WO1995022228A1 - Method and device for encoding image signal - Google Patents
Method and device for encoding image signal Download PDFInfo
- Publication number
- WO1995022228A1 WO1995022228A1 PCT/JP1995/000202 JP9500202W WO9522228A1 WO 1995022228 A1 WO1995022228 A1 WO 1995022228A1 JP 9500202 W JP9500202 W JP 9500202W WO 9522228 A1 WO9522228 A1 WO 9522228A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- quantized
- image signal
- quantization
- quantized data
- 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
- H04N19/10—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
- H04N19/124—Quantisation
-
- 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
- H04N19/124—Quantisation
- H04N19/126—Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
-
- 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
- H04N19/132—Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
-
- 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
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- 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
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/18—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a set of transform coefficients
-
- 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
Definitions
- the present invention relates to an image signal encoding method and apparatus used in various application fields such as a digital storage medium, a television broadcast, and a communication network, and particularly to a moving image sequence converted into digital data in a compressed format. is suitably applied to an image signal coding method and apparatus used to process signals to allow transmission or storage (bACKGROUND
- DCT discrete cosine transform
- the video specialization group decided that the DC ⁇ coefficient in the block It employs a quantization method that can be represented by a set of manageable-size symbols. This process is performed by dividing the DCT coefficients by a constant q (hereinafter referred to as quantization scale). By this processing, each DCT coefficient is approximated to the integer value closest to each DCT coefficient (hereinafter referred to as quantized DCT coefficient). Note that by multiplying the quantized DCT coefficient obtained in this way by the quantization scale q and performing inverse operation, the DCT coefficient before encoding can be restored.
- the quantization scale q is set too large to increase the coding efficiency during quantization, the approximation to the DCT coefficients will be insufficient, and as a result, the eye will not be visible in the decoded image. Visible noise comes to appear. This noise is commonly known as mosquito noise.
- the low-frequency coefficient generally has a characteristic that most of the energy is easily concentrated, and the high-frequency coefficient has a characteristic that the human perception sensitivity tends to be lower than the low-frequency coefficient.
- a typical characteristic of an image signal is that a change between successive frames is small except for a certain kind of movement of an object in a screen.
- a technique generally called motion compensation is used as a coding technique focusing on this characteristic.
- the current image included in each block is represented by a displacement vector from an image encoded earlier.
- This motion compensation method is effective for further reducing the number of bits required to encode a moving image.
- images coded in this manner sometimes had some visible sharp edges between the displaced procks and the procks. This is commonly referred to as block noise.
- the Moving Image Specialization Group describes these two types (discrete cosine transform + quantization and motion compensation) as MPEG-1 and MP2 video coding standards.
- EG 2 is used in each system.
- MPEG1 and MPEG2 have a rate control mechanism. This rate control mechanism controls the allocation of the number of bits to each block that composes the moving image.
- the rate control mechanism sets the size of the quantization scale used in the quantizer for each of the four DCT blocks, that is, for each macroblock corresponding to 16 ⁇ 16 pixels.
- the quantized DCT coefficients generated by the quantizer can be represented as a sequence of zero values followed by a non-zero value.
- Run-length coding is used in MPEG 1 and MPEG 2 to encode such quantized DCT coefficients, because large data compression can be achieved for redundant inputs characterized by long clusters of the same symbol. Have been.
- the quantization scale is reduced, the series of zero values becomes shorter, and the number of bits for coding the quantized DCT coefficient increases. ⁇ Also, the relatively low energy DCT coefficient affects the image quality.
- the present invention has been made in view of the above points, and an object of the present invention is to propose an image signal encoding method and apparatus capable of further improving the image quality of an image sequence as compared with the related art.
- the present invention provides an image signal encoding method for encoding an image signal, in which a dynamic range of the image signal is compressed for each pixel. And a step of orthogonally transforming the image signal of which the dynamic range has been compressed into coefficient data for each unit block, and converting the coefficient data into quantized data to generate quantized data. And a step of converting the same.
- an image signal encoding method for encoding an image signal a step of orthogonally transforming the image signal for each unit block to convert the image signal into coefficient data; Generating the first quantized data by quantizing based on the quantization scale, and converting the coefficient data to the first quantized data only for the coefficient data in which the value of the first quantized data is substantially non-zero.
- one Le Ri by scale in the c also present invention is to include a step of generating a second quantization data, the image signal dies Na Mi click
- the range is compressed for each pixel.
- the dynamic range is generated by transforming the compressed image signal into coefficient data by orthogonally transforming the image signal for each unit block, and quantizing the coefficient data.
- Quantized data In the image signal decoding method for decoding the image data, the quantized data is inversely quantized to generate coefficient data, and the coefficient data is inversely orthogonally transformed for each unit block to generate a decoded image, and the decoding is performed.
- the dynamic range of the decoded image signal is extended for each pixel.
- a dynamic range compression unit for compressing a dynamic range of the image signal for each pixel, Orthogonal transform means for orthogonally transforming the image signal of which the mix range has been compressed into coefficient data by orthogonal transform for each unit block, and quantizing means for quantizing the coefficient data to generate quantized data And was prepared.
- an image signal encoding device for encoding an image signal
- First quantization means for quantizing based on the quantization scale to generate first quantized data
- a second quantum that quantizes based on a second quantization scale smaller than the first quantization scale to generate second quantized data And means for conversion.
- the dynamic range of the image signal is compressed for each pixel, and the image signal having the compressed dynamic range is orthogonally transformed for each unit block to obtain coefficient data.
- An image signal decoding device for decoding the quantized data generated by transforming and quantizing the coefficient data, an inverse quantization means for inversely quantizing the quantized data to generate coefficient data, Inverse orthogonal transform means for inversely orthogonally transforming data for each unit block to generate a decoded image, and dynamic range extension for extending the dynamic range of the decoded image signal for each pixel Means.
- the step of generating the encoded image data includes the step of compressing the dynamic range of the image signal for each pixel.
- the step of generating encoded image data is a step of orthogonally transforming an image signal into coefficient data by unit block. And a step of quantizing the coefficient data based on the first quantization scale to generate first quantized data; and the step of converting the coefficient data to a value in which the first quantized data is substantially non-zero. And generating the second quantized data by quantizing only the coefficient data based on the second quantization scale smaller than the first quantization scale.
- Fig. 1 is a block diagram showing the overall configuration of an encoding / decoding system using the dynamic range compression / expansion method.
- FIG. 2 is a block diagram showing an embodiment of a quantizer using a multi-stage quantization and refinement method.
- FIG. 3 is a block diagram showing another embodiment of the quantizer using the multi-stage quantization and refinement method.
- Fig. 4 is a n-chart showing the quantization procedure by the multi-stage quantization and refinement method of Fig. 3.
- FIG. 5 is a block diagram showing an embodiment of a quantizer using a single-stage quantization and refinement method.
- FIG. 6 is a flowchart showing a quantization procedure according to the single-stage quantization and refinement method
- FIG. 7 is a block diagram showing a configuration of a decoder circuit.
- the quality of an encoded video sequence can be improved by reducing the difficulty of the source image sequence and by refining the quantization levels used in the encoder.
- the difficulty of the video sequence is due to the minimum amount of data required to represent the video sequence when encoded using a fixed set of parameters, or the coding sequence.
- a video sequence that is relatively difficult is represented by an encoded sequence that is longer than a sequence that is relatively difficult.
- a sequence that is difficult is composed of many sharp edges and screen cuts. At the pixel level, these many sharp edges and screen cuts can be interpreted as large differences between neighboring pixels in the same frame or between corresponding pixels in adjacent frames. be able to.
- the difficulty of this sequence can be reduced by reducing the pixel-level dynamic range of the video sequence.
- a linear mapping is used to simplify the implementation.
- the use of a linear mapping can reduce the difference in pixel level between adjacent pixels, and can smooth out sharp edges.
- images encoded in this way can be restored by inverse mapping.
- the pixel-level dynamic range is restored with a slight loss of image texture smoothness, but this loss cannot be detected by human perception. Therefore, compression of image information by linear mapping is effective.
- the discrete cosine transform (DCT) coefficient corresponding to the shortened portion can be reduced by the reduction. It can be used for refinement.
- the refinement of DCT coefficients is achieved, for example, by reducing the quantization scale and increasing the number of coding bits little by little for each macroblock described above to refine the non-zero ⁇ quantization coefficients. can do.
- the number of coded bits is increased from the stage where the difficulty of the source sequence was reduced by reducing the dynamic range.
- an encoding method using these two methods is proposed.
- a method called "multi-stage quantization and refinement method” is introduced in the present application, and a method of improving the accuracy of the non-zero ⁇ quantization coefficient will be described.
- This method is a method of refining non-zero quantized coefficients without reducing the quantization scale and increasing the run length of the zero quantized coefficients.
- Fig. 1 shows an overview of an encoding / decoding system using the dynamic range compression / expansion method that reduces the difficulty of the source video sequence by reducing its dynamic range. explain.
- the dynamic range of each pixel in the source video sequence is sequentially linearly compressed by the dynamic range compression circuit.
- X represents the pixel value of the source image
- y represents the pixel value of the image to be encoded
- a represents a constant (where a ⁇ l).
- the encoder circuit 20 encodes data by performing DCT, quantization, variable length encoding, and the like, which will be described later.
- the encoded data output from the encoder circuit 20 is recorded on the disc 21 via a recording circuit (not shown).
- the encoded data recorded on the disc 21 is read out via a reproducing circuit (not shown) and supplied to the decoder circuit 22.
- the transmission path for transmitting the encoded data may be a television broadcast or a communication network instead of the disk 21.
- the decoder circuit 22 is configured as shown in FIG. Receive buffer 4 1
- the encoded data read out from the disk 21 via the reproducing circuit is temporarily stored in the receiving buffer 41.
- the encoded data read from the reception buffer is subjected to variable length decoding by the variable length decoding circuit 42.
- the quantization scale used for encoding is also decoded.
- the inverse quantizer 43 inversely quantizes the quantized DCT coefficient based on the decoded quantization scale and supplies the result to the inverse DCT circuit 44.
- the inverse DCT circuit 44 performs an inverse DCT on the inversely quantized DCT coefficient and outputs a decoded image.
- the decoder circuit 22 includes a configuration of a predictor and the like in order to correspond to the coded data subjected to the prediction coding, but the description is omitted for simplicity.
- the decoder circuit 22 decodes the encoded data as described above and supplies the decoded data to the dynamic range extending circuit 3, which uses the dynamic range of each pixel as the basis. It is expanded to the size of and restored.
- u represents the pixel value of the decoded image
- V represents the image value viewing
- a represents a constant (where a ⁇ 1).
- the quantization refinement method introduced to enhance the accuracy of the non-zero quantized DCT coefficient will be described.
- the multi-stage quantization refinement method will be described, and then the single-stage quantization refinement method will be described.
- FIG. 2 shows the multi-stage quantization and refinement circuit included in the encoder circuit 20.
- Multi-stage The quantization refinement circuit is composed of a discrete cosine transform (DCT) circuit 4 and a quantization circuit.
- DCT discrete cosine transform
- the input data is converted into DCT coefficients by the DCT circuit 4 and supplied to the quantization circuit.
- the quantization circuit inputs the DCT coefficient to the quantizer 5 q having the largest quantization scale among the n + 1 quantizers 5 q to 5 q-n, and quantizes the DCT coefficient by the quantization scale q. Become This quantized DCT coefficient is given to the non-zero position identification circuit 6.
- the quantization scale q at this time is set as follows.
- the VLC circuit 10 performs run-length coding to convert the input quantized DCT coefficients into data represented by a pair of a zero value length and a subsequent non-zero value, and converts the data into a predetermined transform. It generates variable-length encoded data for transmission by using the table and outputs it to the transmission buffer 11.
- the transmission buffer 11 temporarily stores an encoding code for transmission and outputs it at a predetermined timing. Therefore, the transmission buffer 11 temporarily stores the quantized, run-length coded, and variable-length coded data corresponding to the block before the DCT coefficient to be quantized. become.
- the quantization scale controller 12 monitors the accumulated amount of data stored in the buffer 11 and performs quantization from now on so that the transmission buffer 11 does not overflow and underflow.
- the quantization scale of each quantizer is also controlled in accordance with the quantization scale q.
- the above-described rate control mechanism is realized by such quantization scale control.
- the non-zero n position identification circuit 6 identifies the position of the quantized coefficient having a value other than the zero ⁇ value among the quantized DCT coefficients, and uses the position information as n masking circuits 9 q—1, 9 q ⁇
- these n masking circuits 9 q— 1 and 9 q ⁇ 2 9 q— n form a set (A) of DCT coefficients output from the DCT circuit 4.
- A The DCT coefficients of the set of non-ze ⁇ values ⁇ Z ⁇ are not changed, and the other DCT coefficients, that is, the DCT coefficients whose quantized DCT coefficients are zero by the quantizer 5q, are Replaces each value with zero and supplies it to each quantizer 5 q ⁇ 1, 5 q ⁇ 25 q — n. In the following description, it is said that replacing the input value with the zero value is masked.
- the n quantizers 5 q-1 and 5 q-25 q-n requantize the input DCT coefficients at the quantization scales q-1 and q-2 q-n, respectively. Since the zero-valued DCT coefficient also has a quantized DCT coefficient of zero after quantization, this requantization effectively requantizes only the set of non-Z D values ⁇ Z ⁇ , improving accuracy. It was done.
- the quantized DCT coefficients quantized by these quantizers 5 q, 5 q ⁇ 1, 5 q ⁇ 2 5 d ⁇ n are output from the corresponding check circuits 7 q, 7 q ⁇ 1, 7 q q2... 7 Input to q-n.
- the check circuits 7 q, 7 q-1, 7 q-2... 7 q-n check the quality of the refined quantized coefficients. This includes monitoring the increase in the number of bits required for the quantization refinement scheme and checking the quality of the refined quantized DCT coefficients.
- the quality inspection of the refined quantized DCT coefficient is performed by checking the gain of the signal-to-noise ratio between the actual DCT coefficient (DCT coefficient before quantization obtained from the DCT circuit 4) and the quantized DCT coefficient, An inspection of the mean square error between the DCT coefficient and the quantized DCT coefficient or an inspection of the deviation of the quantized DCT coefficient from the actual DCT coefficient is performed, and the obtained inspection result is supplied to the selection circuit 8.
- each test circuit has the same configuration as VLC 10 internally, and performs provisional run-length coding and variable-length coding on each quantized DCT coefficient, and obtains the obtained code. The length information is supplied to the selection circuit 8.
- the selection circuit 8 selects a quantized DCT coefficient corresponding to the required quality and a corresponding quantization scale. Furthermore, the code length obtained from the test circuit 7 q is compared with the code length obtained from each of the other test circuits, and the code length corresponding to the quantized DCT coefficient selected based on the required quality is compared. If the increase in the code length exceeds the maximum allowable code length increase, the quantized DC coefficient corresponding to the quality close to the required quality within the predetermined maximum allowable code length increase range is used. Output the quantization scale. Further, the selection circuit 8 may select the quantized DCT coefficient corresponding to the one with the smallest quantization scale within the maximum allowable code increase length range. In the case of, as described above, feedback is applied to the setting of the quantization scale q at the time of the quantization of the subsequent DCT coefficient by the rate control mechanism. Quantized DCT coefficients are selected based on the.
- the VLC circuit performs run-length coding and variable-length coding on the quantized DCT coefficients output from the selection circuit 8 and outputs the obtained coding code to the transmission buffer 11 ⁇ Also, the VLC circuit is selected.
- the quantization scale corresponding to the quantized DCT coefficient is also encoded as an encoding code.
- the transmission buffer 11 temporarily stores the code and outputs the coded data at a predetermined timing.
- the masking circuit is provided before the quantizer.
- the masking circuit is provided after the quantizer, and the quantized DCT coefficient by the quantizer 5 q is zero. It is also possible to mask the quantized DCT coefficient corresponding to the value position.
- the masking circuit is provided for each of the quantizers, the masking circuit is made a single unit, and after performing the masking, The DCT coefficient may be distributed to the quantizer.
- the quality is checked by comparing the actual DCT coefficient with each quantized DCT coefficient.
- the quantized DCT coefficient at a certain quantization scale is one step smaller than the quantized DCT coefficient.
- the quantized DCT coefficients at the time when there is no difference in the quality improvement may be selected by comparing the quantized DCT coefficients with the quantized scale.
- the circuit scale is larger than that of a multi-stage quantization / definition circuit 2 described later, but the processing speed is improved by the parallel processing of quantization.
- a set ⁇ A ⁇ of DCT coefficients from the DCT circuit 4 is stored in the memory 13.
- the memory controller 14 controls writing and reading of the memory 13.
- the DCT coefficient read from the memory 13 is supplied to the masking circuit 9.
- the masking circuit 9 outputs the supplied DCT coefficient as it is, and when the quantizer 5 performs requantization.
- the DC ⁇ coefficients other than the set of non-zero values ⁇ detected by the non-zero position detection circuit 6 are replaced with ze in and output.
- the quantization circuit 5 quantizes the DCT coefficient based on the quantization scale set by the quantization scale controller n-17.
- the quantized DCT coefficient output from the quantizer 5 is sequentially overwritten and stored in the memory 15 while the DCT coefficient of the same block is repeatedly quantized.
- the memory controller 16 controls writing and reading of the memory 15.
- the quantized DCT coefficient read from the memory 15 is output via the VLC 10 and the transmission buffer 11.
- Inspection circuit 7 is similar to inspection circuits 7q, 7q-1 and 7q-27q-n in Fig. 2 in that the quality of the quantized DCT coefficients and the code length of the quantized refinement are reduced. Monitor the increase and, if necessary, request the quantization scale controller 17 to switch the quantization scale. The inspection circuit 7 sends a signal requesting the memory controller 14 to reread the DCT coefficient from the memory 13, and outputs the signal to the memory controller 14. For (6), it is required to read out the refined quantized DCT coefficients from memory (15).
- the quantization scale controller 17 configures the above-described rate control mechanism in cooperation with the transmission buffer 11 similarly to the quantization scale controller 12 in FIG. 2, and sets the initial quantization scale q. Set. Further, the quantization scale controller 17 sets a plurality of different quantization scales sequentially according to a request for switching the quantization scale from the inspection circuit 7.
- FIG. 4 shows the results of the encoder circuit 20 using the multi-stage quantization and refinement circuit (part 2).
- the multi-stage quantization refinement processing procedure will be described.
- the processing enclosed by the dotted line in the figure (steps SP5 to SP8) is specific to the multi-stage quantization and refinement method.
- step SP 1 when the quantization process is started (step SP 1), an initial quantization scale q is generated by the rate control mechanism (step SP 2), and the 'DCT coefficient read from the memory 13 is quantized. Supplied to the gasifier 5.
- the quantizer 5 sequentially quantizes the DCT coefficients using the initial quantization scale q (step SP3).
- step SP 4 the quantized DCT coefficient and the generated quantization scale are set.
- the quantization refinement process when the quantization refinement process is selected, the position of the set of non-ze ⁇ quantized DCT coefficients is identified and stored by the non-Z The process proceeds to the refinement processing of the quantized DCT coefficient (step SP5).
- step SP 6 the same DCT coefficient as that quantized in step SP 3 is read out again from the memory 13, and the masking circuit 9 reads the position stored in the non-zero position identification circuit 6. Mask the DCT coefficient at position D using the information.
- the quantizer 5 requantizes the DCT coefficient at the position identified as a non-ze D value using the quantization scale q-1 of one smaller quantization scale.
- step SP7 the quality of the quantized DCT coefficients refined by the quantizer 5 and the amount of increase in the code length are inspected by the inspection circuit 7, and it is determined whether the quality needs to be improved. I do.
- step SP8 If a positive result is obtained here (that is, when it is necessary to improve the quality of the refined quantized DCT coefficient), the process goes to step SP8 to change the value of the parameter X. Increase and return to step SP5. This increases the value of parameter X from 1 to 2.
- the DCT coefficient at the position where the quantized DCT coefficient becomes non-zero is quantized again by a further smaller quantization scale q ⁇ 2, and the quality of the quantized DCT coefficient is checked. Check. Thereafter, if necessary, re-quantization with a smaller quantization scale q—3, q—4,..., Q—n is repeated.
- step SP5 to step SP8 is repeated until a negative result is obtained in step SP7 (that is, there is no need to improve the quality of the refined quantized DCT coefficient).
- the inspection circuit 7 instructs the memory controller 16 to read the refined DCT coefficients from the memory 15.
- the latest quantized DCT coefficient is read out from the memory 15 and a series of multi-stage quantization refinement processing is completed.
- the inspection at step SP7 includes monitoring the increase in the number of bits required for the refinement of quantization and inspecting the quality of the refined quantized DCT coefficients as described above.
- the basis for quality inspection of refined quantized DCT coefficients is the gain of the signal-to-noise ratio between the actual DCT coefficients and the refined quantized DCT coefficients, or the actual DCT coefficients and the refinement. Mean square error with the quantized DCT coefficient, or the deviation of the refined quantized DCT coefficient from the actual DCT coefficient.
- the masking circuit is provided before the quantizer, but when the masking circuit is provided after the quantizer and quantized using the initial quantization scale q, the quantized DCT coefficient corresponding to the zeroed position may be replaced with a zero value.
- the quality is checked by comparing the actual DCT coefficient with each quantized DCT coefficient.
- the quantized DCT coefficient at a certain quantization scale is one step smaller than the quantized DCT coefficient.
- the quantized DCT coefficients may be sequentially compared with the quantized DCT coefficients according to the quantization scale, and the quantized DCT coefficients at the time when there is no difference in the quality improvement may be selected.
- the processing time becomes longer due to the necessity of the repetitive operation, but the circuit scale can be reduced.
- the position where the DCT coefficient becomes zero is masked, and the DCT coefficient is substantially reduced. Only the DCT coefficients at non-zero positions will be requantized using the smaller quantization scale q _ 1.
- Q— 2,..., Q _ n so the smaller quantization scale Does not affect the length of the sequence of zero coefficients. It is important to mask the quantization coefficient at this zero position. Otherwise, it is possible that some of the preceding zero coefficients will become non-zero due to the reduced quantization scale. That is, it can be seen that the accuracy of the non-zero coefficient can be improved without affecting the length of the series of zero coefficients.
- the difficulty of the image data is compressed by linear mapping, then the discrete cosine transform is performed, and when the coefficient data is quantized by the initial quantization scale q, the value of the quantized DCT coefficient is reduced.
- the non-zero DCT coefficients are re-quantized by the quantization scale q-1 and q-2 q--n which are smaller than the initial quantization scale q. It is possible to realize good coding.
- the quantization with the quantization scale q eliminates (that is, masks) the quantized DCT coefficients that do not become 0, thereby reducing the quantization to zero.
- the accuracy of the non-zero coefficient can be further improved without affecting the length of the series of coefficients. As a result, it is possible to realize coding in which mosquito noise / block noise is hardly generated.
- FIG. 5 shows a single-stage quantization and refinement circuit included in the encoder circuit 20.
- the single-stage quantization and refinement circuit also includes a discrete cosine transform (DCT) circuit 4 and a quantization circuit.
- DCT discrete cosine transform
- the initial quantization scale q and one quantization scale q-c which is smaller than the initial quantization scale q are used to eliminate the repetition of the DCT coefficient refinement processing. And two quantization scales are used.
- the data set ⁇ A ⁇ converted to DCT coefficients is given to the quantizer 31q, and quantized with the initial quantization scale q given from the quantizer scale controller 34. I do.
- the initial quantization scale q is set by the rate control mechanism as in the case of multi-stage quantization and refinement.
- a position of the quantized DCT coefficient having a non-zero coefficient value is identified by the non-zero position identification circuit 32, and this is given to the masking circuit 33.
- the masking circuit 33 generates a set ⁇ Z ⁇ (that is, ⁇ A ⁇ ) of non-zero quantized DCT coefficients among the quantized DCT coefficients ⁇ A ⁇ quantized by the quantizer 31qc.
- ⁇ ⁇ Z ⁇ is output by quantizing with substantially smaller quantization scale q — C.
- C is a constant that gives a fixed decrement greater than 0 and less than q
- c is set to c, for intra mode and C2 for inter mode.
- Fig. 6 shows the procedure of the single-stage quantization and refinement processing by the encoder circuit 20.
- the processing (steps SP14 to SP18) of the portion enclosed by the dotted line in the figure is single. This is a process unique to the stage ⁇ refining process.
- an initial quantization scale q is generated by the rate control mechanism in the encoder (step SP11), and supplied to the quantizer 31q. Is done.
- the quantizer 3 1 q uses the initial quantization scale q
- the T coefficients are sequentially quantized (step SP12).
- step SP13 the quantized DCT coefficient and the generated quantization scale are set.
- step SP 1 4 If this quantization fine Seika process is selected for in step SP 1 4 - position of the set of Hize D quantized DCT coefficients is identified stored.
- step SP 15 it is determined whether the encoding mode is less than the inter mode or the intra mode, and the quantization scale controller 34 determines the fixed reduction value c of the quantization scale of the quantizer 31 qc. Is set to a value corresponding to each mode.
- the intra mode is applied to a block coded using only DCT and quantization, while the inter mode is used for a block coded using both motion compression and DCT and quantization. Applied to
- MPEG 1 and MPEG 2 it is well known that coding is performed using so-called I-pictures, ⁇ -pictures, and ⁇ ⁇ ⁇ ⁇ -pictures. All macroblocks in the I-picture are intra-coded.
- a picture generally includes both intra-coded macroblocks and forward-predicted macroblocks.
- ⁇ ⁇ ⁇ ⁇ Pictures generally consist of an intra-coded macroblock, a forward-predicted macroblock, and a backward-predicted macroblock. Includes locks and bidirectional (macroblock) predicted macroblocks.
- the intra mode corresponds to the Mac alpha block that is intra-coded for I, P, and B pictures, and the inter mode corresponds to the other macro blocks. .
- the single-stage quantization refinement method has a simple configuration and can reduce the processing time because the repetitive operation is omitted when refining the DCT coefficient.
- the masking circuit is provided after the quantizer.However, when the masking circuit is provided before the quantizer and quantized using the initial quantization scale Q, the quantized DCT coefficient corresponding to the zeroed position may be replaced with a zero value.
- the difficulty of the image data is compressed by linear mapping, then the discrete cosine transform is performed, and when the coefficient data is quantized by the initial quantization scale q, the quantized DCT coefficient is DCT coefficients with non-zero TP values are quantized again with a fixed quantization scale q-c that is smaller than the initial quantization scale q, so that encoding with high quantization accuracy and good image quality can be realized. .
- mosquito noise and block noise can be further reduced as compared with the related art.
- the DCT coefficients of the image data subjected to the discrete cosine transform are quantized and refined. It can also be applied to the case where the coefficients of the image data converted by the quantization are quantized and refined. In this case, the same effect can be obtained.
- the dynamic range is compressed for each pixel, and then the orthogonal transform is performed. Further, the coefficient data obtained by the orthogonal transform is subjected to the first quantization scale. When quantizing with, the coefficient data for which the value of the first quantized data is non-zero is quantized with the second quantization scale having a smaller quantization scale than the first quantization scale.
- An image signal encoding method and apparatus include a disc creating apparatus for creating a disc on which an MPEG 1 or MPEG 2 bit stream is recorded, and an MPEG 1 or MPEG 2 bit stream. It can be used for CATV and digital satellite broadcasting distribution equipment that transmits video.
- the image signal decoding method and apparatus of the present invention include a disk reproducing apparatus for reproducing a disk on which an MPEG 1 or MPEG 2 bit stream is recorded, an MPEG 1 or MPEG 2 bit stream, It can be used for CATV and digital satellite broadcast receivers that transmit a stream.
- the image recording medium of the present invention can be used as a recording medium on which a bit stream of MPEG 1 or MPEG 2 is recorded.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP95908375A EP0699003A4 (en) | 1994-02-15 | 1995-02-14 | METHOD AND DEVICE FOR ENCODING SIGNAL SIGNALS |
KR1019950704558A KR960702247A (ko) | 1994-02-15 | 1995-02-14 | 화상 신호 부호화 방법 및 장치(Method and device for encoding image signal) |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP4040494 | 1994-02-15 | ||
JP6/40404 | 1994-02-15 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1995022228A1 true WO1995022228A1 (en) | 1995-08-17 |
Family
ID=12579733
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP1995/000202 WO1995022228A1 (en) | 1994-02-15 | 1995-02-14 | Method and device for encoding image signal |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP0699003A4 (ja) |
KR (1) | KR960702247A (ja) |
CN (1) | CN1123593A (ja) |
WO (1) | WO1995022228A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0783230A2 (en) * | 1995-11-29 | 1997-07-09 | Samsung Electronics Co., Ltd. | Transform coding of arbitrarily shaped image segments |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1359766A3 (en) * | 1997-02-14 | 2005-02-16 | AT&T Corp. | A method of generating a dequantized dc luminance or dc chrominance coefficient |
US7269219B1 (en) | 1997-02-14 | 2007-09-11 | At&T Corp. | Non-linear quantizer for video coding |
CN1204751C (zh) * | 1999-04-13 | 2005-06-01 | 松下电器产业株式会社 | 编码数据变换方法、再编码方法、再编码***以及数据记录媒体 |
AU1242901A (en) * | 1999-10-29 | 2001-05-14 | Sensormatic Electronics Corporation | Method of and system for dynamic range compression and expansion |
GB0110748D0 (en) * | 2001-05-02 | 2001-06-27 | Apical Ltd | Image enhancement methods and apparatus therefor |
JP4347687B2 (ja) * | 2001-07-26 | 2009-10-21 | アイピージー・エレクトロニクス・503・リミテッド | スケーラブルでないコード化された映像信号からスケーラブルなコード化された映像信号を生成するための方法とデバイス |
US20210368177A1 (en) * | 2018-05-17 | 2021-11-25 | Amimon Ltd. | Joint Source Channel Coding with H.265 Video Compression |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61147692A (ja) * | 1984-12-21 | 1986-07-05 | Nec Corp | デ−タ圧縮伸張方式 |
JPH01137785A (ja) * | 1987-11-24 | 1989-05-30 | Sony Corp | 高能率符号化装置 |
JPH04156794A (ja) * | 1990-10-19 | 1992-05-29 | Matsushita Electric Ind Co Ltd | 符号化装置 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5166987A (en) * | 1990-04-04 | 1992-11-24 | Sony Corporation | Encoding apparatus with two stages of data compression |
US5150208A (en) * | 1990-10-19 | 1992-09-22 | Matsushita Electric Industrial Co., Ltd. | Encoding apparatus |
JP3227674B2 (ja) * | 1991-06-27 | 2001-11-12 | ソニー株式会社 | 動画像符号化装置および方法 |
-
1995
- 1995-02-14 KR KR1019950704558A patent/KR960702247A/ko not_active Application Discontinuation
- 1995-02-14 WO PCT/JP1995/000202 patent/WO1995022228A1/ja not_active Application Discontinuation
- 1995-02-14 EP EP95908375A patent/EP0699003A4/en not_active Withdrawn
- 1995-09-21 CN CN95190090A patent/CN1123593A/zh active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61147692A (ja) * | 1984-12-21 | 1986-07-05 | Nec Corp | デ−タ圧縮伸張方式 |
JPH01137785A (ja) * | 1987-11-24 | 1989-05-30 | Sony Corp | 高能率符号化装置 |
JPH04156794A (ja) * | 1990-10-19 | 1992-05-29 | Matsushita Electric Ind Co Ltd | 符号化装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP0699003A4 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0783230A2 (en) * | 1995-11-29 | 1997-07-09 | Samsung Electronics Co., Ltd. | Transform coding of arbitrarily shaped image segments |
EP0783230A3 (en) * | 1995-11-29 | 1999-10-20 | Samsung Electronics Co., Ltd. | Transform coding of arbitrarily shaped image segments |
Also Published As
Publication number | Publication date |
---|---|
CN1123593A (zh) | 1996-05-29 |
EP0699003A1 (en) | 1996-02-28 |
KR960702247A (ko) | 1996-03-28 |
EP0699003A4 (en) | 1999-10-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP0987899B1 (en) | Adaptive variable-length coding method for video data | |
US5650782A (en) | Variable length coder using two VLC tables | |
JP3085024B2 (ja) | 画像再圧縮器及び画像記録装置 | |
EP0677969B1 (en) | Moving picture coding method and an apparatus therefor | |
JPH03204290A (ja) | 映像信号符号化方法及びその装置 | |
JPH05505077A (ja) | イメージフレームのランダムスキャンニングを用いるディジタルイメージコーディング法 | |
JPH0686262A (ja) | 画像符号化装置 | |
JPH06125543A (ja) | 符号化装置 | |
WO1995022228A1 (en) | Method and device for encoding image signal | |
WO2000018129A1 (en) | Video signal encoding method, video signal encoder, and program recorded medium | |
JP2963958B2 (ja) | 高能率符号化装置 | |
KR100219133B1 (ko) | 변환부호화시스템의 변환계수선택방법및그장치 | |
JPH10164597A (ja) | 符号化方法およびその方法を用いた符号化装置 | |
JPH1175198A (ja) | 画像信号圧縮装置及び方法、並びに記録媒体 | |
JPH07107464A (ja) | 画像符号化装置および復号化装置 | |
JP4000581B2 (ja) | 画像符号化装置および方法 | |
JPH08130735A (ja) | 画像データ符号化/復号化装置 | |
KR100209879B1 (ko) | 가변양자화를 이용한 부분적 화질강조장치 | |
JPH08340536A (ja) | 画像符号化装置 | |
KR0128862B1 (ko) | 동영상부호화장치의가변길이부호화장치 | |
JPH1175197A (ja) | 画像信号圧縮装置及び方法、並びに記録媒体 | |
JPH0622212A (ja) | 特殊効果発生装置 | |
JPH08275108A (ja) | 動画像の圧縮記録装置 | |
JPH09261634A (ja) | 映像データ圧縮装置およびその方法 | |
JP2006136010A (ja) | 映像データ圧縮装置およびその方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): CN JP KR US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): DE FR GB |
|
WWE | Wipo information: entry into national phase |
Ref document number: 95190090.0 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1995908375 Country of ref document: EP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
ENP | Entry into the national phase |
Ref document number: 1996 530302 Country of ref document: US Date of ref document: 19960129 Kind code of ref document: A |
|
WWP | Wipo information: published in national office |
Ref document number: 1995908375 Country of ref document: EP |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 1995908375 Country of ref document: EP |