.
- An upsampling filter configured to receive the downsampled intra prediction coded signal, perform upsampling on the received downsampled intraframe predictive coded signal, obtain an upsampled video sequence, and send to the upsampled intraframe and intraframe coding unit.
- An upsampled intra prediction and intra coding unit configured to receive an upsampled video sequence sent by the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain an upsampled intra prediction coding signal, and The upsampled intra prediction encoded signal is transmitted to the multiplexing unit.
- a multiplexing unit configured to receive the downsampled intra prediction encoded signal, the motion prediction encoded signal, and the upsampled intra prediction encoded signal, and perform synthesis processing on the received signal to obtain a video encoded signal.
- the downsampling filter and the upsampling filter used in this implementation are down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal, according to the bi-orthogonal wavelet theory with parameters.
- the video signal conforming to the Markov model used to generate the downsampling filter and the upsampling filter may be the same signal or a different signal as the encoded input, and is generally different. signal.
- the multiplexing unit mainly performs synthesis processing on the downsampled intra prediction encoded signal, the motion prediction encoded signal, and the upsampled intra prediction encoded signal to obtain a video encoded signal. If the spatial scalable video coding system does not synthesize these three signals, the multiplexing unit may not be included in this embodiment.
- the introduction of the embodiment shown in Fig. 9 is ended.
- the up and down sampling filters appearing in FIGS. 2 to 6 are formulas, and the up and down sampling filters in FIGS. 7 to 9 are actual physical units.
- the upsampling filter and the downsampling filter generated by the embodiment of the present invention are upper and lower sampling filters generated when the energy model is maximized, and are optimal for the energy model in ensuring the same filter length.
- the upsampling filter and the downsampling filter generated by the embodiment of the present invention are used for encoding processing of a video signal, which can reduce image energy loss, improve quality of encoding the video signal, minimize energy loss, and restore video quality. maximize.
- the bi-orthogonal wavelet theory with parameters is applied to the design of the upsampling filter and the downsampling filter, so that the designed upsampling filter and downsampling filter are related.
- the reconstruction effect of the video signal under high resolution is optimized.
- the generated integer filter can be tailored to different lengths to generate an optimized integer filter, thereby generating the most trade-off between the operation speed and the video signal reconstruction performance.
- Optimized integer upsampling filters and optimized integer downsampling filters that meet actual needs.
- embodiments of the present invention may be implemented in a computer program language, or may be implemented by hardware components of a preset program, other related components, or a combination of hardware and software.
- the methods, apparatus, and systems shown in the embodiments of the present invention can be fabricated into a computer program product for use in a computer system.
- it can be fabricated as a series of computer instructions that are solidified in actual media (CD, CD-ROM, ROM or hard disk), or media implemented using wireless technology.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
A method for generating an upsample filter and a downsample filter. The method comprises the steps of: upsampling and downsampling the video signal in conformity with Markov Model so as to derive the reconstructed signal; calculating the energy of the reconstructed signal; calculating the parameters of the wavelet filter corresponding to the maximum value of the energy of the reconstructed signal on the basis of the biorthogonal wavelet theory with parameters and the energy of the reconstructed signal, and generating an upsample filter and a downsample filter. Meanwhile, also provided are a method, device and system for performing the spatial scalable video encoding by use of the upsample filter and the downsample filter according to the present invention. With the solution according to the embodiment of the present invention, it is possible to minimize the loss of the encoding energy so that the qulity and performance of the video signal encoding could be improved.
Description
生成上下釆样滤波器及实现编码的方法、 ***和装置 技术领域 Method, system and device for generating upper and lower sampling filters and realizing coding
本发明涉及编码技术， 特别涉及生成上下采样滤波器及实现编码的 方法、 ***和装置。 发明背景 The present invention relates to coding techniques, and more particularly to a method, system and apparatus for generating up and down sampling filters and implementing coding. Background of the invention
随着互联网、 无线通讯等技术的发展， 人们对以视频为主的多媒体 数据的需求也在迅速增加。 传输不经过压缩的多媒体数据需要占用很大 的带宽，需要容量很大的存储介质。多媒体数据的传输量往往非常巨大， 所以需要先对多媒体数据进行压缩编码， 然后再进行传输。 此外， 在传 输过程中要求编码和传输具有随时间变化的网络带宽自适应能力、 抗误 码性能， 以及支持空间分级、 时间分级和基于峰值信噪比（PSNR )质 量分级的能力， 从而可以允许具有不同终端硬件设施条件和需求的用户 可以根据自身情况截取嵌入码流， 并将截取的嵌入码流还原成不同质量 的信息。 With the development of technologies such as the Internet and wireless communications, the demand for video-based multimedia data is rapidly increasing. The transmission of uncompressed multimedia data requires a large amount of bandwidth and requires a large storage medium. The amount of multimedia data transmission is often very large, so it is necessary to compress and encode the multimedia data before transmitting. In addition, encoding and transmission are required to have time-varying network bandwidth adaptation capabilities, error-resistance performance, and support for spatial grading, time grading, and peak signal-to-noise ratio (PSNR) quality grading during transmission, allowing for Users with different terminal hardware facilities and requirements can intercept the embedded code stream according to their own conditions and restore the intercepted embedded code stream to different quality information.
在可分级视频编码（SVC ) 空间可分级编码过程中， 对视频序列进 行下采样和上采样两重采样处理。 对原始视频序列进行下采样处理， 得 到低分辨率的视频信号。 然后对经下采样处理得到的相应的宏块进行上 采样处理得到高分辨率的视频， 利用该高分辨率的视频进行增强层预测 或残差图像预测。 In the scalable video coding (SVC) spatial scalable coding process, the video sequence is downsampled and upsampled into two samples. The original video sequence is downsampled to obtain a low resolution video signal. The corresponding macroblock obtained by the downsampling process is then subjected to upsampling processing to obtain a high resolution video, and the high resolution video is used for enhancement layer prediction or residual image prediction.
利用现有技术方案设计得到的上采样滤波器和下采样滤波器分别 为： [1 0 -5 0 20 32 20 0 -5 0 1]/32 和 [2 0 -4 -3 5 19 26 19 5 -3 -4 0 2]/64, 层间残差预测所采用的上采样滤波器和下采样滤波器分别为： [1 1]/2和 [2 0 -4 -3 5 19 26 19 5 -3 -4 0 2]/64。 可见， 现有技术在设计用于视频编码
的上采样滤波器和下采样滤波器时， 并未考虑上采样滤波器和下采样滤 波器的关联性， 即设计出的上采样滤波器和下采样滤波器是独立的。 The upsampling filter and downsampling filter designed by using the prior art scheme are: [1 0 -5 0 20 32 20 0 -5 0 1]/32 and [2 0 -4 -3 5 19 26 19 5 -3 -4 0 2]/64, The upper and lower sampling filters used for inter-layer residual prediction are: [1 1]/2 and [2 0 -4 -3 5 19 26 19 5 - 3 -4 0 2]/64. It can be seen that the prior art is designed for video coding. When the upsampling filter and downsampling filter are used, the correlation between the upsampling filter and the downsampling filter is not considered, that is, the designed upsampling filter and downsampling filter are independent.
在实际应用中， 利用上述上采样滤波器和下采样滤波器进行的上下 采样操作之间是相互独立的、 缺乏关联。 因此， 在编码过程中采用这些 滤波器编码得到的视频信号会出现失真、 快效应等现象， 给观众带来较 差的视觉效果。 In practical applications, the up-and-down sampling operations using the above-described upsampling filter and downsampling filter are independent of each other and lack of correlation. Therefore, the video signals obtained by using these filters in the encoding process will have distortion, fast effects, etc., and bring a poor visual effect to the viewer.
综上所述， 由于现有技术设计出的上采样滤波器和下采样滤波器缺 乏关联性， 且滤波性能较差， 从而使用现有的上采样滤波器和下采样滤 波器进行编码的质量较差。 发明内容 In summary, due to the lack of correlation between the upsampling filter and the downsampling filter designed in the prior art, and the filtering performance is poor, the quality of encoding using the existing upsampling filter and downsampling filter is relatively high. difference. Summary of the invention
本发明的实施例提供一种生成上下采样滤波器的方法， 使用该方法 可以生成滤波性能较好的上采样滤波器和下采样滤波器。 Embodiments of the present invention provide a method of generating an up-and-down sampling filter, by which an upsampling filter and a downsampling filter with better filtering performance can be generated.
本发明实施例提供一种利用下采样滤波器实现空间可分级视频编码 的方法， 通过该方法可以提高编码的质量。 Embodiments of the present invention provide a method for implementing spatial scalable video coding by using a downsampling filter, by which the quality of coding can be improved.
本发明实施例提供一种利用上采样滤波器实现空间可分级视频编码 的方法， 通过该方法可以提高编码的质量。 Embodiments of the present invention provide a method for implementing spatial scalable video coding by using an upsampling filter, by which the quality of coding can be improved.
本发明实施例提供一种利用下采样滤波器实现空间可分级视频编码 的装置， 通过使用该装置可以提高编码的质量。 Embodiments of the present invention provide a device for implementing spatial scalable video coding by using a downsampling filter, and the quality of the coding can be improved by using the device.
本发明实施例提供一种利用上采样滤波器实现空间可分级视频编码 的装置， 通过使用该装置可以提高编码的质量。 Embodiments of the present invention provide a device for implementing spatial scalable video coding by using an upsampling filter, and the quality of the coding can be improved by using the device.
本发明实施例提供一种利用下采样滤波器和上采样滤波器实现空间 可分级视频编码的***， 通过使用该***可以提高编码的质量。 Embodiments of the present invention provide a system for implementing spatial scalable video coding by using a downsampling filter and an upsampling filter, and the quality of the coding can be improved by using the system.
本发明的实施例提供一种生成上下采样滤波器的装置， 使用该装置 可以生成滤波性能较好的上采样滤波器和下采样滤波器。
本发明的实施例提供了一种生成上下采样滤波器的方法， 该方法包 括： Embodiments of the present invention provide an apparatus for generating an up-and-down sampling filter, by which an upsampling filter and a downsampling filter having better filtering performance can be generated. Embodiments of the present invention provide a method of generating an up-and-down sampling filter, the method comprising:
对符合马尔科夫模型的视频信号进行下采样和上采样， 得到重建信 号； Downsampling and upsampling the video signal conforming to the Markov model to obtain a reconstructed signal;
计算所述重建信号的能量； Calculating the energy of the reconstructed signal;
根据带参数的双正交小波理论和所述重建信号的能量， 计算所述重 建信号的能量的最大值所对应的小波滤波器参数， 生成上采样滤波器和 下采样滤波器。 And calculating a wavelet filter parameter corresponding to a maximum value of the energy of the reconstructed signal according to a bi-orthogonal wavelet theory with parameters and an energy of the reconstructed signal, and generating an upsampling filter and a downsampling filter.
本发明实施例提供了一种利用下采样滤波器实现空间可分级视频编 码的方法， 该方法包括： Embodiments of the present invention provide a method for implementing spatial scalable video coding by using a downsampling filter, where the method includes:
使用下采样滤波器对用于编码的视频信号进行下采样， 得到下采样 视频序列； Downsampling the video signal used for encoding using a downsampling filter to obtain a downsampled video sequence;
对所述下采样视频序列进行时域分解， 得到内部 I帧、 预测 P帧和 双向预测 B帧； Performing time domain decomposition on the downsampled video sequence to obtain an internal I frame, a predicted P frame, and a bidirectional predicted B frame;
对所述 I帧进行帧内预测和帧内编码， 得到下采样帧内预测编码信 号； 对所述 P帧和 B帧进行运动编码， 得到运动预测编码信号； Performing intra prediction and intra coding on the I frame to obtain a downsampled intra prediction coding signal; performing motion coding on the P frame and the B frame to obtain a motion prediction coding signal;
所述下采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的下采样滤波器。 The downsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the downsampling filter.
本发明提供了一种利用上采样滤波器实现空间可分级视频编码的方 法， 该方法包括： The present invention provides a method for implementing spatial scalable video coding using an upsampling filter, the method comprising:
使用上采样滤波器对下采样帧内预测编码信号进行上采样， 得到上 采样视频序列； Upsampling the downsampled intraframe encoded signal using an upsampling filter to obtain an upsampled video sequence;
对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内
预测编码信号； Performing intra prediction and intra coding on the upsampled video sequence to obtain an upsampled frame Predictive coded signal;
所述上采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的上采样滤波器。 The upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
本发明实施例提供了一种利用下采样滤波器实现空间可分级视频编 码的装置， 该装置包括： An embodiment of the present invention provides a device for implementing spatial scalable video coding by using a downsampling filter, where the device includes:
下采样滤波器， 用于接收用于编码的视频信号， 对所述视频信号进 行下采样， 得到下采样视频序列； a downsampling filter, configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence;
时域分解单元，用于从所述下采样滤波器接收所述下采样视频序列， 将所述下采样视频序列分解成 I帧、 P帧和 B帧； a time domain decomposition unit, configured to receive the downsampled video sequence from the downsampling filter, and decompose the downsampled video sequence into an I frame, a P frame, and a B frame;
运动编码单元， 用于从时域分解单元接收所述 P帧和 B帧， 对所述 P帧和 B帧进行运动编码得到运动预测编码信号； a motion coding unit, configured to receive the P frame and the B frame from a time domain decomposition unit, and perform motion coding on the P frame and the B frame to obtain a motion prediction encoded signal;
下采样帧内预测和帧内编码单元， 用于从所述时域分解单元接收所 述 I帧， 对所述 I帧进行帧内预测和帧内编码， 得到下采样帧内预测编 码信号； a downsampling intra prediction and intra coding unit, configured to receive the I frame from the time domain decomposition unit, perform intra prediction and intra coding on the I frame, to obtain a downsampled intra prediction coding signal;
所述下采样滤波器是对符合马尔科夫模型的视频信号进行下采样处 理和上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信 号的能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参 数， 生成的下采样滤波器。 The downsampling filter is configured to perform downsampling processing and upsampling the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The wavelet filter parameters corresponding to the maximum energy, the generated downsampling filter.
本发明实施例提供了一种利用上采样滤波器实现空间可分级视频编 码的装置， 该装置包括： Embodiments of the present invention provide an apparatus for implementing spatial scalable video coding by using an upsampling filter, the apparatus comprising:
上采样滤波器， 用于接收下采样帧内预测编码信号， 并对所述下采 样帧内预测编码信号进行上采样， 得到上采样视频序列； An upsampling filter, configured to receive a downsampled intra prediction encoded signal, and upsample the downsampled intra prediction encoded signal to obtain an upsampled video sequence;
上采样帧内预测和帧内编码单元， 用于从所述上采样滤波器接收所
述上采样视频序列， 对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内预测编码信号； Upsampling intra prediction and intra coding units for receiving from the upsampling filter An upsampled video sequence, performing intra prediction and intra coding on the upsampled video sequence to obtain an upsampled intra prediction encoded signal;
所述上采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的上采样滤波器。 The upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
本发明实施例提供了一种利用下采样滤波器和上采样滤波器实现空 间可分级视频编码的***， 该***包括： Embodiments of the present invention provide a system for implementing spatial scalable video coding by using a downsampling filter and an upsampling filter, the system comprising:
下采样滤波器， 用于接收用于编码的视频信号， 对所述视频信号进 行下采样， 得到下采样视频序列； a downsampling filter, configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence;
下采样编码单元， 用于从所述下采样滤波器接收所述下采样视频序 列， 对所述下采样视频序列进行预测和编码， 生成运动预测编码信号和 下采样帧内预测编码信号； And a downsampling coding unit, configured to receive the downsampled video sequence from the downsampling filter, perform prediction and encoding on the downsampled video sequence, and generate a motion prediction coded signal and a downsampled intra prediction coded signal;
上采样滤波器， 用于从所述下采样编码单元接收所述下采样帧内预 测编码信号， 对所述下采样帧内预测编码信号进行上采样处理， 得到上 采样视频序列； An upsampling filter, configured to receive the downsampled intra prediction encoded signal from the downsampling coding unit, and perform upsampling processing on the downsampled intra prediction encoded signal to obtain an upsampled video sequence;
上采样帧内预测和帧内编码单元， 用于从所述上采样滤波器接收所 述上采样视频序列， 对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内预测编码信号； Upsampling intra prediction and intra coding unit, configured to receive the upsampled video sequence from the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain upsampled intra prediction Coded signal
所述下采样滤波器和上采样滤波器是对符合马尔科夫模型的视频信 号进行下采样和上采样得到重建信号， 根据带参数的双正交小波理论和 所述重建信号的能量， 计算所述重建信号的能量的最大值所对应的小波 滤波器参数， 生成的下采样滤波器和上采样滤波器。 The downsampling filter and the upsampling filter are down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal, according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, The wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal, the generated downsampling filter and the upsampling filter.
本发明的实施例提供了一种生成上下采样滤波器的装置， 该装置包 括：
用于对符合马尔科夫模型的视频信号进行下采样和上采样， 得到重 建信号的单元； Embodiments of the present invention provide an apparatus for generating an up-and-down sampling filter, the apparatus comprising: For downsampling and upsampling a video signal conforming to the Markov model to obtain a unit for reconstructing the signal;
计算所述重建信号的能量的单元； 和 Means for calculating the energy of the reconstructed signal; and
根据带参数的双正交小波理论和所述重建信号的能量， 计算所述重 建信号的能量的最大值所对应的小波滤波器参数， 生成上采样滤波器和 下采样滤波器的单元。 附图简要说明 The wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal are calculated according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, and the units of the upsampling filter and the downsampling filter are generated. BRIEF DESCRIPTION OF THE DRAWINGS
图 1为本发明实施例的一个应用场景中的两级视频编码的 JSVM 的结构示意图； 1 is a schematic structural diagram of a JSVM of two-level video coding in an application scenario according to an embodiment of the present invention;
图 2 为本发明实施例的生成上采样滤波器和下采样滤波器的方 法的流程示意图； 2 is a schematic flow chart of a method for generating an upsampling filter and a downsampling filter according to an embodiment of the present invention;
图 2a为本发明实施例的实现视频信号重建的流程示意图； 图 3为利用图 2所示实施例生成的下采样滤波器实现空间可分级 视频编码的方法的流程示意图； 2a is a schematic flowchart of implementing video signal reconstruction according to an embodiment of the present invention; FIG. 3 is a schematic flowchart of a method for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. 2;
图 3a为视频信号帧中全部像素的结构示意图； 3a is a schematic structural diagram of all pixels in a video signal frame;
图 4为利用图 2所示实施例生成的上采样滤波器实现空间可分级 视频编码的方法的流程示意图； 4 is a schematic flow chart of a method for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2;
图 5为使用本发明生成的滤波器对图像进行处理得到的效果图； 图 6a是对运动物体（ Mobile ) 进行层 0帧内编码所得的图形； 图 6b是对工头（foreman) 进行层 0帧内编码所得的图形； 图 6c是对足球 (football) 进行层 0帧内编码所得的图形； 图 6d是对公共汽车 (bus) 进行层 0帧内编码所得的图形； 图 6al是对 Mobile进行层 1帧内编码所得到的图形； 5 is an effect diagram of processing an image using a filter generated by the present invention; FIG. 6a is a graph of layer 0 intra-frame coding of a moving object (Mobile); FIG. 6b is a layer 0 frame for a foreman (foreman) Figure 6c is a graph of layer 0 intra-frame coding of a football; Figure 6d is a graph of layer 0 intra-frame coding of a bus; Figure 6al is a layer for Mobile The graphics obtained by encoding within 1 frame;
图 6b 1是对 foreman进行层 1帧内编码所得的图形；
图 6cl是对 football进行层 1帧内编码所得的图形； Figure 6b 1 is a graph of layer 1 intra-frame coding of foreman; Figure 6cl is a graph of layer 1 intraframe coding of the football;
图 6dl是对 bus进行层 1帧内编码所得的图形； Figure 6dl is a graph of layer 1 intra-frame coding of the bus;
图 7为利用图 2所示实施例生成的下采样滤波器实现空间可分级 视频编码的装置的结构示意图； 7 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. 2;
图 8为利用图 2所示实施例生成的上采样滤波器实现空间可分级 视频编码的装置的结构示意图； 8 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2;
图 9为利用图 2所示实施例生成的下采样滤波器和上采样滤波器 实现空间可分级视频编码的***的结构示意图。 实施本发明的方式 FIG. 9 is a schematic structural diagram of a system for realizing spatial scalable video coding by using a downsampling filter and an upsampling filter generated by the embodiment shown in FIG. Mode for carrying out the invention
在本发明实施例的一个应用场景中， 采用二进可分二维（2D )采样 技术对信号进行采样处理。 一般下采样的方法为： 首先， 将高分辨率的 宏块用下采样滤波器进行滤波； 然后， 直接舍弃偶行偶列的像素， 得到 低分辨率的宏块。 上采样的具体操作为： 将下采样处理得到的低分辨率 的宏块进行隔行或隔列补零， 得到水平方向和竖直方向的分辨率均为原 分辨率两倍的高分辨率宏块。 例如， 设 fH和 t分别代表高分辨率和低分 辨率的图像， 令 fH ( 2x, 2y ) =fL ( x, y )。 然后按照水平方向和竖直方向 分别用上采样滤波器对高分辨率宏块进行卷积处理， 合并得到高分辨率 图像。 实际操作中， 利用上采样滤波器自身的特性， 在高分辨率宏块上 直接填充与低分辨率宏块的像素对应的像素， 利用插值方法处理高分辨 率宏块的剩余像素。 In an application scenario of the embodiment of the present invention, a signal is sampled by a binary binary (2D) sampling technique. The general downsampling method is as follows: First, the high resolution macroblock is filtered by a downsampling filter; then, the pixels of the even row even columns are directly discarded to obtain a low resolution macroblock. The specific operation of the upsampling is as follows: the low-resolution macroblock obtained by the downsampling process is interlaced or interlaced to obtain a high-resolution macroblock with the resolution of both the horizontal and vertical directions being twice the original resolution. . For example, let f H and t represent high resolution and low resolution images, respectively, such that f H ( 2x, 2y ) = fL ( x, y ). Then, the high-resolution macroblocks are convoluted by the upsampling filter in the horizontal direction and the vertical direction, respectively, and combined to obtain a high-resolution image. In actual operation, the pixels corresponding to the pixels of the low-resolution macroblock are directly filled on the high-resolution macroblock by using the characteristics of the up-sampling filter, and the remaining pixels of the high-resolution macroblock are processed by the interpolation method.
目前， 联合视频组（JVT ) SVC所采用的编码模型为联合视频组视 频分层编码模型 （JSVM ) 6.3.1。 图 1为本发明实施例的一个应用场景中 的两级视频编码的 JSVM的结构示意图。 Currently, the coding model used by the Joint Video Group (JVT) SVC is the Joint Video Group Video Layered Coding Model (JSVM) 6.3.1. FIG. 1 is a schematic structural diagram of a JSVM of two-level video coding in an application scenario according to an embodiment of the present invention.
如图 1所示， 该模型包括： 视频单元 100、 2D空域下采样单元 110、
核心编码器 120、 与 H.264主架构兼容的编码器 130、 2D空域上采样单元 140、 复用单元 150和比特流输出单元 160。 其中， 核心编码器 120包括： 时域分解单元 121、运动编码单元 122和帧内预测 /编码单元 123; 与 H.264 主架构兼容的编码器 130包括： 时域分解单元 131、 运动编码单元 132和 帧内预测 /编码单元 133。 As shown in FIG. 1, the model includes: a video unit 100, a 2D spatial domain downsampling unit 110, The core encoder 120, an encoder 130 compatible with the H.264 main architecture, a 2D spatial domain upsampling unit 140, a multiplexing unit 150, and a bitstream output unit 160. The core encoder 120 includes: a time domain decomposition unit 121, a motion coding unit 122, and an intra prediction/coding unit 123. The encoder 130 compatible with the H.264 main architecture includes: a time domain decomposition unit 131 and a motion coding unit 132. And intra prediction/encoding unit 133.
视频单元 100将视频信号输入到 2D空域下采样单元 110 , 2D空域下采 样单元 110对该视频信号进行二进下采样处理生成低分辨率的视频信 号， 将生成的低分辨率的视频信号输出到时域分解单元 121和时域分解 单元 131。 The video unit 100 inputs the video signal to the 2D spatial domain downsampling unit 110, and the 2D spatial domain downsampling unit 110 performs a binary downsampling process on the video signal to generate a low resolution video signal, and outputs the generated low resolution video signal to the video signal. The time domain decomposition unit 121 and the time domain decomposition unit 131.
时域分解单元 121和时域分解单元 131分别对接收到的低分辨率视频 信号进行时域分级， 将分级后的视频信号分别输出到运动编码单元 122 和运动编码单元 132, 以及帧内预测 /编码单元 123和帧内预测 /编码单元 133。 The time domain decomposition unit 121 and the time domain decomposition unit 131 respectively perform time domain grading on the received low resolution video signals, and output the gradation video signals to the motion coding unit 122 and the motion coding unit 132, respectively, and intra prediction/ The encoding unit 123 and the intra prediction/encoding unit 133.
运动编码单元 122和运动编码单元 132对接收到的经时域分级处理的 视频信号进行运动编码处理， 并输出到复用单元 150。 The motion encoding unit 122 and the motion encoding unit 132 perform motion encoding processing on the received video signal subjected to the time domain hierarchical processing, and output it to the multiplexing unit 150.
帧内预测 /编码单元 133对接收到的经时域分级处理的视频信号进行 帧内预测和帧内编码， 并发送给复用单元 150, 并将该视频信号经 2D空 域上采样单元 140的填充和插值处理后发送给帧内预测 /编码单元 123。 The intra prediction/encoding unit 133 performs intra prediction and intra coding on the received video signal subjected to the time domain hierarchical processing, and transmits the video signal to the multiplexing unit 150, and the video signal is filled by the 2D spatial domain upsampling unit 140. And the interpolation process is sent to the intra prediction/encoding unit 123.
帧内预测 /编码单元 123接收 2D空域上采样单元 140发送的信号，对该 信号进行帧内预测， 并将经预测处理后的视频信号发送给复用单元 150。 The intra prediction/encoding unit 123 receives the signal transmitted by the 2D spatial domain upsampling unit 140, performs intra prediction on the signal, and transmits the predicted video signal to the multiplexing unit 150.
复用单元 150接收运动编码单元 122、 运动编码单元 132、 帧内预测 / 编码单元 123和帧内预测 /编码单元 133发送的视频信号，对接收到的视频 信号进行复用合并后发送给比特流输出单元 160。 The multiplexing unit 150 receives the video signals transmitted by the motion coding unit 122, the motion coding unit 132, the intra prediction/coding unit 123, and the intra prediction/encoding unit 133, and multiplexes the received video signals and transmits them to the bit stream. Output unit 160.
帧内预测包括： 层间帧内纹理预测和层间残差预测。 其中， 层间帧 内纹理预测是指没有利用运动信息， 将低分辨率层的宏块通过上采样得
到高分辨率层预测宏块的过程； 层间残差预测是指对低分辨率层的残差 信息进行上采样得到高分辨率层残差信息预测的过程。 在帧预测过程 中， 可以利用相邻宏块或其他层相关宏块的信息对高分辨率层进行运动 信息的预测， 也可以利用对较低层信息的上采样处理对高分辨率层进行 层间帧内纹理预测和层间残差预测。 Intra prediction includes: inter-layer intra texture prediction and inter-layer residual prediction. Among them, inter-layer intra-textural prediction means that the macroblock of the low-resolution layer is not upsampled by using motion information. The process of predicting a macroblock to a high resolution layer; inter-layer residual prediction refers to a process of up-sampling residual information of a low-resolution layer to obtain high-resolution layer residual information prediction. In the frame prediction process, the motion information of the high-resolution layer may be predicted by using information of adjacent macroblocks or other layer-related macroblocks, or the high-resolution layer may be layered by upsampling processing of lower layer information. Inter-frame texture prediction and inter-layer residual prediction.
在本发明实施例中， 要生成上采样滤波器和下采样滤波器， 首先需 要对符合马尔科夫模型的视频信号进行下采样和上采样处理， 得到重建 信号， 然后计算出重建信号的能量； 再根据带参数的双正交小波理论和 所述重建信号的能量， 计算出所述重建信号的能量的最大值所对应的小 波滤波器参数， 生成上采样滤波器和下采样滤波器。 In the embodiment of the present invention, to generate an upsampling filter and a downsampling filter, first, a video signal conforming to the Markov model needs to be downsampled and upsampled to obtain a reconstructed signal, and then the energy of the reconstructed signal is calculated; Then, according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, the wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal are calculated, and an upsampling filter and a downsampling filter are generated.
图 2为本发明实施例的生成上采样滤波器和下采样滤波器的方法的 流程示意图。 如图 2所示， 该方法包括如下步骤： 2 is a flow chart showing a method of generating an upsampling filter and a downsampling filter according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps:
步骤 21 :对符合马尔科夫模型的视频信号进行下采样和上采样处理， 得到重建信号。 Step 21: Down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal.
步骤 22: 计算出重建信号的能量。 Step 22: Calculate the energy of the reconstructed signal.
步骤 23: 根据带参数的双正交小波理论和重建信号的能量， 计算出 重建信号的最大值所对应的小波滤波器参数， 生成上采样滤波器和下采 样滤波器。 Step 23: According to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, calculate the wavelet filter parameters corresponding to the maximum value of the reconstructed signal, and generate an upsampling filter and a downsampling filter.
步骤 24: 对上采样滤波器和下采样滤波器进行整数化处理， 得到整 数上采样滤波器和整数下采样滤波器。 Step 24: Integring the upsampling filter and the downsampling filter to obtain an integer upsampling filter and an integer downsampling filter.
步骤 25: 对整数上采样滤波器和整数下采样滤波器进行长度裁减， 得到优化整数上采样滤波器和优化整数下采样滤波器。 Step 25: Perform length reduction on the integer upsampling filter and the integer downsampling filter to obtain an optimized integer upsampling filter and an optimized integer downsampling filter.
在实际操作中， 上述步骤 24和步骤 25属于可选步骤。 In practice, steps 24 and 25 above are optional steps.
其中， 上述步骤 21〜步骤 25的具体操作为： The specific operations of the foregoing steps 21 to 25 are as follows:
在步骤 21中， 对符合马尔科夫模型的视频信号进行下采样处理和上
采样处理， 得到重建信号的方法包括： In step 21, the video signal conforming to the Markov model is subjected to downsampling processing and The sampling process, the method for obtaining the reconstructed signal includes:
S = {sn}0≤n<M是一个平稳随机视频信号序列， H(1) =
S = {s n } 0 ≤ n < M is a sequence of stationary random video signals, H (1 ) =
和 ^2) ^2)。，/^，…，^2^是两个有限线性相位脉沖响应（FIR)低通滤 波器， 且均为因果滤波器。 And ^ 2 ) ^ 2 ). , /^,...,^ 2 ^ are two finite linear phase impulse response (FIR) low-pass filters, both of which are causal filters.
首先，用 H(1)对 ^ = k}。≤„<M进行滤波处理，再进行下采样处理后得到： First, use H (1 ) for ^ = k}. ≤ „ <M performs filtering processing, and then performs downsampling processing to obtain:
然后， 对 信号序列进行上采样处理，
Then, the signal sequence is upsampled,
并用 H(2)对该信号序列进行滤波， 得到重建信号： And filtering the signal sequence with H( 2 ) to obtain a reconstructed signal:
Sl = {sp l \sp l =∑h^p_2kSk 0}(0<p<M) 7 其中， =∑^),— 2Α·。 S l = {s p l \s p l =∑h^ p _ 2kSk 0 }(0<p<M) 7 where, =∑^), — 2 Α·.
k j k j
在实际应用中， 95%的视频信号都是符合马尔科夫模型的视频信号， 将上述对视频信号进行下采样处理和上采样处理， 得到重建信号的过程 用流程示意图的形式表示出来。 In practical applications, 95% of the video signals are video signals conforming to the Markov model, and the process of downsampling and upsampling the video signals to obtain the reconstructed signals is represented in the form of a flow diagram.
图 2a为本发明实施例的实现视频信号重建的流程示意图。 如图 2a所 示, 2a is a schematic flowchart of implementing video signal reconstruction according to an embodiment of the present invention. As shown in Figure 2a,
使用 ω对视频信号 S进行滤波处理， 再经下采样处理得到信号 S0。 然后对信号 S。进行上采样处理， 使用 (2)滤波器对经上采样处理的 S。进 行滤波处理， 得到重建信号 Si。 The video signal S is subjected to filtering processing using ω , and then subjected to downsampling processing to obtain a signal S 0 . Then on the signal S. The upsampling process is performed, and the S(S) processed by the upsampling is used by the ( 2 ) filter. A filtering process is performed to obtain a reconstructed signal Si.
在步骤 22中， 根据重建信号计算得到重建信号的能量的方法为： 在本步骤中， 通过将所述重建信号 入能量模型， 计算所述重建 信号的能量， 具体包括： In the step 22, the method for calculating the energy of the reconstructed signal according to the reconstructed signal is: In this step, calculating the energy of the reconstructed signal by using the reconstructed signal into the energy model, specifically:
首先， 选择一个能量模型： maxi I A I +
First, choose an energy model: maxi IAI +
其中， Si是对视频信号 S进行下采样和上采样处理得到的重建信号； S2 是将所述视频信号 S与所述 SI进行相关得到的视频信号； S3是对所述 S进
行下采样得到的视频信号。 是进行下采样和上采样处理后的视频 信号的能量。 是进行下采样和上采样处理前的视频信号和所述重 建信号的相关能量。 是进行下采样处理后的视频信号的能量。 所述 α和 β均为大于零的正整数， γ为大于等于零的正整数。 Wherein, Si is a reconstructed signal obtained by down-sampling and up-sampling the video signal S; S 2 is a video signal obtained by correlating the video signal S with the SI; and S 3 is for the S Line down the resulting video signal. It is the energy of the video signal after downsampling and upsampling. Is the correlation energy of the video signal before the downsampling and upsampling process and the reconstructed signal. It is the energy of the video signal after the downsampling process. Both α and β are positive integers greater than zero, and γ is a positive integer greater than or equal to zero.
当 α、 β和 γ取不同值时可以得到不同的能量模型， 例如： Different energy models can be obtained when α, β and γ take different values, for example:
当 α=1, β=λ, γ=0时， 能量模型为： El = max{E[e(W] + ^[e( )]}; 当 α=1/2, β=1, γ=0时， 能量模型为： 2 = max{E[(l/2)e(W] + [e(^)]}; 当 α=1/2, β=λ, γ=0时，能量模型为： 3 = max{E[(l/2 (W] + ； 当 α=1/2, β=λ, γ=θ时， 能量模型为： When α = 1 , β = λ , γ = 0 , the energy model is: El = m ax {E[ e (W] + ^[e( )]} ; when α = 1/2, β = 1, γ When =0, the energy model is: 2 = m ax {E[(l/2) e (W] + [e(^)]} ; when α=1/2, β=λ, γ=0, energy The model is: 3 = m ax {E[(l/2 (W) + ; When α=1/2, β=λ, γ=θ, the energy model is:
Ε4 = max {E[(l 12)e(Sl )] + AE[e(S2 )] + 0E[e(S3 )] } 在本实施例中， 选择的能量模型为： ^ maxi ^d)] ^?^^)]^ 进行下采样和上采样处理后的视频信号的能量为：
Ε4 = max {E[(l 12)e(S l )] + AE[e(S 2 )] + 0E[e(S 3 )] } In this embodiment, the selected energy model is: ^ maxi ^ d)] ^?^^)]^ The energy of the video signal after downsampling and upsampling is:
进行下采样和上采样处理前的视频信号和所述重建信号的相关能量 The correlation between the video signal before the downsampling and the upsampling process and the reconstructed signal
E } = E } =
为：
for:
因此得到的重建信号的能量为： The energy of the reconstructed signal thus obtained is:
EieiS, )} + E{e(S2 )} = E{∑ (sp l f } + E{∑ sp l +lsp } 其中， E[e( )]和 E[e(S 2 )]可以分别近似为： E{e(S } - 2M ¾ h(2) ph(2) pi ^ ^ ( 1" "21 EieiS, )} + E{e(S 2 )} = E{∑ (s p l f } + E{∑ s p l +l s p } where E[e( )] and E[e(S 2 )] can be approximated as: E{e(S } - 2M 3⁄4 h (2) p h (2) pi ^ ^ ( 1 "" 21
E{e(S2)} = E{∑ sp l +lsp} ¾ ∑∑Λ(2 (
在步骤 23中， 在本实施例中， 带参数双正交小波理论为带参数的双 正交小波滤波器参数化公式。 根据带参数的双正交小波滤波器参数化公 式和重建信号的能量， 计算出重建信号的最大值所对应的小波滤波器参 数， 生成上采样滤波器和下采样滤波器的方法包括： E {e (S 2)} = E {Σ s p l + l s p} ¾ ΣΣΛ (2 ( In step 23, in the present embodiment, the parameterized bi-orthogonal wavelet theory is a parameterized formula of a bi-orthogonal wavelet filter with parameters. According to the parameterization formula of the bi-orthogonal wavelet filter with parameters and the energy of the reconstructed signal, the wavelet filter parameters corresponding to the maximum value of the reconstructed signal are calculated, and the methods for generating the upsampling filter and the downsampling filter include:
由步骤 22可知， 在本实施例中的重建信号的能量公式为： It can be seen from step 22 that the energy formula of the reconstructed signal in this embodiment is:
EieiS, )} + E{e(S2 )} = E{∑ (sp l f } + E{∑
} o 在本实施例中， 采用长度为 EieiS, )} + E{e(S 2 )} = E{∑ (s p l f } + E{∑ } o In this embodiment, the length is
(7, 13) 的单参数双正交小波参数化公式。 (7, 13) Single parameter bi-orthogonal wavelet parameterization formula.
长度为 （7, 13) 的单参数双正交小波参数化公式如下： = (-丄 +丄 -丄 +丄 7ϊ,丄 + I, 丄 + 7ϊ, -丄 +丄 ϊ, -丄 +丄 I) The single parameter bi-orthogonal wavelet parameterization formula of length (7, 13) is as follows: = (-丄+丄-丄+丄7ϊ,丄+ I, 丄+ 7ϊ, -丄+丄ϊ, -丄+丄I )
4 32 2 4 4 32 4 32 2 4 4 32 4 32 2 4 4 32 4 32 2 4 4 32
L =— (-157-352Λ>/2 +13552Λ2 +1048576Λ9>/2 +4554752Λ7>/2 L =—(-157-352Λ>/2 +13552Λ 2 +1048576Λ 9 >/2 +4554752Λ 7 >/2
0 64 0 64
+993792Λ/2Λ5 -25784Λ3Λ/2 -162624Λ4 -4225024Λ6 -4849664Λ8) +993792Λ/2Λ 5 -25784Λ 3 Λ/2 -162624Λ 4 -4225024Λ 6 -4849664Λ 8 )
/(-8Λ2 + 4hf2 - 1)(V2 - 256Λ3 +20Λ + 16h2 ¾(2 - 1); /(-8Λ 2 + 4hf2 - 1)(V2 - 256Λ 3 +20Λ + 16h 2 3⁄4(2 - 1);
h = 68 + 6120Λ2 + 262144/i9V2 +1298432Λ7Λ/2h = 68 + 6120Λ 2 + 262144/i 9 V2 +1298432Λ 7 Λ/2
+408768V2/i5 -10160Λ3Λ/2 -89856Λ4 -1413632Λ6 -1277952Λ8) +408768V2/i 5 -10160Λ 3 Λ/2 -89856Λ 4 -1413632Λ 6 -1277952Λ 8 )
/(-8Λ2 + 4 - 1)(Λ/Ϊ - 256Λ3 +20A + 16h2 ¾(2 - 1);
/(-8Λ 2 + 4 - 1)(Λ/Ϊ - 256Λ 3 +20A + 16h 2 3⁄4(2 - 1);
+1500672V2/i5 - 19904Λ3 2 - 446784Λ4 - 4823040Λ6 - 4849664Λ8 )
+1500672V2/i 5 - 19904Λ 3 2 - 446784Λ 4 - 4823040Λ 6 - 4849664Λ 8 )
+1123392 2¾5 -30160Λ3>/2 +16008Λ2 -223168Λ4 -4095488Λ6) +1123392 23⁄4 5 -30160Λ 3 >/2 +16008Λ 2 -223168Λ 4 -4095488Λ 6 )
/(-8Λ2 +
+20Λ + 16h2 ¾(2 - 1)； /(-8Λ 2 + +20Λ + 16h 2 3⁄4(2 - 1);
h, = -"— (-1808Λ + 125>/2 -12864Λ3 - 4096Λ5 + 4936Λ2 V2 + 8192h4 Ϊ)
£ =— (-55-219h l + 4172h2 -1496Λ3 Ϊ - 750 4 h, = -"- (-1808Λ + 125>/2 -12864Λ 3 - 4096Λ 5 + 4936Λ 2 V2 + 8192h 4 Ϊ) £ =— (-55-219h l + 4172h 2 -1496Λ 3 Ϊ - 750 4
+209024Λ/2Λ5 + 286720Λ7 Λ/Ϊ - 497664Λ6 -131072k8 ) +209024Λ/2Λ 5 + 286720Λ 7 Λ/Ϊ - 497664Λ 6 -131072k 8 )
/(-8k2 + 4ky/2 - 1)(V2 - 256k3 +20k + l 6k2 ¾; /(-8k 2 + 4ky/2 - 1)(V2 - 256k 3 +20k + l 6k 2 3⁄4;
=丄 (-4560896Λ6 - 4849664Λ8 -165 + 1048576Λ9 Ϊ =丄(-4560896Λ 6 - 4849664Λ 8 -165 + 1048576Λ 9 Ϊ
6 256 6 256
+1156608>/2Λ5 +4620288/i7V2 -272ΛΛ/2 -34240Λ3>/2 -196928Λ4 +15472Λ2) +1156608>/2Λ 5 +4620288/i 7 V2 -272ΛΛ/2 -34240Λ 3 >/2 -196928Λ 4 +15472Λ 2 )
/(-8Λ2 + h l - 1)(V2 - 256Λ3 +20Λ + 16h2
/(-8Λ 2 + hl - 1)(V2 - 256Λ 3 +20Λ + 16h 2
需要指出的是， 可以采用的带参数的双正交小波滤波器参数化公式 可以包括但不限于： 单参数双正交小波参数化公式， 或双参数双正交小 波参数化公式， 或多参数双正交小波参数化公式， 或单参数和多参数双 正交小波参数化公式。 利用双参数双正交小波参数化公式、 多参数双正 交小波参数化公式， 以及单参数和多参数双正交小波参数化公式求取上 采样滤波器和下采样滤波器的方法与利用单参数双正交 d、波参数化公 式求取上采样滤波器和下采样滤波器的方法相同， 也在本发明的保护范 围之内。 It should be noted that the parameterized formula of the parameterized bi-orthogonal wavelet filter may include, but is not limited to: a single parameter double orthogonal wavelet parameterization formula, or a two parameter double orthogonal wavelet parameterization formula, or multiple parameters. Biorthogonal wavelet parameterization formula, or single parameter and multi-parameter bi-orthogonal wavelet parameterization formula. Method and utilization list for obtaining up-sampling filter and down-sampling filter by using two-parameter bi-orthogonal wavelet parameterization formula, multi-parameter bi-orthogonal wavelet parameterization formula, and single-parameter and multi-parameter bi-orthogonal wavelet parameterization formula The parameter bi-orthogonal d, the wave parameterization formula is the same as the method of obtaining the up-sampling filter and the down-sampling filter, and is also within the protection scope of the present invention.
求解重建信号取得最大值时所对应的小波滤波器参数的方法为： 首先， 将带参数的双正交小波滤波器参数化公式代入重建信号的能 量公式， 即将长度为 （7, 13) 的单参数双正交小波参数化公式代入到 E{e(S )} + )} = E{ ( )2 } + E{ + ^ }公式中。 其次，对 λ=1和 λ=0.1分别进行 60次迭代计算，计算得出， 当 h=0.7时， 重建信号的能量处于最大值附近。 The method for solving the wavelet filter parameters corresponding to the maximum value of the reconstructed signal is as follows: First, the parameterized formula of the bi-orthogonal wavelet filter with parameters is substituted into the energy formula of the reconstructed signal, that is, the single length (7, 13) The parameter bi-orthogonal wavelet parameterization formula is substituted into the E{e(S )} + )} = E{ ( ) 2 } + E{ + ^ } formula. Secondly, 60 iterations are calculated for λ=1 and λ=0.1, respectively. It is calculated that when h=0.7, the energy of the reconstructed signal is near the maximum value.
再次， 将 h=0.7代入单参数双正交小波参数化公式得到如表 1所示的 值， 表 1为小波滤波器参数值。 Again, substituting h=0.7 into the one-parameter bi-orthogonal wavelet parameterization formula yields the values shown in Table 1, and Table 1 shows the wavelet filter parameter values.
H7 H13 H7 H13
_.4241747852752234e-l -.2864426763622237e-2
.3553390593273753e-2 -.2399583254419701e-3 _.4241747852752234e-l -.2864426763622237e-2 .3553390593273753e-2 -.2399583254419701e-3
.3959708691207962 .5084782097542918e- l .3959708691207962 .5084782097542918e- l
.7000000000000000 -.4525099414445236e- l .7000000000000000 -.4525099414445236e- l
.3959708691207962 -.1811648336223731 .3959708691207962 -.1811648336223731
.3553390593273753e-2 .3990443430631642 .3553390593273753e-2 .3990443430631642
_.4241747852752234e- l .9734696600076678 _.4241747852752234e- l .9734696600076678
.3990443430631642 .3990443430631642
-.1811648336223731 -.1811648336223731
-.4525099414445236e- l -.4525099414445236e- l
.5084782097542918e- l .5084782097542918e- l
-.2399583254419701e-3 -.2399583254419701e-3
-.2864426763622237e-2 表 1 -.2864426763622237e-2 Table 1
最后，将表 1中浮点形式的小波滤波器参数分别写成矩阵的形式， 即 可生成上采样滤波器和下采样滤波器。 Finally, the wavelet filter parameters in the floating point form in Table 1 are written in the form of a matrix, that is, an upsampling filter and a downsampling filter can be generated.
在步骤 24中， 可以对步骤 23生成的上采样滤波器和下采样滤波器进 行整数化处理， 生成整数上采样滤波器和整数下采样滤波器的方法为： 对表 1中的浮点形式的小波滤波器作二进制乘法，然后对二进制乘法 的结果取近似值， 可以生成的整数上采样滤波器和整数下采样滤波器分 别为： In step 24, the upsampling filter and the downsampling filter generated in step 23 may be integerized, and the method of generating the integer upsampling filter and the integer downsampling filter is as follows: The wavelet filter is used for binary multiplication, and then the approximate result of the binary multiplication is obtained. The integer upsampling filter and the integer downsampling filter that can be generated are:
H7 =[-1 0 9 16 9 0 -1] /16和 H9=[ l -1 -4 9 22 9 -4 -1 1 ]/32 , H7 = [-1 0 9 16 9 0 -1] /16 and H9=[ l -1 -4 9 22 9 -4 -1 1 ]/32 ,
或者， HH7=[-8 1 72 126 72 1 -8]/128和 HH13=[-1 0 18 -16 -65 144 352 144 -65 -16 18 0 -1]/512； Or, HH7=[-8 1 72 126 72 1 -8]/128 and HH13=[-1 0 18 -16 -65 144 352 144 -65 -16 18 0 -1]/512;
或， H7 = [-5 0 21 32 21 0 -5]/32和 H3 = [17 30 17]/64。 Or, H7 = [-5 0 21 32 21 0 -5]/32 and H3 = [17 30 17]/64.
在步骤 25中， 可以对步骤 24生成的整数上采样滤波器和整数下采样
滤波器进行长度裁减， 生成优化整数上采样滤波器和优化整数下采样滤 波器的方法为： In step 25, the integer upsampling filter and integer downsampling generated in step 24 may be performed. The filter is length-reduced to generate an optimized integer upsampling filter and an optimized integer downsampling filter:
滤波器的长度直接影响到其运算速度和视频信号的重建质量， 在对 运算速度和视频信号重建性能之间进行权衡地基础上， 将整数上采样滤 波器和整数下采样滤波器两端影响比较小的值省略， 并累加到中心值 上， 生成优化整数上采样滤波器和优化整数下采样滤波器。 The length of the filter directly affects the speed of its operation and the reconstruction quality of the video signal. Based on the trade-off between the operation speed and the reconstruction performance of the video signal, the effects of the integer up-sampling filter and the integer down-sampling filter are compared. Small values are omitted and added to the center value to generate an optimized integer upsampling filter and an optimized integer downsampling filter.
对步骤 24中的 H7 =[-l 0 9 16 9 0 -1] /16^H9=[1 -1 -4 9 22 9 -4 -1 1 ]/32进行长度裁减，生成的优化整数上采样滤波器和优化整数下采样滤 波器分别为： H7 =[-l 0 9 16 9 0 -1] /16； H5=[-4 9 22 9 -4]/32。 Perform length reduction on H7 =[-l 0 9 16 9 0 -1] /16^H9=[1 -1 -4 9 22 9 -4 -1 1 ]/32 in step 24, and generate optimized integer upsampling The filter and the optimized integer downsampling filter are: H7 = [-l 0 9 16 9 0 -1] /16; H5 = [-4 9 22 9 -4] / 32.
需要指出的是， 在本实施例中可以利用步骤 23中的其它能量模型和 带参数的双正交小波滤波器参数化公式对视频信号的色度分量或亮度 分量的上采样滤波器和下采样滤波器进行设计和优化处理， 例如， 可以 生成另外一组上采样滤波器和下采样滤波器： H7 = [-5, 0, 21, 32, 21, 0, -5]/64和 H3 = [17, 30, 17]/64。 It should be noted that, in this embodiment, the up-sampling filter and the downsampling of the chrominance component or the luminance component of the video signal may be utilized by using other energy models in step 23 and a parameterized bi-orthogonal wavelet filter parameterization formula. The filter is designed and optimized. For example, another set of upsampling filters and downsampling filters can be generated: H7 = [-5, 0, 21, 32, 21, 0, -5]/64 and H3 = [ 17, 30, 17]/64.
执行步骤 21〜步骤 23 即可实现本发明的上采样滤波器和下采样滤波 器的设计， 进一步地执行步骤 24可以对步骤 23生成的浮点形式的滤波器 进行整数化处理， 生成整数滤波器， 同样进一步地执行步骤 25可以对步 骤 24生成的整数滤波器进行长度优化处理， 生成优化整数上采样滤波器 和优化整数下采样滤波器。 The steps of the upsampling filter and the downsampling filter of the present invention can be implemented by performing steps 21 to 23. Further, step 24 can be used to perform integer processing on the floating point filter generated in step 23 to generate an integer filter. Similarly, step 25 may be further performed to perform length optimization processing on the integer filter generated in step 24 to generate an optimized integer upsampling filter and an optimized integer downsampling filter.
从本实施例可以看出，本实施例主要介绍了基于 H.264的空间可分级 编码过程中上采样滤波器和下采样滤波器的设计。 与现有技术相比， 具 有: ¾口下优点： As can be seen from the embodiment, this embodiment mainly introduces the design of the upsampling filter and the downsampling filter in the spatial scalable coding process based on H.264. Compared with the prior art, it has the following advantages: 3⁄4 mouth:
首先， 将带参数的双正交小波理论用于上采样滤波器和下采样滤波 器的设计， 从而使设计出的上采样滤波器和下采样滤波器具有关联性。 First, the bi-orthogonal wavelet theory with parameters is used in the design of the upsampling filter and the downsampling filter, so that the designed upsampling filter and downsampling filter are related.
其次， 本实施例生成的上采样滤波器和下采样滤波器是对能量模型
求解最大化的情况下生成的上、 下采样滤波器， 在保证同等滤波器长度 中针对能量模型是最优的。 Secondly, the upsampling filter and the downsampling filter generated in this embodiment are energy models. The upper and lower sampling filters generated in the case of maximizing the solution are optimal for the energy model in ensuring the same filter length.
最后， 根据不同视频信号编码复杂度的要求， 可以对生成的整数滤 波器进行不同长度的裁剪， 生成优化整数滤波器， 从而通过在运算速度 和视频信号重建性能之间的权衡， 生成最符合实际需要的优化整数上采 样滤波器和优化整数下采样滤波器。 Finally, according to the coding complexity of different video signals, the generated integer filters can be clipped to different lengths to generate an optimized integer filter, so that the trade-off between the operation speed and the video signal reconstruction performance is the most realistic. Optimized integer upsampling filters and optimized integer downsampling filters are required.
至此， 结束对图 2所示实施例的介绍。 So far, the introduction of the embodiment shown in Fig. 2 is ended.
图 2所示的实施例给出了如何生成上采样滤波器和下采样滤波器 的方法， 可以将设计生成的上采样滤波器和下采样滤波器用于空间可分 级视频编码操作。 空间可分级视频信号编码过程分为： 利用下采样滤波 器进行的编码过程和利用上采样滤波器进行的编码过程。 下面分别以具 体的实施例予以介绍： The embodiment shown in Figure 2 shows how to generate an upsampling filter and a downsampling filter, which can be used for spatially scalable video encoding operations. The spatially scalable video signal encoding process is divided into: an encoding process using a downsampling filter and an encoding process using an upsampling filter. The following are introduced in specific embodiments:
图 3为利用图 2所示实施例生成的下采样滤波器实现空间可分级视频 编码的方法的流程示意图。 如图 3所示， 该方法包括以下步骤： FIG. 3 is a schematic flow chart of a method for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. As shown in FIG. 3, the method includes the following steps:
步骤 31 : 利用下采样滤波器对作为编码输入的视频信号进行下采样 处理， 得到下采样视频序列。 Step 31: Down-sampling the video signal as the encoded input by using a downsampling filter to obtain a downsampled video sequence.
在本实施例中， 使用的下采样滤波器是步骤 24中所生成的下采样滤 波器： H9=[l - 4 9 22 9 -4 -l l ]/32。 In the present embodiment, the downsampling filter used is the downsampling filter generated in step 24: H9 = [l - 4 9 22 9 -4 - l l ] / 32.
步骤 32: 对视频信号序列进行时域分解，得到内部（I )帧、预测（P ) 帧和双向预测 （B ) 帧。 Step 32: Perform time domain decomposition on the video signal sequence to obtain an internal (I) frame, a prediction (P) frame, and a bidirectional prediction (B) frame.
步骤 33:对 I帧进行帧内预测和帧内编码得到下采样帧内预测编码信 号， 对 P帧和 B帧进行运动编码， 得到运动预测编码信号。 Step 33: Perform intra prediction and intra coding on the I frame to obtain a downsampled intra prediction coding signal, and perform motion coding on the P frame and the B frame to obtain a motion prediction coded signal.
上述步骤 31〜步骤 33的具体操作为： The specific operations of steps 31 to 33 above are as follows:
在步骤 31中， 利用下采样滤波器对作为编码输入的视频信号进行下
采样处理， 得到下采样视频序列的具体操作为： In step 31, the video signal as the encoded input is subjected to the downsampling filter. The sampling operation, the specific operation of obtaining the downsampled video sequence is:
将作为编码输入的视频信号中同行或隔行中的各像素与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行垂直方向上的卷积计算， 得到各像素的像素值， 然后将各像素的像素值与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行水平方向上 的卷积计算， 得到下采样视频序列； 或者， 也可以先进行水平方向上的 卷积运算得到各像素的像素值，然后再将各像素的像素值与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行垂直方向上的卷积计算， 得到下采样视频序。 针 对上述过程， 下面以一个具体的例子予以介绍： Convolution calculation of each pixel in the peer or interlaced line of the video signal as the code input with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32 in the vertical direction to obtain the pixel value of each pixel Then, the pixel value of each pixel is convoluted with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32 in the horizontal direction to obtain a downsampled video sequence; or, the level may be first The convolution operation in the direction obtains the pixel value of each pixel, and then calculates the pixel value of each pixel in a vertical direction with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32. Get the downsampled video sequence. For the above process, the following is a specific example:
图 3a为视频信号帧中全部像素的结构示意图。 图 3a中包含了： 包 括像素 E、 像素 e、 像素 F、 像素 f、 像素 G、 像素 g、 像素 H、 像素 h 和像素 I的行； 包括像素 像素 K、 像素 L、 像素 k、 像素 M和像素 N 的行； 包括像素八、 像素 a、 像素 C、 像素 b、 像素 G、 像素 c、 像素 L、 像素 d和像素 P的列； 以及包括像素 B、 像素 D、 像素 H、 像素 n、 像 素 M和像素 Q的列。 FIG. 3a is a schematic structural diagram of all pixels in a video signal frame. Figure 3a includes: a row including pixel E, pixel e, pixel F, pixel f, pixel G, pixel g, pixel H, pixel h, and pixel I; including pixel pixel K, pixel L, pixel k, pixel M, and a row of pixels N; a column including pixel eight, pixel a, pixel C, pixel b, pixel G, pixel c, pixel L, pixel d, and pixel P; and includes pixel B, pixel D, pixel H, pixel n, pixel M and the column of pixel Q.
以像素 G为例， 第一步，将 G点的像素值以及垂直方向上的各像素值 与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行卷积运算， 得到像素值 G，= Taking the pixel G as an example, in the first step, the pixel value of the G point and the pixel value in the vertical direction are convoluted with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32 to obtain a convolution operation. Pixel value G, =
( A-a-4*C+9*b+22*G+9*c-4*L-d+P ) /32。 同样对于像素 G同行或隔行中 的各像素也按此公式与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行卷积运算，分别 得到像素值 E'、 像素值 e'、 像素值 F'、 像素值 f'、 像素值 G'、 像素 值 g'、 像素值 H'、 像素值 h'和像素值 Γ等。 (A-a-4*C+9*b+22*G+9*c-4*L-d+P) /32. Similarly, for each pixel in the pixel G peer or interlaced line, convolution operation is performed according to this formula with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32 to obtain the pixel value E' and the pixel value e, respectively. ', pixel value F', pixel value f', pixel value G', pixel value g', pixel value H', pixel value h', pixel value Γ, and the like.
第二步， 将像素值 G，以及水平方向上的各像素值与 H9=[l -1 -4 9 22 9 -4 -1 1 ]/32进行卷积运算， 得到 G"= In the second step, the pixel value G and the pixel values in the horizontal direction are convoluted with H9=[l -1 -4 9 22 9 -4 -1 1 ]/32 to obtain G"=
( E'-e'-4*F'+9*f'+22*G'+9*g'-4*H'-h'+r ) /32, 其中等式右侧的各像素 值是通过第一步的卷积运算得到的像素值。 ( E'-e'-4*F'+9*f'+22*G'+9*g'-4*H'-h'+r ) /32, where the pixel values on the right side of the equation are The pixel value obtained by the convolution operation of the first step.
将图 3a中灰色方格中的像素值单独取出，即可得到下采样视频序列。
需要指出的是上述第一步和第二步的操作可以倒换次序， 得到的结果相 上与下采样滤波器进行卷积运算。 The pixel values in the gray squares in Figure 3a are taken separately to obtain a downsampled video sequence. It should be noted that the operations of the first step and the second step described above can be reversed, and the obtained result is convoluted with the downsampling filter.
步骤 32中对视频信号序列进行时域分解，得到 I帧、 P帧和 B帧的操作， 以及步骤 33中对 I帧进行帧内预测和帧内编码得到下采样帧内预测编码 信号， 对 P帧和 B帧进行运动编码， 得到运动预测编码信号操作， 可以与 现有技术中的相应操作相同。 In step 32, the video signal sequence is subjected to time domain decomposition to obtain an operation of an I frame, a P frame, and a B frame, and in step 33, the I frame is subjected to intra prediction and intra coding to obtain a downsampled intra prediction encoded signal. The frame and the B frame are motion-encoded to obtain a motion prediction coded signal operation, which can be the same as the corresponding operation in the prior art.
至此， 结束对图 3所示实施例的介绍。 So far, the introduction of the embodiment shown in Fig. 3 is ended.
图 4为利用图 2所示实施例生成的上采样滤波器实现空间可分级视频 编码的方法的流程示意图。 如图 4所示， 该方法包括以下步骤： 4 is a flow chart showing a method for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. 2. As shown in Figure 4, the method includes the following steps:
步骤 41： 使用上采样滤波器对下采样帧内预测编码信号进行上采样 处理， 得到上采样视频序列。 Step 41: Upsampling the downsampled intra prediction encoded signal by using an upsampling filter to obtain an upsampled video sequence.
在本步骤中， 所使用的上采样滤波器为步骤 24中所生成的整数上采 样滤波器： H7 =[-1 0 9 16 9 0 -1] /16,所涉及的下采样帧内预测编码信号 是步骤 33中对 I帧进行帧内预测和帧内编码得到下采样帧内预测编码信 号。 In this step, the upsampling filter used is the integer upsampling filter generated in step 24: H7 = [-1 0 9 16 9 0 -1] /16, the downsampling intra prediction coding involved The signal is an intra-frame prediction and intra-frame coding of the I frame in step 33 to obtain a downsampled intra-predicted coded signal.
步骤 42: 对上采样视频序列进行帧内预测和帧内编码， 得到上采样 帧内预测编码信号。 Step 42: Perform intra prediction and intra coding on the upsampled video sequence to obtain an upsampled intra prediction encoded signal.
其中， 步骤 41〜步骤 42的具体操作为： The specific operations of step 41 to step 42 are as follows:
在步骤 41中， 使用上采样滤波器对下采样帧内预测编码信号进行上 采样处理， 得到上采样视频序列得具体方法为： 使用上采样滤波器对下 采样帧内预测编码信号进行层间帧内纹理预测和层间残差预测的 2D空 域上采样处理， 得到上采样视频序列。 In step 41, the upsampling encoded signal is upsampled using an upsampling filter to obtain an upsampled video sequence. The specific method is: using an upsampling filter to perform inter-frame processing on the downsampled intraframe encoded signal. The 2D spatial domain upsampling process of inner texture prediction and inter-layer residual prediction yields an upsampled video sequence.
仍参见图 3a, 图 3a中灰色方格表示下采样帧内预测编码信号中宏块 的像素， 所有的方格表示上采样视频序列中宏块的像素。 使用上采样滤
波器对下采样帧内预测编码信号进行层间帧内纹理预测和层间残差预 测的 2D空域上采样处理， 得到上采样视频序列具体方法为： Still referring to Fig. 3a, the gray squares in Fig. 3a represent the pixels of the macroblock in the downsampled intra prediction encoded signal, all of which represent the pixels of the macroblock in the upsampled video sequence. Use upsampling filter The waver performs intra-layer intratextural prediction and inter-layer residual prediction on the downsampled intra prediction encoded signal by 2D spatial domain upsampling, and the specific method for obtaining the upsampled video sequence is:
对于灰色方格位置的像素： 保持其像素值不变。 For pixels in gray square positions: Keep their pixel values unchanged.
对于在水平方向上紧邻灰色方格，形如 g的像素： 将下采样帧内预测 编码信号序列隔列补零后按水平方向与 H7 =[-1 0 9 16 9 0 -1] /16做卷积， 得到 g的像素值为： g= ( -F+9*G+9*H-i ) /16。 此外， 对于其他与 g在水平 方向上和垂直方向上相差偶数个像素的像素也按同样的方法进行处理。 For a pixel in the horizontal direction immediately adjacent to the gray square, shaped like g: the subsampled intra prediction encoded signal sequence is padded with zeros and then horizontally aligned with H7 = [-1 0 9 16 9 0 -1] /16 Convolution, the pixel value of g is: g = ( -F+9*G+9*Hi ) /16. In addition, pixels which are evenly spaced apart from g in the horizontal direction and the vertical direction are processed in the same manner.
对于在垂直方向上紧邻灰色方格， 形如 c的像素： 将下采样帧内预测 编码信号序列隔行补零后按垂直方向与 H7 =[-1 0 9 16 9 0 -1] /16做卷积， 得到 c的像素值为: c= ( -C+9*G+9*L-P ) /16ο 此外， 对于其他与 c在水 平方向上和垂直方向上相差偶数个像素的像素也按同样的方法进行处 理。 For a pixel in the vertical direction immediately adjacent to the gray square, shaped like c: The subsampled intra prediction encoded signal sequence is padded with zeros and then scrolled in the vertical direction with H7 = [-1 0 9 16 9 0 -1] /16 Product, the pixel value of c is obtained: c= ( -C+9*G+9*LP ) /16ο In addition, for other pixels that are different from c in the horizontal direction and the vertical direction by an even number of pixels, the same method is used. Process it.
对于与灰色方格处于斜对角， 形如 i的像素： 将下采样帧内预测编码 信号序列按垂直方向或水平方向与 H7 =[-1 0 9 16 9 0 -1] /16做卷积，得到 i的像素值1= ( -m+9*c+9*n-o ) /16或1= ( -j+9*g+9*k-l ) /16, 这两个值是 相等的， 取其一即可。 For pixels diagonally diagonal to the gray square, shaped like i: Convolute the downsampled intra-predictive coded signal sequence with H7 = [-1 0 9 16 9 0 -1] /16 in the vertical or horizontal direction , get the pixel value of i 1 = ( -m+9*c+9*no ) /16 or 1= ( -j+9*g+9*kl ) /16, these two values are equal, take it One can be.
在步骤 42中， 利用步骤 41得到的上采样视频序列和原有的纹理信息 在增强层进行帧内预测和变换 /熵编码， 形成上采样帧内预测编码信号。 In step 42, the upsampled intra-predictive coded signal is formed by performing intra prediction and transform/entropy coding on the enhancement layer using the upsampled video sequence obtained in step 41 and the original texture information.
最后， 将全部编码过程得到的下采样帧内预测编码信号、 运动预测 编码信号和上采样帧内预测编码信号进行复用， 得到重建视频信号。 Finally, the downsampled intra prediction encoded signal, the motion prediction encoded signal and the upsampled intra prediction encoded signal obtained by all encoding processes are multiplexed to obtain a reconstructed video signal.
其中， 本步骤的操作可以与现有技术中的相应操作类似。 The operation of this step can be similar to the corresponding operation in the prior art.
至此， 结束对图 4所示实施例的介绍。 So far, the introduction of the embodiment shown in Fig. 4 is ended.
图 2所示实施例中生成的上采样滤波器和下采样滤波器针对能量模 型是最优的， 且是相互关联的。 将图 2所示实施例生成的上采样滤波器 和下采样滤波器分别用于图 3和图 4所示视频信号的编码处理中， 可以减
少图像能量的损耗， 提高对视频信号进行编码的质量， 达到能量损失的 最小化和视频恢复质量的最大化， 且能够提高视频信号的重构质量， 优 化高分辨率下视频信号的重构效果。 The upsampling filter and downsampling filter generated in the embodiment shown in Figure 2 are optimal for the energy model and are interrelated. The upsampling filter and the downsampling filter generated by the embodiment shown in FIG. 2 are respectively used in the encoding process of the video signals shown in FIG. 3 and FIG. 4, and can be reduced. Less image energy loss, improve the quality of encoding video signals, minimize energy loss and maximize video recovery quality, and improve the reconstruction quality of video signals, optimize the reconstruction of video signals at high resolution .
使用图 2所示方法设计得到的上采样滤波器和下采样滤波器对图像 进行处理， 可以显著提高高分辨率的重建效果。 下面以具体的实施例介 绍对图像进行处理所得到的定量和定性测试结果。 The upsampling filter and downsampling filter designed using the method shown in Figure 2 process the image to significantly improve the resolution of high resolution. The quantitative and qualitative test results obtained by processing the image are described below in the specific examples.
表 2示出了对图像进行定量测试得到的结果。 Table 2 shows the results obtained by quantitatively testing the images.
如表 2所示， 列表示八幅标准图像的名称，行表示对图像进行测试的 滤波器， 依次为 SVC/H.264、 h5/h7、 h7/h9、 H7/H9和 HH7/HH13滤波器， As shown in Table 2, the columns represent the names of the eight standard images, and the rows represent the filters for testing the images, followed by SVC/H.264, h5/h7, h7/h9, H7/H9, and HH7/HH13 filters. ,
测试数据， 单位为峰值信噪比 （PSNR ), 其中， PSNR的值越大表示测 试结果越为理想。 The test data is expressed in units of peak signal-to-noise ratio (PSNR), where a larger value of PSNR indicates that the test result is more ideal.
通过比较图 2中的数值，可以明显看出采用后两组滤波器对图像进行 处理得到的 PSNR的值明显大于使用现有的滤波器对图像进行处理得到
的 PSNR的值。 因此， 可以得出这样的结论： 使用本发明方法设计得到 的上采样滤波器和下采样滤波器对图像进行处理， 可以显著提高高分辨 率的重建效果。 By comparing the values in Figure 2, it can be clearly seen that the PSNR value obtained by processing the image after using the latter two sets of filters is significantly larger than that obtained by using the existing filter. The value of PSNR. Therefore, it can be concluded that the image is processed by the upsampling filter and the downsampling filter designed by the method of the present invention, and the high-resolution reconstruction effect can be remarkably improved.
图 5为使用本发明生成的滤波器对图像进行处理得到的效果图。从图 理， 可以得到很清晰的视图效果。 Fig. 5 is a view showing an effect obtained by processing an image using a filter generated by the present invention. From the graphics, you can get a very clear view.
至此， 结束对图像的定量测试结果的介绍。 At this point, the introduction of the quantitative test results of the image is ended.
下面以具体的实施例介绍， 使用本发明生成的上下采样滤波器对图 像进行处理所得到的定性测试结果。 The qualitative test results obtained by processing the image using the up-and-down sampling filter generated by the present invention are described below with reference to specific embodiments.
图 6a是对运动物体（Mobile )进行层 0帧内编码所得的图形。 如图 6a 所示，层 0表示核心层，横坐标表示分辨率，取值范围为 500bps~3000bps; 纵坐标表示 PSNR , PSNR越大表示测试结果越为理想。 Fig. 6a is a diagram of layer 0 intraframe coding of a moving object (Mobile). As shown in Figure 6a, layer 0 represents the core layer, the abscissa represents the resolution, and the value ranges from 500bps to 3000bps; the ordinate represents PSNR, and the larger the PSNR, the more ideal the test result is.
图 6a中示出了 4条测试曲线，均是采用下采样滤波器和上采样滤波器 对层间帧内纹理预测中亮度分量进行优化处理后得到的曲线， 从上到下 依次为： 使用 our— Filter— 3— 7滤波器组、 Default JS VM滤波器组、 our— Filter— 5— 7滤波器组和 JVT-U 147— Filter— 5— 7滤波器组进行测试得到的 曲线， Default JSVM滤波器组和 JVT-U 147— Filter— 5— 7滤波器组的曲线用 虚线来表示。 Four test curves are shown in Fig. 6a, which are the curves obtained by optimizing the luminance components in the inter-layer intratextural prediction using the downsampling filter and the upsampling filter. From top to bottom: — Filter— 3—7 Filter Set, Default JS VM Filter Set, our—Filter—5—7 Filter Set, and JVT-U 147—Filter— 5—7 Curves from the Filter Set, Default JSVM Filter The curves of the bank and JVT-U 147—Filter—5-7 filter bank are indicated by dashed lines.
our— Filter— 3— 7滤波器组和 our— Filter— 5— 7滤波器组是使用本发明的方 法设计得到的滤波器组， Default JSVM滤波器组和 JVT-U 147— Filter— 5— 7 滤波器组是现有的滤波器组。 从图 6a中可以看出， 在相同的分辨率上， 使用 our— Filter— 3— 7滤波器组和 our— Filter— 5— 7滤波器组测试得到的 PSNR 的值要分别大于使用 Default JSVM滤波器组和 JVT-U147— Filter— 5— 7滤波 器组测试得到的 PSNR的值， 因此可以得到这样的结论： 使用本发明技 术方案生成的滤波器较现有的 JS VM6.3.1滤波器， 在对图像进行处理时，
可以显著提高高分辨率的重建效果。 Our_Filter—3-7 filter bank and our_Filter—5-7 filter bank are filter banks designed using the method of the present invention, Default JSVM filter bank and JVT-U 147—Filter— 5-7 The filter bank is an existing filter bank. As can be seen from Figure 6a, at the same resolution, the values of PSNR obtained using the Our-Filter-3-7 filter bank and our-Filter-7-7 filter bank are greater than those using Default JSVM filtering. The value of the PSNR obtained by the filter set and the JVT-U147-Filter- 5-7 filter bank, so it can be concluded that the filter generated by the technical solution of the present invention is more than the existing JS VM 6.3.1 filter. When processing an image, Can significantly improve the reconstruction of high resolution.
图 6b是对工头 (foreman) 进行层 0帧内编码所得的图形。 Figure 6b is a graph of layer 0 intraframe coding of the foreman.
图 6c是对足球 (football) 进行层 0帧内编码所得的图形。 Figure 6c is a graph of layer 0 intra-frame coding of a football.
图 6d是对公共汽车 (bus) 进行层 0帧内编码所得的图形。 Figure 6d is a graph of layer 0 intraframe coding of a bus.
图 6al是对 Mobile进行层 1帧内编码所得到的图形。如图 6al所示的图 形与图 6a相比， 差别仅在于图 6al是对层 1进行帧内编码所得到的图形。 从上到下依次为： 使用 our— Filter— 3— 7滤波器组、 Default JSVM滤波器组、 our— Filter— 5— 7滤波器组和 JVT-U 147— Filter— 5— 7滤波器组进行测试得到的 曲线， Default JSVM滤波器组和 JVT-U 147— Filter— 5— 7滤波器组的曲线用 虚线来表示。 Figure 6al is a graph obtained by layer 1 intra-frame coding of Mobile. The pattern shown in Fig. 6al is different from Fig. 6a only in that Fig. 6al is a graph obtained by intraframe coding layer 1. From top to bottom: using the “our—Filter—3-7 filter bank, the Default JSVM filter bank, our—Filter—5-7 filter bank, and the JVT-U 147—Filter—5-7 filter bank. The curve obtained by the test, the default JSVM filter bank and the curve of the JVT-U 147-Filter-5-7 filter bank are indicated by dashed lines.
图 6b 1是对 foreman进行层 1帧内编码所得的图形。 Figure 6b 1 is a graph of layer 1 intra-frame coding of foreman.
图 6c 1是对 football进行层 1帧内编码所得的图形。 Figure 6c 1 is a graph of layer 1 intraframe coding of football.
图 6dl是对 bus进行层 1帧内编码所得的图形。 Figure 6dl is a graph of layer 1 intra-frame coding of the bus.
从图 6b、 图 6c、 图 6d、 图 6bl、 图 6cl、 图 6dl得出的结论与从图 6a 得出的结论相同， 即使用本发明技术方案生成的滤波器较现有的 The conclusions drawn from Fig. 6b, Fig. 6c, Fig. 6d, Fig. 6bl, Fig. 6cl, Fig. 6dl are the same as those obtained from Fig. 6a, that is, the filter generated by the technical solution of the present invention is more conventional than
JSVM6.3.1滤波器， 在对图像进行处理时， 可以显著提高高分辨率图像 的重建效果。 The JSVM6.3.1 filter can significantly improve the reconstruction of high-resolution images when processing images.
至此， 结束对图像的定性测试结果的介绍。 At this point, the introduction of the qualitative test results of the image is ended.
在接下来的实施例中，介绍利用图 2所示实施例生成的下采样滤波器 和上采样滤波器实现空间可分级视频编码的装置和***。 In the following embodiments, an apparatus and system for implementing spatial scalable video coding using the downsampling filter and upsampling filter generated by the embodiment shown in Fig. 2 are described.
图 7为利用图 2所示实施例生成的下采样滤波器实现空间可分级视频 编码的装置的结构示意图。 如图 7所示， 该装置包括： 下采样滤波器、 时域分解单元、 运动编码单元， 以及下采样帧内预测和帧内编码单元。 FIG. 7 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using a downsampling filter generated by the embodiment shown in FIG. As shown in FIG. 7, the apparatus includes: a downsampling filter, a time domain decomposition unit, a motion coding unit, and downsampling intra prediction and intra coding units.
其中， 下采样滤波器， 用于接收作为编码输入的视频信号， 对该视 频信号进行下采样处理， 得到下采样视频序列， 将该下采样视频序列发
送给时域分解单元。 The downsampling filter is configured to receive a video signal as an encoded input, and perform downsampling processing on the video signal to obtain a downsampled video sequence, and send the downsampled video sequence Send to the time domain decomposition unit.
时域分解单元接收下采样滤波器发送的下采样视频序列， 将该下采 样视频序列分解成 I帧、 P帧和 B帧， 将 I帧发送给下采样帧内预测和帧内 编码单元， 将 P帧和 B帧发送给运动编码单元。 The time domain decomposition unit receives the downsampled video sequence sent by the downsampling filter, decomposes the downsampled video sequence into I frames, P frames, and B frames, and sends the I frames to the downsampled intra prediction and intra coding units, The P frame and the B frame are sent to the motion coding unit.
运动编码单元接收时域分解单元发送的 P帧和 B帧，对 P帧和 B帧进行 运动编码得到运动预测编码信号。 The motion coding unit receives the P frame and the B frame transmitted by the time domain decomposition unit, and performs motion coding on the P frame and the B frame to obtain a motion prediction coded signal.
下采样帧内预测和帧内编码单元接收 I帧， 对该 I帧进行帧内预测和 帧内编码， 得到下采样预测编码信号。 The downsampled intra prediction and intra coding unit receives the I frame, and performs intra prediction and intra coding on the I frame to obtain a downsampled predictive coded signal.
需要指出的是， 本实施例中用到的下采样滤波器是对符合马尔科夫 模型的视频信号进行下采样处理和上采样处理得到重建信号， 然后根据 带参数的双正交小波理论和重建信号的能量， 计算出重建信号的能量最 大值所对应的小波滤波器参数， 生成的下采样滤波器， 在本实施例中， 带参数的双正交小波理论即带参数的双正交小波滤波器参数化公式。 需 要指出的是， 生成下采样滤波器所用的符合马尔科夫模型的视频信号和 作为编码输入的视频信号可以是相同的信号， 也可以是不同的信号。 作 为编码输入的视频信号可以为所有的视频信号， 而所有的这些视频信号 中大概有 95%是符合马尔科夫模型的视频信号。 It should be noted that the downsampling filter used in this embodiment is to perform downsampling processing and upsampling processing on the video signal conforming to the Markov model to obtain a reconstructed signal, and then according to the bi-orthogonal wavelet theory with parameters and reconstruction. The energy of the signal, the wavelet filter parameter corresponding to the maximum energy of the reconstructed signal is calculated, and the generated downsampling filter is used. In this embodiment, the bi-orthogonal wavelet theory with parameters is a bi-orthogonal wavelet filter with parameters. Parameterization formula. It should be noted that the video signal conforming to the Markov model used to generate the downsampling filter and the video signal as the coded input may be the same signal or different signals. The video signal input as a code can be all video signals, and about 95% of all of these video signals are video signals conforming to the Markov model.
至此， 结束对图 7所示实施例的介绍。 So far, the introduction of the embodiment shown in Fig. 7 is ended.
图 8为利用图 2所示实施例生成的上采样滤波器实现空间可分级视频 编码的装置的结构示意图。 如图 8所示， 该装置包括： 上采样滤波器和 上采样帧内预测和帧内编码单元。 FIG. 8 is a schematic structural diagram of an apparatus for implementing spatial scalable video coding by using an upsampling filter generated by the embodiment shown in FIG. As shown in Figure 8, the apparatus includes: an upsampling filter and upsampled intra prediction and intra coding units.
其中，上采样滤波器，用于接收图 7中的下采样帧内预测和帧内编码 单元输出的下采样帧内预测编码信号， 对接收到的下采样帧内预测编码 信号进行上采样处理， 得到上采样视频序列， 将该上采样视频序列发送 给上采样帧内预测和帧内编码单元。
上采样帧内预测和帧内编码单元， 用于接收上采样滤波器发送的上 采样视频序列， 对接收到上采样视频序列进行帧内预测和帧内编码， 得 到上采样帧内预测编码信号。 The upsampling filter is configured to receive the downsampled intra prediction coding signal outputted by the downsampled intra prediction and the intra coding unit in FIG. 7, and perform upsampling processing on the received downsampled intra prediction coding signal, An upsampled video sequence is obtained, which is sent to the upsampled intra prediction and intra coding units. The upsampling intra prediction and intra coding unit is configured to receive an upsampled video sequence sent by the upsampling filter, perform intra prediction and intraframe coding on the received upsampled video sequence, to obtain an upsampled intra prediction encoded signal.
需要指出的是， 本实施例中用到的上采样滤波器是对视频信号进行 下采样处理和上采样处理得到重建信号， 然后根据带参数的双正交小波 理论和重建信号的能量， 计算出重建信号的能量最大值所对应的小波滤 波器参数， 生成的上采样滤波器， 在本实施例中， 带参数的双正交小波 理论即带参数的双正交小波滤波器参数化公式。 需要指出的是， 生成上 采样滤波器所用的符合马尔科夫模型的视频信号和作为编码输入的视 频信号可以是相同的信号，也可以是不同的信号，一般情况下并不相同。 It should be noted that the upsampling filter used in this embodiment is to perform downsampling processing and upsampling processing on the video signal to obtain a reconstructed signal, and then calculate according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The wavelet filter parameters corresponding to the energy maximum of the reconstructed signal, the generated upsampling filter, in this embodiment, the parameterized bi-orthogonal wavelet theory, that is, the parameterized formula of the parameterized bi-orthogonal wavelet filter. It should be noted that the video signal conforming to the Markov model used to generate the upsampling filter and the video signal used as the code input may be the same signal or different signals, which are generally not the same.
至此， 结束对图 8所示实施例的介绍。 So far, the introduction of the embodiment shown in Fig. 8 is ended.
图 9为利用图 2所示实施例生成的下采样滤波器和上采样滤波器实现 空间可分级视频编码的***的结构示意图。 如图 9所示， 该***包括： 下采样滤波器、 下采样编码单元、 上采样滤波器、 上采样帧内预测和帧 内编码单元、 以及复用单元。 其中， 下采样编码单元包括： 时域分解单 元、 运动编码单元和下采样帧内预测和帧内编码单元。 FIG. 9 is a schematic structural diagram of a system for realizing spatial scalable video coding by using a downsampling filter and an upsampling filter generated by the embodiment shown in FIG. As shown in FIG. 9, the system includes: a downsampling filter, a downsampling coding unit, an upsampling filter, upsampling intra prediction and intra coding units, and a multiplexing unit. The downsampling coding unit includes: a time domain decomposition unit, a motion coding unit, and a downsampling intra prediction and intra coding unit.
其中， 下采样滤波器， 用于接收作为编码输入的视频信号， 对该视 频信号进行下采样处理， 得到下采样视频序列， 将该下采样视频序列发 送给下采样编码单元的时域分解单元。 The downsampling filter is configured to receive a video signal as an encoded input, perform downsampling processing on the video signal, obtain a downsampled video sequence, and send the downsampled video sequence to a time domain decomposition unit of the downsampling coding unit.
时域分解单元， 用于接收下采样滤波器发送的下采样视频序列， 将 接收到的下采样视频序列分解成 I帧、 P帧和 B帧， 将 I帧发送给下采样帧 内预测和帧内编码单元， 将 P帧和 B帧发送给运动编码单元。 a time domain decomposition unit, configured to receive a downsampled video sequence sent by the downsampling filter, decompose the received downsampled video sequence into I frames, P frames, and B frames, and send the I frames to the downsampled intra prediction and frame. The inner coding unit transmits the P frame and the B frame to the motion coding unit.
运动编码单元， 用于接收 P帧和 B帧，对接收到的 P帧和 B帧进行运动 编码得到运动编码信号， 将该运动预测编码信号发送给复用单元。 The motion coding unit is configured to receive the P frame and the B frame, perform motion coding on the received P frame and the B frame to obtain a motion coded signal, and send the motion prediction coded signal to the multiplexing unit.
下采样帧内预的和帧内编码单元， 用于接收 I帧， 对接收到的 I帧进
行帧内预测和帧内编码， 得到下采样帧内预测和编码信号， 将该下采样 帧内预测编码信号发送给上采样滤波器和复用单元。 Down-sampling intra-pre and intra-coded units for receiving I-frames, for incoming I-frames The intra-frame prediction and the intra-frame coding obtain a downsampled intra prediction and coding signal, and send the downsampled intra prediction coding signal to the upsampling filter and the multiplexing unit.
上采样滤波器， 用于接收下采样帧内预测编码信号， 对接收到的下 采样帧内预测编码信号进行上采样处理， 得到上采样视频序列， 并发送 给上采样帧内预测和帧内编码单元。 An upsampling filter, configured to receive the downsampled intra prediction coded signal, perform upsampling on the received downsampled intraframe predictive coded signal, obtain an upsampled video sequence, and send to the upsampled intraframe and intraframe coding unit.
上采样帧内预测和帧内编码单元， 用于接收上采样滤波器发送的上 采样视频序列， 对该上采样视频序列进行帧内预测和帧内编码， 得到上 采样帧内预测编码信号， 并将该上采样帧内预测编码信号发送给复用单 元。 An upsampled intra prediction and intra coding unit, configured to receive an upsampled video sequence sent by the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain an upsampled intra prediction coding signal, and The upsampled intra prediction encoded signal is transmitted to the multiplexing unit.
复用单元， 用于接收下采样帧内预测编码信号、 运动预测编码信号 和上采样帧内预测编码信号， 对接收到的信号进行合成处理， 得到视频 编码信号。 And a multiplexing unit, configured to receive the downsampled intra prediction encoded signal, the motion prediction encoded signal, and the upsampled intra prediction encoded signal, and perform synthesis processing on the received signal to obtain a video encoded signal.
需要指出的是， 本实施中所用到的下采样滤波器和上采样滤波器是 对符合马尔科夫模型的视频信号进行下采样和上采样处理得到重建信 号， 根据带参数的双正交小波理论和所述重建信号的能量， 计算出所述 能量的最大值所对应的 d、波滤波器参数， 生成的下采样滤波器和上采样 滤波器， 在本实施例中， 带参数的双正交小波理论即带参数的双正交小 波滤波器参数化公式。 需要指出的是， 生成下采样滤波器和上采样滤波 器所用的符合马尔科夫模型的视频信号与作为编码输入的视频信号可 以是同一信号， 也可以是不同的信号， 一般情况下是不同的信号。 It should be noted that the downsampling filter and the upsampling filter used in this implementation are down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal, according to the bi-orthogonal wavelet theory with parameters. And the energy of the reconstructed signal, the d, the filter parameter corresponding to the maximum value of the energy, the generated downsampling filter and the upsampling filter, in this embodiment, the bi-orthogonal with parameters Wavelet theory is a parametric formula of a bi-orthogonal wavelet filter with parameters. It should be noted that the video signal conforming to the Markov model used to generate the downsampling filter and the upsampling filter may be the same signal or a different signal as the encoded input, and is generally different. signal.
此外， 复用单元的作用主要是对下采样帧内预测编码信号、 运动预 测编码信号和上采样帧内预测编码信号进行合成处理， 得到视频编码信 号。 如果空间可分级视频编码***不对这三路信号进行合成处理， 在本 实施例中也可以不包括复用单元。 In addition, the multiplexing unit mainly performs synthesis processing on the downsampled intra prediction encoded signal, the motion prediction encoded signal, and the upsampled intra prediction encoded signal to obtain a video encoded signal. If the spatial scalable video coding system does not synthesize these three signals, the multiplexing unit may not be included in this embodiment.
至此， 结束对图 9所示实施例的介绍。
由以上实施例可以看出，图 2〜图 6中出现的上、下采样滤波器为公式， 图 7〜图 9中的上、 下采样滤波器为实际的物理单元。 So far, the introduction of the embodiment shown in Fig. 9 is ended. As can be seen from the above embodiment, the up and down sampling filters appearing in FIGS. 2 to 6 are formulas, and the up and down sampling filters in FIGS. 7 to 9 are actual physical units.
本发明实施例具有如下优点： Embodiments of the present invention have the following advantages:
首先， 本发明实施例生成的上采样滤波器和下采样滤波器是对能量 模型求解最大化的情况下生成的上下采样滤波器， 在保证同等滤波器长 用中针对能量模型是最优的。 将本发明实施例生成的上采样滤波器和下 采样滤波器用于视频信号的编码处理， 可以减少图像能量的损耗， 提高 对视频信号进行编码的质量， 达到能量损失的最小化和视频恢复质量的 最大化。 First, the upsampling filter and the downsampling filter generated by the embodiment of the present invention are upper and lower sampling filters generated when the energy model is maximized, and are optimal for the energy model in ensuring the same filter length. The upsampling filter and the downsampling filter generated by the embodiment of the present invention are used for encoding processing of a video signal, which can reduce image energy loss, improve quality of encoding the video signal, minimize energy loss, and restore video quality. maximize.
其次， 将带参数的双正交小波理论用于上采样滤波器和下采样滤波 器的设计， 从而使设计出的上采样滤波器和下采样滤波器具有关联性。 以提高视频信号的重构质量， 优化高分辨率下视频信号的重构效果。 Secondly, the bi-orthogonal wavelet theory with parameters is applied to the design of the upsampling filter and the downsampling filter, so that the designed upsampling filter and downsampling filter are related. In order to improve the reconstruction quality of the video signal, the reconstruction effect of the video signal under high resolution is optimized.
最后， 可以根据不同的视频信号编码复杂度的要求， 可以对生成的 整数滤波器进行不同长度的裁剪， 生成优化整数滤波器， 从而通过在运 算速度和视频信号重建性能之间的权衡， 生成最符合实际需要的优化整 数上采样滤波器和优化整数下采样滤波器。 Finally, according to different video signal coding complexity requirements, the generated integer filter can be tailored to different lengths to generate an optimized integer filter, thereby generating the most trade-off between the operation speed and the video signal reconstruction performance. Optimized integer upsampling filters and optimized integer downsampling filters that meet actual needs.
需要指出的是，本发明实施例可以用计算机程序语言来实现，或者， 可以由预置程序的硬件元件、 其他相关的组件， 或者硬件和软件相结合 的组件来实现。 It should be noted that the embodiments of the present invention may be implemented in a computer program language, or may be implemented by hardware components of a preset program, other related components, or a combination of hardware and software.
此外， 可以将本发明实施例所示的方法、 装置和***制作成计算机 程序产品应用于计算机***。 例如， 可以制作成一系列计算机指令固化 在实际介质中 （光盘、 CD-ROM, ROM或者硬盘）， 或者利用无线技术 实现的介质。
综上所述， 以上仅为本发明的较佳实施例而已， 并非用于限定本发 明的保护范围。 凡在本发明的精神和原则之内， 所作的任何修改、 等同 替换、 改进等， 均应包含在本发明的保护范围之内。
Moreover, the methods, apparatus, and systems shown in the embodiments of the present invention can be fabricated into a computer program product for use in a computer system. For example, it can be fabricated as a series of computer instructions that are solidified in actual media (CD, CD-ROM, ROM or hard disk), or media implemented using wireless technology. In conclusion, the above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and scope of the present invention are intended to be included within the scope of the present invention.
Claims
1、 一种生成上下采样滤波器的方法， 其特征在于， 包括： 对符合马尔科夫模型的视频信号进行下采样和上采样， 得到重建信 号； A method for generating an up-and-down sampling filter, comprising: down-sampling and up-sampling a video signal conforming to a Markov model to obtain a reconstructed signal;
计算所述重建信号的能量； Calculating the energy of the reconstructed signal;
根据带参数的双正交小波理论和所述重建信号的能量， 计算所述重 建信号的能量的最大值所对应的小波滤波器参数， 生成上采样滤波器和 下采样滤波器。 And calculating a wavelet filter parameter corresponding to a maximum value of the energy of the reconstructed signal according to a bi-orthogonal wavelet theory with parameters and an energy of the reconstructed signal, and generating an upsampling filter and a downsampling filter.
2、根据权利要求 1所述的方法， 其特征在于， 所述计算所述重建信 号的能量包括： The method according to claim 1, wherein the calculating the energy of the reconstructed signal comprises:
将所述重建信号代入能量模型， 计算所述重建信号的能量， 得到重 建信号的能量公式。 Substituting the reconstructed signal into an energy model, calculating the energy of the reconstructed signal, and obtaining an energy formula for reconstructing the signal.
3、根据权利要求 2所述的方法， 其特征在于， 计算所述重建信号的 能量的最大值所对应的小波滤波器参数包括： The method according to claim 2, wherein the calculating the wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal comprises:
将带参数的双正交小波滤波器参数化公式代入所述重建信号的能量 公式， 计算所述重建信号的能量的最大值所对应的小波滤波器参数。 The parameterized formula of the parameterized bi-orthogonal wavelet filter is substituted into the energy formula of the reconstructed signal, and the wavelet filter parameter corresponding to the maximum value of the energy of the reconstructed signal is calculated.
4、根据权利要求 2或 3所述的方法，其特征在于，所述能量模型为： El = max{aE[e(W] + βΕ[ε(Ξ2 )] + /E[e(S3 )]}； 4. Method according to claim 2 or 3, characterized in that the energy model is: El = max{aE[e(W] + βΕ[ε(Ξ 2 )] + /E[e(S 3 )]};
其中， 所述 是进行下采样和上采样处理后的视频信号的能 量，所述 是进行下采样和上采样处理前的视频信号和所述重建信 号的相关能量， 所述 £[ 3 )]是进行下采样处理后的视频信号的能量； 所 述 α和 β均为大于零的正整数， γ为大于等于零的正整数。 Wherein, the energy of the video signal after performing the downsampling and the upsampling process, wherein the video signal before the downsampling and the upsampling process and the reconstructed signal are related energy, wherein the £[ 3]] is The energy of the video signal after the downsampling process; the α and β are positive integers greater than zero, and γ is a positive integer greater than or equal to zero.
5、根据权利要求 3所述的方法， 其特征在于， 所述带参数的双正交 小波滤波器参数化公式包括： 单参数双正交小波参数化公式， 或双参数
双正交小波参数化公式， 或多参数双正交小波参数化公式， 或单参数和 多参数双正交小波参数化公式。 The method according to claim 3, wherein the parametric formula of the parameterized bi-orthogonal wavelet filter comprises: a single parameter double orthogonal wavelet parameterization formula, or a double parameter A bi-orthogonal wavelet parameterization formula, or a multi-parameter bi-orthogonal wavelet parameterization formula, or a single-parameter and multi-parameter bi-orthogonal wavelet parameterization formula.
6、 根据权利要求 1所述的方法， 其特征在于， 进一步包括： 对所述上采样滤波器和下采样滤波器进行整数化处理， 生成整数上 采样滤波器和整数下采样滤波器。 6. The method according to claim 1, further comprising: performing integer processing on the upsampling filter and the downsampling filter to generate an integer upsampling filter and an integer downsampling filter.
7、 根据权利要求 6所述的方法， 其特征在于， 进一步包括： 对所述整数上采样滤波器和整数下采样滤波器进行长度裁减， 生成 优化整数上采样滤波器和优化整数下采样滤波器。 The method according to claim 6, further comprising: performing length reduction on the integer upsampling filter and the integer downsampling filter to generate an optimized integer upsampling filter and an optimized integer downsampling filter. .
8、根据权利要求 6所述的方法， 其特征在于， 所述整数上采样滤波 器和整数下采样滤波器分别为： H7 =[-1 0 9 16 9 0 -1] /16和 H9=[l -1 -4 9 The method according to claim 6, wherein the integer upsampling filter and the integer downsampling filter are: H7 = [-1 0 9 16 9 0 -1] /16 and H9=[ l -1 -4 9
22 9 -4 -1 1 ]/32； 22 9 -4 -1 1 ]/32;
或，分别为： HH7=[-8 1 72 126 72 1 -8]/128和 HH13=[-1 0 18 -16 -65 144 352 144 -65 -16 18 0 -1]/512； Or, respectively: HH7=[-8 1 72 126 72 1 -8]/128 and HH13=[-1 0 18 -16 -65 144 352 144 -65 -16 18 0 -1]/512;
或， 分别为： H7 = [-5 0 21 32 21 0 -5]/32和 H3 = [17 30 17]/64。 9、根据权利要求 7所述的方法， 其特征在于， 对所述整数上采样滤 波器和整数下采样滤波器进行长度裁减包括： Or, respectively: H7 = [-5 0 21 32 21 0 -5]/32 and H3 = [17 30 17]/64. The method according to claim 7, wherein the length reduction of the integer upsampling filter and the integer downsampling filter comprises:
将整数上采样滤波器和整数下采样滤波器两端影响小的值省略， 并 累加到中心值上， 分别生成优化整数上采样滤波器和优化整数下采样滤 波器。 The small-valued values of the integer up-sampling filter and the integer down-sampling filter are omitted and added to the center value to generate an optimized integer upsampling filter and an optimized integer downsampling filter, respectively.
10、 根据权利要求 9所述的方法， 其特征在于， 所述生成的优化整 数上采样滤波器和优化整数下采样滤波器分别为： H7 =[-1 0 9 16 9 0 -1] /16和 H5=[-4 9 22 9 -4]/32。 10. The method according to claim 9, wherein the generated optimized integer upsampling filter and optimized integer downsampling filter are: H7 = [-1 0 9 16 9 0 -1] /16 And H5=[-4 9 22 9 -4]/32.
11、 一种利用下采样滤波器实现空间可分级视频编码的方法， 其特 征在于， 包括： 11. A method for spatially scalable video coding using a downsampling filter, the method comprising:
使用下采样滤波器对用于编码的视频信号进行下采样， 得到下采样
视频序列； Downsampling the video signal used for encoding using a downsampling filter Video sequence
对所述下采样视频序列进行时域分解， 得到内部 I帧、 预测 P帧和 双向预测 B帧； Performing time domain decomposition on the downsampled video sequence to obtain an internal I frame, a predicted P frame, and a bidirectional predicted B frame;
对所述 I帧进行帧内预测和帧内编码， 得到下采样帧内预测编码信 号； 对所述 P帧和 B帧进行运动编码， 得到运动预测编码信号； Performing intra prediction and intra coding on the I frame to obtain a downsampled intra prediction coding signal; performing motion coding on the P frame and the B frame to obtain a motion prediction coding signal;
所述下采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的下采样滤波器。 The downsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the downsampling filter.
12、根据权利要求 11所述的方法， 其特征在于， 使用下采样滤波器 对用于编码的视频信号进行下采样， 得到下采样视频序列包括： 12. The method according to claim 11, wherein the downsampling filter is used to downsample the video signal for encoding, and obtaining the downsampled video sequence comprises:
将所述用于编码的视频信号中同行或隔行中的各像素与所述下采样 滤波器进行垂直方向上的卷积计算， 得到各像素的像素值； 将所述各像 样视频序列； Performing convolution calculation of each pixel in the peer or interlaced line of the video signal for encoding in the vertical direction with the downsampling filter to obtain pixel values of each pixel; and the respective video sequences;
或， 将所述视频信号中同行或隔行中的各像素与所述下采样滤波器 进行水平方向上的卷积计算， 得到各像素的像素值； 将所述各像素的像 素值与所述下采样滤波器进行垂直方向上的卷积计算， 得到下采样视频 序列。 Or performing a convolution calculation in the horizontal direction of each pixel in the video signal or in the interlaced signal with the downsampling filter to obtain a pixel value of each pixel; and setting the pixel value of each pixel to the lower The sampling filter performs a convolution calculation in the vertical direction to obtain a downsampled video sequence.
13、根据权利要求 11所述的方法， 其特征在于， 所述用于编码的视 频信号和所述符合马尔科夫模型的视频信号是相同类型的信号。 The method according to claim 11, wherein the video signal for encoding and the video signal conforming to the Markov model are the same type of signal.
14、 一种利用上采样滤波器实现空间可分级视频编码的方法， 其特 征在于， 包括： 14. A method for spatially scalable video coding using an upsampling filter, the method comprising:
使用上采样滤波器对下采样帧内预测编码信号进行上采样， 得到上 采样视频序列；
对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内 预测编码信号； Upsampling the downsampled intra prediction encoded signal using an upsampling filter to obtain an upsampled video sequence; Performing intra prediction and intra coding on the upsampled video sequence to obtain an upsampled intra prediction encoded signal;
所述上采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的上采样滤波器。 The upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
15、根据权利要求 14所述的方法， 其特征在于， 所述使用上采样滤 波器对下采样帧内预测编码信号进行上采样， 得到上采样视频序列包 括： 理预测和层间残差预测的 2维 2D空域上采样， 得到上采样视频序列。 The method according to claim 14, wherein the upsampling the intra-predictive coded signal is upsampled using an upsampling filter, and the obtained upsampled video sequence comprises: a prediction and an inter-layer residual prediction. The 2D 2D spatial domain is upsampled to obtain an upsampled video sequence.
16、 一种利用下采样滤波器实现空间可分级视频编码的装置， 其特 征在于， 包括： 16. A device for spatially scalable video coding using a downsampling filter, the method comprising:
下采样滤波器， 用于接收用于编码的视频信号， 对所述视频信号进 行下采样， 得到下采样视频序列； a downsampling filter, configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence;
时域分解单元，用于从所述下采样滤波器接收所述下采样视频序列， 将所述下采样视频序列分解成 I帧、 P帧和 B帧； a time domain decomposition unit, configured to receive the downsampled video sequence from the downsampling filter, and decompose the downsampled video sequence into an I frame, a P frame, and a B frame;
运动编码单元， 用于从时域分解单元接收所述 P帧和 B帧， 对所述 P帧和 B帧进行运动编码得到运动预测编码信号； a motion coding unit, configured to receive the P frame and the B frame from a time domain decomposition unit, and perform motion coding on the P frame and the B frame to obtain a motion prediction encoded signal;
下采样帧内预测和帧内编码单元， 用于从所述时域分解单元接收所 述 I帧， 对所述 I帧进行帧内预测和帧内编码， 得到下采样帧内预测编 码信号； a downsampling intra prediction and intra coding unit, configured to receive the I frame from the time domain decomposition unit, perform intra prediction and intra coding on the I frame, to obtain a downsampled intra prediction coding signal;
所述下采样滤波器是对符合马尔科夫模型的视频信号进行下采样处 理和上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信 号的能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参
数， 生成的下采样滤波器。 The downsampling filter is configured to perform downsampling processing and upsampling the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. Wavelet filter parameter corresponding to the maximum value of energy Number, generated downsampling filter.
17、 一种利用上采样滤波器实现空间可分级视频编码的装置， 其特 征在于， 包括： 17. An apparatus for spatially scalable video coding using an upsampling filter, the method comprising:
上采样滤波器， 用于接收下采样帧内预测编码信号， 并对所述下采 样帧内预测编码信号进行上采样， 得到上采样视频序列； An upsampling filter, configured to receive a downsampled intra prediction encoded signal, and upsample the downsampled intra prediction encoded signal to obtain an upsampled video sequence;
上采样帧内预测和帧内编码单元， 用于从所述上采样滤波器接收所 述上采样视频序列， 对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内预测编码信号； Upsampling intra prediction and intra coding unit, configured to receive the upsampled video sequence from the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain upsampled intra prediction Coded signal
所述上采样滤波器是对符合马尔科夫模型的视频信号进行下采样和 上采样得到重建信号， 根据带参数的双正交小波理论和所述重建信号的 能量， 计算所述重建信号的能量的最大值所对应的小波滤波器参数， 生 成的上采样滤波器。 The upsampling filter is configured to downsample and upsample the video signal conforming to the Markov model to obtain a reconstructed signal, and calculate the energy of the reconstructed signal according to the parameter of the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal. The maximum value corresponds to the wavelet filter parameters generated by the upsampling filter.
18、 一种利用下采样滤波器和上采样滤波器实现空间可分级视频编 码的***， 其特征在于， 包括： 18. A system for spatially scalable video coding using a downsampling filter and an upsampling filter, comprising:
下采样滤波器， 用于接收用于编码的视频信号， 对所述视频信号进 行下采样， 得到下采样视频序列； a downsampling filter, configured to receive a video signal for encoding, and downsample the video signal to obtain a downsampled video sequence;
下采样编码单元， 用于从所述下采样滤波器接收所述下采样视频序 列， 对所述下采样视频序列进行预测和编码， 生成运动预测编码信号和 下采样帧内预测编码信号； And a downsampling coding unit, configured to receive the downsampled video sequence from the downsampling filter, perform prediction and encoding on the downsampled video sequence, and generate a motion prediction coded signal and a downsampled intra prediction coded signal;
上采样滤波器， 用于从所述下采样编码单元接收所述下采样帧内预 测编码信号， 对所述下采样帧内预测编码信号进行上采样处理， 得到上 采样视频序列； An upsampling filter, configured to receive the downsampled intra prediction encoded signal from the downsampling coding unit, and perform upsampling processing on the downsampled intra prediction encoded signal to obtain an upsampled video sequence;
上采样帧内预测和帧内编码单元， 用于从所述上采样滤波器接收所 述上采样视频序列， 对所述上采样视频序列进行帧内预测和帧内编码， 得到上采样帧内预测编码信号；
所述下采样滤波器和上采样滤波器是对符合马尔科夫模型的视频信 号进行下采样和上采样得到重建信号， 根据带参数的双正交小波理论和 所述重建信号的能量， 计算所述重建信号的能量的最大值所对应的小波 滤波器参数， 生成的下采样滤波器和上采样滤波器。 Upsampling intra prediction and intra coding unit, configured to receive the upsampled video sequence from the upsampling filter, perform intra prediction and intraframe coding on the upsampled video sequence, to obtain upsampled intra prediction Coded signal The downsampling filter and the upsampling filter are down-sampling and up-sampling the video signal conforming to the Markov model to obtain a reconstructed signal, according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, The wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal, the generated downsampling filter and the upsampling filter.
19、根据权利要求 18所述的***， 其特征在于， 所述下采样编码单 元包括： The system according to claim 18, wherein the downsampling coding unit comprises:
时域分解单元，用于从所述下采样滤波器接收所述下采样视频序列， 将所述下采样视频序列分解成 I帧、 P帧和 B帧； a time domain decomposition unit, configured to receive the downsampled video sequence from the downsampling filter, and decompose the downsampled video sequence into an I frame, a P frame, and a B frame;
运动编码单元， 用于从所述时域分解单元接收所述 P帧和 B帧， 对 所述 P帧和 B帧进行运动编码得到运动预测编码信号； a motion coding unit, configured to receive the P frame and the B frame from the time domain decomposition unit, and perform motion coding on the P frame and the B frame to obtain a motion prediction encoded signal;
下采样帧内预测和帧内编码单元， 用于从所述时域分解单元接收所 述 I帧， 对所述 I帧进行帧内预测和帧内编码， 得到下采样帧内预测编 码信号； a downsampling intra prediction and intra coding unit, configured to receive the I frame from the time domain decomposition unit, perform intra prediction and intra coding on the I frame, to obtain a downsampled intra prediction coding signal;
所述上采样滤波器， 进一步用于接收所述下采样帧内预测和帧内编 码单元发送的下采样帧内预测编码信号。 The upsampling filter is further configured to receive the downsampled intra prediction encoded signal sent by the downsampled intra prediction and the intra coding unit.
20、 根据权利要求 19所述的***， 其特征在于， 进一步包括： 复用 单元； 20. The system of claim 19, further comprising: a multiplexing unit;
所述下采样帧内预测和帧内编码单元， 进一步用于将所述下采样帧 内预测编码信号发送给复用单元； The downsampling intra prediction and intra coding unit is further configured to send the downsampled intra prediction encoded signal to a multiplexing unit;
所述运动编码单元， 进一步用于将运动预测编码信号发送给复用单 元； The motion coding unit is further configured to send the motion prediction encoded signal to the multiplexing unit;
所述上采样帧内预测和帧内编码单元， 进一步用于将所述上采样帧 内预测编码信号发送给复用单元； The upsampling intra prediction and intra coding unit is further configured to send the upsampled intra prediction encoded signal to a multiplexing unit;
所述复用单元， 用于接收所述下采样帧内预测编码信号、 运动预测 编码信号和上采样帧内预测编码信号， 对接收到的信号进行合成处理，
得到视频编码信号。 The multiplexing unit is configured to receive the downsampled intra prediction encoded signal, the motion prediction encoded signal, and the upsampled intra prediction encoded signal, and perform synthesis processing on the received signal, The video encoded signal is obtained.
21、 一种生成上下采样滤波器的装置， 其特征在于， 包括： 用于对符合马尔科夫模型的视频信号进行下采样和上采样， 得到重 建信号的单元； An apparatus for generating an up-and-down sampling filter, comprising: means for down-sampling and up-sampling a video signal conforming to a Markov model to obtain a unit for reconstructing a signal;
计算所述重建信号的能量的单元； 和 Means for calculating the energy of the reconstructed signal; and
根据带参数的双正交小波理论和所述重建信号的能量， 计算所述重 建信号的能量的最大值所对应的小波滤波器参数， 生成上采样滤波器和 下采样滤波器的单元。 The wavelet filter parameters corresponding to the maximum value of the energy of the reconstructed signal are calculated according to the bi-orthogonal wavelet theory with parameters and the energy of the reconstructed signal, and the units of the upsampling filter and the downsampling filter are generated.
22、 根据权利要求 21所述的装置， 其特征在于， 进一步包括： 对所述上采样滤波器和下采样滤波器进行整数化处理， 生成整数上 采样滤波器和整数下采样滤波器的单元。 22. The apparatus according to claim 21, further comprising: performing integer processing on the upsampling filter and the downsampling filter to generate a unit of an integer upsampling filter and an integer downsampling filter.
23、 根据权利要求 22所述的装置， 其特征在于， 进一步包括： 对所述整数上采样滤波器和整数下采样滤波器进行长度裁减， 生成 优化整数上采样滤波器和优化整数下采样滤波器的单元。
The apparatus according to claim 22, further comprising: performing length reduction on the integer upsampling filter and the integer downsampling filter to generate an optimized integer upsampling filter and an optimized integer downsampling filter Unit.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710106131 CN101312529B (en) | 2007-05-24 | 2007-05-24 | Method, system and apparatus generating up and down sampling filter |
| CN200710106131.8 | 2007-05-24 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2008145039A1 true WO2008145039A1 (en) | 2008-12-04 |
Family
ID=40074575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2008/070902 WO2008145039A1 (en) | 2007-05-24 | 2008-05-07 | Methods, systems and devices for generating upsample filter and downsample filter and for performing encoding |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN101312529B (en) |
| WO (1) | WO2008145039A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8483342B2 (en) | 2009-08-10 | 2013-07-09 | Huawei Device Co., Ltd. | Down sampling method and down sampling device |
| CN114324185A (en) * | 2022-01-04 | 2022-04-12 | 浙江大学 | Underwater polarization detection device based on Stokes vector |
| CN116708770A (en) * | 2018-10-08 | 2023-09-05 | 北京达佳互联信息技术有限公司 | Method, computing device and storage medium for decoding video signal |
| CN117254960A (en) * | 2023-09-25 | 2023-12-19 | 深圳市云钜天成信息技术有限公司 | Detection method for detecting API interface verification risk from flow data |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2524505B1 (en) * | 2010-01-15 | 2015-11-25 | Dolby Laboratories Licensing Corporation | Edge enhancement for temporal scaling with metadata |
| CN102263942A (en) * | 2010-05-31 | 2011-11-30 | 苏州闻道网络科技有限公司 | Scalable video transcoding device and method |
| CN102595125B (en) * | 2011-01-17 | 2016-04-06 | 深圳市云宙多媒体技术有限公司 | A kind of bi-directional predicted method of P frame and device |
| CN104656494A (en) * | 2013-11-19 | 2015-05-27 | 北大方正集团有限公司 | Signal real-time processing device |
| CN104574277A (en) * | 2015-01-30 | 2015-04-29 | 京东方科技集团股份有限公司 | Image interpolation method and image interpolation device |
| CN108848381B (en) * | 2018-06-20 | 2021-09-24 | 腾讯科技（深圳）有限公司 | Video encoding method, decoding method, device, computer device and storage medium |
| CN108776954B (en) * | 2018-06-26 | 2020-04-28 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image |
| CN108921801B (en) * | 2018-06-26 | 2020-01-07 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image |
| CN109672885B (en) * | 2019-01-08 | 2020-08-04 | 中国矿业大学(北京) | Video image coding and decoding method for intelligent monitoring of mine |
| CN114501012B (en) * | 2021-12-31 | 2024-06-11 | 浙江大华技术股份有限公司 | Image filtering, encoding and decoding methods and related equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030005007A1 (en) * | 2001-02-20 | 2003-01-02 | Nicola Neretti | Signal adaptive filter bank optimization |
| US6587507B1 (en) * | 1997-06-02 | 2003-07-01 | Picsurf, Inc. | System and method for encoding video data using computationally efficient adaptive spline wavelets |
| JP2006295913A (en) * | 2005-04-11 | 2006-10-26 | Sharp Corp | Method and device for adaptive upsampling for spatial scalable coding |
| CN1926873A (en) * | 2004-03-04 | 2007-03-07 | 三星电子株式会社 | Video encoding and decoding methods and systems for video streaming service |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1152493C (en) * | 2002-06-20 | 2004-06-02 | 上海交通大学 | Method for inhibiting peak-to-average ratio in orthogonal frequency division complex communication system |
| US7336208B2 (en) * | 2003-03-31 | 2008-02-26 | Nxp B.V. | Up and down sample rate converter |
| CN100410925C (en) * | 2004-12-30 | 2008-08-13 | 中国科学院长春光学精密机械与物理研究所 | Digital signal processing method for ultrasonic signals |
-
2007
- 2007-05-24 CN CN 200710106131 patent/CN101312529B/en not_active Expired - Fee Related
-
2008
- 2008-05-07 WO PCT/CN2008/070902 patent/WO2008145039A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6587507B1 (en) * | 1997-06-02 | 2003-07-01 | Picsurf, Inc. | System and method for encoding video data using computationally efficient adaptive spline wavelets |
| US20030005007A1 (en) * | 2001-02-20 | 2003-01-02 | Nicola Neretti | Signal adaptive filter bank optimization |
| CN1926873A (en) * | 2004-03-04 | 2007-03-07 | 三星电子株式会社 | Video encoding and decoding methods and systems for video streaming service |
| JP2006295913A (en) * | 2005-04-11 | 2006-10-26 | Sharp Corp | Method and device for adaptive upsampling for spatial scalable coding |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8483342B2 (en) | 2009-08-10 | 2013-07-09 | Huawei Device Co., Ltd. | Down sampling method and down sampling device |
| CN116708770A (en) * | 2018-10-08 | 2023-09-05 | 北京达佳互联信息技术有限公司 | Method, computing device and storage medium for decoding video signal |
| CN116708770B (en) * | 2018-10-08 | 2024-03-26 | 北京达佳互联信息技术有限公司 | Method, computing device and storage medium for decoding video signal |
| US11962789B2 (en) | 2018-10-08 | 2024-04-16 | Beijing Dajia Internet Information Technology Co., Ltd. | Simplifications of cross-component linear model |
| CN114324185A (en) * | 2022-01-04 | 2022-04-12 | 浙江大学 | Underwater polarization detection device based on Stokes vector |
| CN117254960A (en) * | 2023-09-25 | 2023-12-19 | 深圳市云钜天成信息技术有限公司 | Detection method for detecting API interface verification risk from flow data |
| CN117254960B (en) * | 2023-09-25 | 2024-05-10 | 深圳市云钜天成信息技术有限公司 | Detection method for detecting API interface verification risk from flow data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101312529A (en) | 2008-11-26 |
| CN101312529B (en) | 2010-07-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2008145039A1 (en) | Methods, systems and devices for generating upsample filter and downsample filter and for performing encoding | |
| KR100746011B1 (en) | Method for enhancing performance of residual prediction, video encoder, and video decoder using it | |
| KR100772882B1 (en) | Deblocking filtering method considering intra BL mode, and video encoder/decoder based on multi-layer using the method | |
| KR100664928B1 (en) | Video coding method and apparatus thereof | |
| JP4999340B2 (en) | Moving picture encoding apparatus, moving picture decoding apparatus, moving picture encoding method, and moving picture decoding method | |
| JP5014989B2 (en) | Frame compression method, video coding method, frame restoration method, video decoding method, video encoder, video decoder, and recording medium using base layer | |
| KR100703788B1 (en) | Video encoding method, video decoding method, video encoder, and video decoder, which use smoothing prediction | |
| JP5300921B2 (en) | Method and apparatus for encoding and decoding | |
| JP2008522537A (en) | Multi-layer video encoding method, decoding method and apparatus using the method | |
| KR20070074453A (en) | Method for encoding and decoding video signal | |
| US20060159359A1 (en) | Fine granularity scalable video encoding and decoding method and apparatus capable of controlling deblocking | |
| JP2005507588A (en) | Coding | |
| JP5301097B2 (en) | Transform domain sub-sampling for video transcoding | |
| WO2006006764A1 (en) | Video decoding method using smoothing filter and video decoder therefor | |
| WO2007041958A1 (en) | An up-sampling method and system for spatial layered coding video pictures | |
| CN102006483B (en) | Video coding and decoding method and device | |
| CN101360236B (en) | Wyner-ziv video encoding and decoding method | |
| JP2007006496A (en) | Method for hierarchical encoding of video source picture and method for decoding video source picture from subband information | |
| WO2006132509A1 (en) | Multilayer-based video encoding method, decoding method, video encoder, and video decoder using smoothing prediction | |
| Zhao et al. | Wavelet inpainting driven image compression via collaborative sparsity at low bit rates | |
| EP1817918A1 (en) | Method and apparatus for encoding/decoding multi-layer video using dct upsampling | |
| Misu et al. | Real-time implementation of UHDTV video coding system with super-resolution techniques | |
| Zhang et al. | Color image compression based on directional all phase biorthogonal transform | |
| WO2023202097A1 (en) | Loop filtering method, video coding method and apparatus, video decoding method and apparatus, medium, program product, and electronic device | |
| JP4385928B2 (en) | Scalable video encoding / decoding method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 08734257 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 08734257 Country of ref document: EP Kind code of ref document: A1 |