WO2018014301A1 - 视频编码方法及装置 - Google Patents

视频编码方法及装置 Download PDF

Info

Publication number
WO2018014301A1
WO2018014301A1 PCT/CN2016/090884 CN2016090884W WO2018014301A1 WO 2018014301 A1 WO2018014301 A1 WO 2018014301A1 CN 2016090884 W CN2016090884 W CN 2016090884W WO 2018014301 A1 WO2018014301 A1 WO 2018014301A1
Authority
WO
WIPO (PCT)
Prior art keywords
prediction mode
discrete cosine
cosine transform
value
transform
Prior art date
Application number
PCT/CN2016/090884
Other languages
English (en)
French (fr)
Inventor
张金雷
邹天玱
王妙锋
石中博
王世通
薛东
罗巍
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to PCT/CN2016/090884 priority Critical patent/WO2018014301A1/zh
Priority to CN201680054222.9A priority patent/CN108028938A/zh
Publication of WO2018014301A1 publication Critical patent/WO2018014301A1/zh

Links

Images

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
    • 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 the field of image processing technologies, and in particular, to a video encoding method and apparatus.
  • the digital video compression format H.264 standard and the High Efficiency Video Coding (HEVC) standard are two video coding schemes currently used, and the HEVC standard may also be referred to as the H.265 standard.
  • the prediction is to compress the image.
  • the prediction part obtains a prediction value for the current coded image by removing spatial correlation and temporal correlation of the video content, and obtains a residual of the original image and the predicted image.
  • the residual is transformed and quantized.
  • the information such as the quantized coefficients of the prediction residual and the prediction mode are entropy encoded to form a code stream output.
  • the prediction mode may be divided into inter prediction and intra prediction.
  • the inter prediction is to predict the next frame image by using the temporally adjacent previous frame
  • the intra prediction mode is to predict the current frame image by using the current intraframe spatial correlation.
  • the inter prediction mode can be predicted by using video adjacent frame information, and the compression ratio is higher.
  • the compression ratio of the intra prediction mode is lower than that of the inter prediction mode, but the spatial redundancy of adjacent blocks inside the current frame can be removed.
  • each prediction mode has different prediction values and prediction residuals, and the entropy-encoded code rate (R) and reconstruction video of each prediction mode.
  • the distortion (D) is different.
  • Rate Distortion Optimization (RDO) is used to calculate the cost of each mode, and the mode with the smallest cost function is selected as the optimal mode.
  • RDO uses R and D
  • Two-quantity design cost function calculates the cost value J of each prediction mode, as shown in equation (1):
  • the prior art outputs encoded data of an optimal prediction mode and preserves reconstruction residuals and reconstructed video of the optimal prediction mode.
  • the reconstruction residual is transformed, quantized, inverse quantized, and inverse transformed to obtain a reconstruction residual, and the distortion D is obtained according to the difference between the prediction residual and the reconstruction residual.
  • the prediction residual is spatial domain data, and the transform operation converts the spatial domain data into frequency domain data. Therefore, the prior art is also called a spatial domain distortion calculation method.
  • the prior art it is necessary to transform, quantize, inverse quantize, and inverse transform the prediction residuals of each prediction mode to select an optimal prediction mode.
  • the complexity of the prior art is high, wherein the complexity of the transform and the inverse transform increases exponentially with the transform size, and the transform size of the HEVC standard is increased compared to H.264, and the number of prediction modes in the HEVC standard is also compared.
  • the H.264 standard has grown exponentially, and as a result, the prior art has led to a dramatic increase in the complexity of the HEVC standard.
  • the transform and the inverse transform are modules with high power consumption.
  • the prediction residuals of each prediction mode are transformed and inverse transformed, and the power consumption is too high.
  • Embodiments of the present invention relate to a video encoding method and apparatus, which solve the problem of high complexity in the prior art and large power consumption in hardware implementation.
  • an embodiment of the present invention provides a video encoding method, which includes performing discrete cosine transform on a prediction residual of a current prediction mode to obtain a transform coefficient, where the predicted image is based on an inter prediction mode or an intraframe.
  • the prediction mode predicts that the inter prediction mode predicts the next frame image by using the temporally adjacent previous frame image, and the intra prediction mode predicts the current frame image by using the current intraframe spatial correlation.
  • the transform coefficients are quantized to obtain quantized coefficients.
  • the quantized coefficients are inverse quantized to obtain inverse quantized coefficients.
  • the distortion value of the current prediction mode is obtained from the difference between the transform coefficient and the inverse quantization coefficient.
  • the current prediction mode cost value is obtained according to the code rate value and the distortion value of the current prediction mode.
  • the prediction mode with the lowest generation value among the multiple prediction modes is selected as the optimal prediction mode.
  • Optimal prediction mode The inverse quantization coefficient of the equation is inverse discrete cosine transform to obtain the reconstructed residual.
  • the discrete cosine transform converts the spatial domain data into frequency domain data.
  • the transform coefficients and inverse quantized coefficients belong to the frequency domain data.
  • the embodiment of the invention adopts the frequency domain distortion estimation method. Compared with the existing spatial domain distortion calculation method, the discrete cosine transform and the inverse discrete cosine transform are not required for each prediction mode, and the distortion can be calculated, and the optimal prediction mode is further selected. Low complexity and low power consumption.
  • the current prediction mode cost value is obtained according to the code rate value and the distortion value of the current prediction mode, including: calculating the current prediction mode cost value by the rate distortion optimization function according to the code rate value and the distortion value of the current prediction mode.
  • the rate-distortion optimization function is used to weigh the code rate value and the distortion value.
  • the discrete residual cosine transform is performed on the prediction residual of the current prediction mode to obtain transform coefficients, including: performing discrete cosine transform on the prediction residual of the current prediction mode by using a discrete cosine transform matrix to obtain transform coefficients and discretizing The cosine transform matrix is orthogonally reversible.
  • the method provided by the embodiment of the present invention can be applied to the HEVC/H.265 standard.
  • the discrete residual cosine transform is performed on the prediction residual of the current prediction mode to obtain transform coefficients, including: performing integer discrete cosine transform on the prediction residual of the current prediction mode by using an integer discrete cosine transform matrix to obtain an integer discrete Cosine transform coefficient, integer discrete cosine transform matrix non-orthogonal reversible; point multiplication of integer discrete cosine transform coefficients to obtain transform coefficients.
  • the method provided by the embodiment of the present invention can be applied to the H.264 standard.
  • an embodiment of the present invention provides a video coding distortion estimating apparatus, where the apparatus includes: a discrete cosine transform unit, configured to perform discrete cosine transform on a prediction residual of a current prediction mode to obtain a transform coefficient, where the prediction residual The difference is the difference between the pixel values of the original image and the predicted image, and the predicted image is predicted according to an inter prediction mode or an intra prediction mode, and the inter prediction mode is to predict the next frame image by using the temporally adjacent previous frame image, The intra prediction mode is to predict the current frame image using the current intra spatial correlation.
  • a quantization unit for quantizing the transform coefficients to obtain quantized coefficients.
  • An inverse quantization unit is configured to inverse quantize the quantized coefficients to obtain inverse quantized coefficients.
  • a distortion value calculation unit configured to obtain a distortion value of the current prediction mode according to a difference between the transform coefficient and the inverse quantization coefficient.
  • the optimal prediction mode selecting unit is configured to select a prediction mode with the lowest generation value among the plurality of prediction modes as the optimal prediction mode.
  • the residual unit is reconstructed for inverse discrete cosine transform on the inverse quantized coefficients of the optimal prediction mode to obtain a reconstructed residual.
  • the prediction mode includes multiple, and the prediction images of each prediction mode are different, and the quantization coefficients of the current prediction mode are entropy encoded to obtain entropy coding information, and the code rate value is determined according to the entropy coding information, where the current prediction mode is multiple. Any of a variety of prediction modes.
  • the value calculation unit is specifically configured to calculate a current prediction mode cost value by using a rate distortion optimization function according to a code rate value and a distortion value of the current prediction mode, and the rate distortion optimization function is used to weigh the code rate value and Distortion value.
  • the discrete cosine transform unit is specifically configured to: perform discrete cosine transform on the prediction residual of the current prediction mode by using a discrete cosine transform matrix to obtain transform coefficients, and the discrete cosine transform matrix is orthogonally reversible.
  • the discrete cosine transform unit is specifically configured to perform integer discrete cosine transform on the prediction residual of the current prediction mode by using an integer discrete cosine transform matrix to obtain an integer discrete cosine transform coefficient, and the integer discrete cosine transform matrix is not positive.
  • an embodiment of the present invention provides a video coding distortion estimating apparatus, where the apparatus includes:
  • a memory for storing program instructions.
  • a processor configured to perform a discrete cosine transform on the prediction residual of the current prediction mode according to the program instruction stored in the memory, to obtain a transform coefficient, where the prediction residual is a difference between a pixel value of the original image and the predicted image,
  • the predicted image is predicted according to an inter prediction mode or an intra prediction mode, and the inter prediction mode is to predict the next frame image by using the temporally adjacent previous frame image, and the intra prediction mode is to use the current intra spatial correlation prediction.
  • Current frame image The transform coefficients are quantized to obtain quantized coefficients.
  • the quantized coefficients are inverse quantized to obtain inverse quantized coefficients.
  • the distortion value of the current prediction mode is obtained from the difference between the transform coefficient and the inverse quantization coefficient.
  • the current prediction mode cost value is obtained according to the code rate value and the distortion value of the current prediction mode. Choose The prediction mode with the lowest generation value among the multiple prediction modes is the optimal prediction mode. The inverse discrete cosine transform is performed on the inverse quantized coefficients of the optimal prediction mode to obtain a reconstructed residual.
  • the processor performs the current prediction mode cost value according to the code rate value and the distortion value of the current prediction mode, including: calculating the current prediction by using the rate distortion optimization function according to the code rate value and the distortion value of the current prediction mode.
  • the mode generation value, rate distortion optimization function is used to weigh the code rate value and the distortion value.
  • the processor performs discrete cosine transform on the prediction residual of the current prediction mode to obtain transform coefficients, including: performing discrete cosine transform on the prediction residual of the current prediction mode by using a discrete cosine transform matrix to obtain a transform
  • the coefficient, discrete cosine transform matrix is orthogonally reversible.
  • the processor performs a discrete cosine transform on the prediction residual of the current prediction mode to obtain transform coefficients, including: performing an integer discrete cosine transform on the prediction residual of the current prediction mode by using an integer discrete cosine transform matrix, An integer discrete cosine transform coefficient is obtained, and the integer discrete cosine transform matrix is non-orthogonally reversible.
  • the integer discrete cosine transform coefficients are multiplied to obtain transform coefficients.
  • an embodiment of the present invention provides a computer storage medium for storing a program, where the step of executing the program includes the steps of the foregoing first aspect.
  • the embodiment of the invention provides a video encoding method and device, which uses a frequency domain distortion calculation method to obtain a distortion value, and selects a most predictive mode according to a rate distortion optimization function, and performs an inverse transformation on the inverse quantization coefficient of the optimal prediction mode. Rebuild the residual.
  • the prediction residual of each prediction mode needs to be transformed, quantized, inverse quantized, and inverse transformed to calculate the distortion value of each prediction mode and obtain the reconstruction residual.
  • the embodiment of the present invention has low complexity and low power. The advantage is high and the reliability is high.
  • FIG. 2 is a schematic flowchart of a video encoding method according to an embodiment of the present invention
  • FIG. 3 is a schematic flowchart of still another video encoding method according to an embodiment of the present disclosure
  • FIG. 5 is a flowchart of implementing a video encoding method according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of comparison between distortion calculated by the distortion estimation method of the present invention and distortion calculated by the prior art according to an embodiment of the present invention
  • FIG. 7 is a schematic diagram showing a comparison between the distortion calculated by the distortion estimation method of the present invention and the distortion calculated by the prior art according to an embodiment of the present invention
  • FIG. 8 is a schematic structural diagram of a video coding distortion estimation apparatus according to an embodiment of the present invention.
  • FIG. 9 is a schematic structural diagram of still another video coding distortion estimation apparatus according to an embodiment of the present invention.
  • the transform operation is usually a Discrete Cosine Transformation (DCT).
  • DCT Discrete Cosine Transformation
  • transform coefficients the quantized coefficients, and the inverse quantized coefficients involved in the present application are respectively: transform coefficients, quantized coefficients, and dequantized coefficients. The description will not be repeated below.
  • FIG. 2 is a schematic flowchart of a video encoding method according to an embodiment of the present invention. Referring to FIG. 2, the method includes:
  • Step 201 performing DCT transform on the prediction residual of the current prediction mode to obtain a transform coefficient, wherein the prediction residual is a difference between pixel values of the original image and the predicted image, and the predicted image is spatial correlation and time correlation according to the original image.
  • the sex is predicted according to the current prediction mode.
  • the predicted image is predicted according to an inter prediction mode or an intra prediction mode
  • the inter prediction is The measurement mode is to predict the next frame image by using the temporally adjacent previous frame image
  • the intra prediction mode is to predict the current frame image by using the current intraframe spatial correlation.
  • the prediction residual of the current prediction mode is DCT-transformed by the DCT transform matrix to obtain transform coefficients, and the DCT transform matrix is orthogonally reversible.
  • the prediction residual of the current prediction mode is integer DCT transformed by the integer DCT transform matrix to obtain integer DCT transform coefficients, which are non-orthogonally reversible; and the integer DCT coefficients are dot-multiplied to obtain transform coefficients.
  • the prediction residual is airspace data.
  • the DCT transform converts spatial domain data into frequency domain data. Quantize the data for compression. Video coding distortion is introduced by quantization operations, which are lossless transforms.
  • the video coding method provided by the embodiment of the present invention is applied to the HEVC standard and the H.264 standard by taking a transform size of 4 ⁇ 4 as an example.
  • the DCT transform includes the following steps:
  • Step 201a in order to implement the DCT transform, first improve the precision of the integer DCT transform matrix A, multiply each matrix element in A by 128 (2 7 ), and approximate each matrix element to obtain a DCT transform matrix.
  • C
  • step 201b the prediction residual is X
  • the prediction residual X is DC-transformed by the DCT transform matrix C to obtain a transform coefficient Y, as shown in formula (2):
  • the formula (2) indicates that the prediction residual X is operated by the DCT transformation matrix C, and is shifted to the right by 9 bits, that is, divided by 2 9 .
  • X is spatial domain data and Y is frequency domain data.
  • step 201c since the DCT transform matrix C is an orthogonal invertible matrix, the inverse transform of Y is lossless, as in equation (3):
  • the DCT transform is a lossless transform, and the entire encoding process does not introduce errors due to the transform, and at the same time, since the encoding process is: prediction, transform, quantization And entropy coding. Therefore, in the HEVC standard, only the quantization introduces an error, so HEVC can estimate the distortion D by using the transform coefficient before quantization and the inverse quantization coefficient after inverse quantization.
  • the DCT transform includes integer DCT transform and point multiplication.
  • the integer DCT transform can be named as a kernel transform, and its integer DCT transform matrix C f is:
  • the integer DCT transform matrix Cf is non-orthogonally reversible.
  • the transform module when the coding method is actually applied to the hardware module, the transform module includes a core transform, and the quantization module includes point multiplication and quantization.
  • the integer DCT transform matrix C f if the kernel transform coefficient is directly used as the transform coefficient in the H.264 standard, the lossless transform described in equations (2) and (3) cannot be satisfied. . Therefore, the point multiplication needs to be attributed to the transform module, and the kernel transform and point multiplication are completed into a complete DCT transform. Therefore, the H.264 standard can also estimate the distortion D by using the transform coefficient before quantization and the inverse quantization coefficient after inverse quantization.
  • the point multiplication coefficient after the point multiplication of the kernel transform coefficients is the transform coefficient described in step 201.
  • the video coding method provided by the embodiment of the present invention is applicable to the HEVC standard and the H.264 standard only by using a transform size of 4 ⁇ 4.
  • the present invention is implemented. The same applies.
  • Step 202 Quantize the transform coefficients to obtain quantized coefficients.
  • the transform coefficients belong to frequency domain data
  • the quantized quantized quantized coefficients also belong to frequency domain data.
  • Step 203 Perform inverse quantization on the quantized coefficients to obtain inverse quantized coefficients.
  • inverse quantization is to understand compression.
  • the distortion is further estimated by using the transform coefficients before quantization and the inverse quantized coefficients after inverse quantization.
  • Step 204 Estimate a distortion value of the current prediction mode according to a difference between the transform coefficient and the inverse quantization coefficient.
  • the quantization coefficient is inverse quantized to obtain a corresponding inverse quantization coefficient Q', and the video coding distortion D' is directly estimated by using the inverse quantization coefficient Q' and the error of the transform coefficient T before quantization.
  • the specific calculation method is as shown in formula (4):
  • blocksize represents the size of the current block
  • (i, j) represents the coordinate value of the current block.
  • the inverse DCT transform converts the frequency domain data into spatial domain data.
  • the prior art calculates coding distortion using two spatial domain data of prediction residual and reconstruction residual.
  • Embodiments of the present invention estimate distortion using two frequency domain data of transform coefficients and inverse quantization coefficients. Therefore, the distortion calculation method provided in the embodiment of the present invention is also referred to as a frequency domain distortion estimation method.
  • Step 205 Obtain a current prediction mode cost value according to a code rate value and a distortion value of the current prediction mode.
  • the current prediction mode cost value is calculated by the rate distortion optimization function as described in the formula (1) according to the code rate value and the distortion value of the current prediction mode, and the rate distortion optimization function is used to weigh the code rate value and the distortion value.
  • Step 206 Select a prediction mode with the lowest generation value among the plurality of prediction modes as the optimal prediction mode.
  • Step 207 Perform inverse discrete cosine transform on the inverse quantized coefficients of the optimal prediction mode to obtain a reconstructed residual.
  • the frequency domain distortion calculation method is adopted.
  • the present invention can estimate the distortion of each prediction mode without performing an inverse transform operation on each prediction mode.
  • the present invention performs an inverse discrete cosine transform only on the optimal prediction mode, and obtains a reconstruction residual and saves it.
  • the sum of the predicted image and the reconstruction residual in the optimal prediction mode is saved as a reconstructed video.
  • the invention greatly reduces video coding complexity and power consumption.
  • FIG. 3 is a schematic flowchart of still another video encoding method according to an embodiment of the present invention. Referring to FIG. 3, the method includes:
  • the distortion value D of each prediction mode is obtained by frequency domain error.
  • the quantized coefficients of the current prediction mode are entropy encoded to obtain entropy coding information, and the code rate value R is determined according to the entropy coding information, wherein the current prediction mode is any one of multiple prediction modes.
  • the current prediction mode cost value is obtained according to the code rate value R and the distortion value D of the current prediction mode.
  • the cost value J of each prediction mode is calculated by a rate distortion optimization function.
  • the rate-distortion optimization function is used to weigh the code rate value R and the distortion value D.
  • the prediction mode includes multiple types, and the prediction images of each prediction mode are different, and therefore, the prediction residual X of each prediction mode is different. Therefore, the code rate value R and the distortion value D of each prediction mode are different, and the cost value J of each prediction mode is different.
  • the prediction mode with the smallest value J in the plurality of prediction modes is selected as the optimal prediction mode; and the inverse quantization coefficient of the optimal prediction mode is inverse DCT transformed to obtain the reconstruction residual.
  • the reconstruction residual is added to the predicted value of the optimal prediction mode to form a reconstructed video.
  • the mode information and the quantized coefficients of the optimal prediction mode are entropy encoded and output.
  • the inverse prediction operation is performed on the optimal prediction mode to obtain a reconstruction residual and a reconstructed video.
  • the frequency domain distortion estimation method is adopted, and the distortion can be obtained by performing DCT transform and inverse DCT transform on each prediction mode.
  • only an inverse DCT transform is performed on the optimal prediction mode to obtain a reconstruction residual.
  • FIG. 4 and FIG. 5 are flowcharts showing implementations of a video encoding method provided by the prior art and the embodiments of the present invention, respectively. 4 and 5 are exemplified by the case where only five prediction modes are implemented in parallel, respectively, to explain the difference between the present invention and the prior art.
  • the prior art needs to perform transform, quantization, inverse quantization, and inverse transform operations on the prediction residuals of each prediction mode to obtain reconstruction residuals for each prediction mode. Transforming and quantizing the residual between the predicted value obtained from each prediction mode and the original value, and entropy encoding the obtained quantized coefficient to obtain the coded bit number (code rate) R of the current block, and simultaneously performing the quantized coefficient.
  • the inverse quantization inverse transform is performed to obtain a decoding (reconstruction) residual, and the distortion is obtained by using the inverse transform to reconstruct the residual and the prediction residual before the transform to obtain the distortion D.
  • the specific calculation method is as shown in formula (5):
  • P org (i, j) represents the pixel value at the corresponding position (i, j) of the original block
  • p rec (i, j) represents the pixel value at the corresponding position (i, j) of the reconstructed block
  • p pred (i , j) represents the pixel value at the corresponding position (i, j) of the prediction block.
  • the distortion value D of each prediction mode is calculated from the prediction residual and the reconstruction residual. As shown by D 11 - D 15 in Fig. 4. And use R and D to calculate the rate distortion cost in the current mode according to formula (1).
  • the invention mainly relates to the rate distortion optimization technique in the prediction process, and the selection of the optimal prediction mode is obtained. Get the best predictive value. In particular, it relates to the acquisition of distortion D in rate-distortion optimization techniques.
  • the acquisition rate of the code rate R is consistent with the scheme shown in the prior art of FIG.
  • the quantized coefficients are inverse quantized to obtain a corresponding inverse quantized coefficient Q', and the error D of the inverse quantized coefficient Q' and the pre-quantized transform coefficient T is directly used, as shown in Figure 21 , D 21 - D 25 is shown.
  • the specific calculation method is shown in formula (3).
  • the embodiment of the present invention replaces the original distortion calculation method using the spatial domain by designing a distortion estimation algorithm in the frequency domain, so that the coding end does not need to perform inverse transformation operations on all prediction modes, and only needs to perform a complete coding strategy for the selected optimal mode. .
  • the video coding method provided by the embodiment of the invention has the advantages of low complexity and low power consumption.
  • the following is a schematic diagram of the comparison between the distortion calculated by the distortion estimation method of the present invention and the distortion calculated by the prior art provided by FIG. 6 and FIG. 7 to illustrate the reliability of the frequency domain distortion estimation method provided by the embodiment of the present invention.
  • the embodiment of the present invention compares D obtained by using the prior art and the D obtained by the frequency domain distortion estimation method of the present invention after block processing of one frame of video. details as follows:
  • FIG. 6 is a schematic diagram of comparison between distortion calculated by the distortion estimation method of the present invention and distortion calculated by the prior art according to an embodiment of the present invention.
  • the figure is in the H.264 standard. Since the image resolution is 1920 ⁇ 1080, the processing is performed in blocks 4 ⁇ 4, and the distortion values of some of the video blocks are selected for comparison to illustrate the embodiment of the present invention.
  • the abscissa of each point represents the D value of the current block obtained by the existing spatial domain distortion calculation method in the H.264 standard, and the ordinate represents the frequency domain distortion estimation method provided by the embodiment of the present invention.
  • the resulting D' value of the current block is the D value of the current block obtained by the existing spatial domain distortion calculation method in the H.264 standard, and the ordinate represents the frequency domain distortion estimation method provided by the embodiment of the present invention. The resulting D' value of the current block.
  • FIG. 7 is a schematic diagram showing a comparison between the distortion calculated by the distortion estimation method of the present invention and the distortion calculated by the prior art according to an embodiment of the present invention.
  • the abscissa of each point represents the D value of the current block obtained by the existing spatial domain distortion calculation method in the HEVC standard, and the ordinate represents the frequency domain distortion estimation method provided by the embodiment of the present invention.
  • the D' value of the current block represents the D value of the current block obtained by the existing spatial domain distortion calculation method in the HEVC standard, and the ordinate represents the frequency domain distortion estimation method provided by the embodiment of the present invention.
  • Distortion calculation method The distortion values D obtained by the two distortion calculation methods are basically similar. It can be seen that the frequency domain distortion estimation method provided by the embodiment of the present invention is very close to the distortion value calculated by the prior art.
  • FIG. 6 and FIG. 7 a distortion value calculation comparison of a partial video block is illustrated.
  • the distortion estimation method and the existing method provided by the embodiment of the present invention are used.
  • the technique calculates the distortion of any video block and satisfies the rules shown in FIG. 6 and FIG. 7. I will not repeat them here. Therefore, the frequency domain distortion estimation method provided by the embodiment of the present invention has high reliability.
  • the distortion value obtained by the frequency domain distortion estimation method provided by the embodiment of the present invention is closer to the calculation result of the HEVC standard than the calculation result of the H.264 standard. Since the HEVC standard is more complicated than H.264, it is further verified that the distortion estimation method provided by the embodiment of the present invention can be well applied to the HEVC standard with more transform sizes and prediction modes.
  • the embodiment of the present invention replaces the original distortion calculation method using the spatial domain by designing a distortion estimation algorithm in the frequency domain, so that the coding end does not need to perform inverse transformation operations on all prediction modes, and only needs to perform a complete coding strategy for the selected optimal mode.
  • the distortion value estimated by the embodiment of the present invention is very close to the distortion value obtained by the existing spatial domain distortion calculation method. Therefore, the embodiment of the invention has the advantages of low complexity, low power consumption, and high reliability.
  • FIG. 8 is a schematic structural diagram of a video encoding apparatus according to an embodiment of the present disclosure. and referring to FIG. 8, the apparatus includes:
  • the discrete cosine transform unit 801 is configured to perform discrete cosine transform on the prediction residual of the current prediction mode to obtain transform coefficients, where the prediction residual is a residual of the pixel value of the original image and the predicted image, and the predicted image is a space according to the original image. Correlation and time correlation are predicted according to current prediction mode get.
  • the predicted image is obtained according to an inter prediction mode or an intra prediction mode, wherein the inter prediction mode is to predict a next frame image by using a temporally adjacent previous frame image, where the intra prediction mode is utilized.
  • the current intra spatial correlation predicts the current frame image.
  • the discrete cosine transform unit 801 is specifically configured to perform discrete cosine transform on the prediction residual of the current prediction mode by using a discrete cosine transform matrix under the HEVC standard to obtain transform coefficients.
  • the discrete cosine transform matrix is orthogonally reversible.
  • the integer residual cosine transform is performed on the prediction residual of the current prediction mode by an integer discrete cosine transform matrix to obtain an integer discrete cosine transform coefficient, and the integer discrete cosine transform matrix is non-orthogonally reversible.
  • the integer discrete cosine transform coefficients are multiplied to obtain transform coefficients.
  • the quantization unit 802 is configured to quantize the transform coefficients to obtain quantized coefficients.
  • the inverse quantization unit 803 is configured to inverse quantize the quantized coefficients to obtain inverse quantized coefficients.
  • the distortion value calculation unit 804 is configured to estimate the distortion value of the current prediction mode according to the difference between the transform coefficient and the inverse quantization coefficient.
  • the cost representative unit 805 is configured to obtain the current prediction mode cost value according to the code rate value and the distortion value of the current prediction mode.
  • the cost value calculation unit 805 is specifically configured to: perform entropy coding on the quantized coefficients of the current prediction mode, obtain entropy coding information, and determine a code rate value according to the entropy coding information, where the current prediction mode is any one of multiple prediction modes.
  • the current prediction mode cost value is calculated by the rate distortion optimization function according to the code rate value and the distortion value of the current prediction mode, and the rate distortion optimization function is used to weigh the code rate value and the distortion value.
  • the optimal prediction mode selecting unit 806 is configured to select a prediction mode with the lowest generation value among the plurality of prediction modes as the optimal prediction mode.
  • the reconstruction residual unit 807 is configured to perform inverse discrete cosine transform on the inverse quantized coefficients of the optimal prediction mode to obtain a reconstruction residual.
  • the 801-807 units and the like are for implementing the functions of the above method embodiments, and include corresponding hardware structures and/or software modules for performing the respective functions.
  • the present invention can be implemented in a combination of hardware or hardware and computer software in combination with the elements and algorithm steps of the various examples described in the embodiments disclosed herein. Whether a function is implemented in hardware or computer software to drive hardware depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.
  • the function modules may be divided into 801-807 and the like according to the foregoing method embodiments.
  • each function module may be divided according to each function, or two or more functions may be integrated into one processing module.
  • the above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of the module in the embodiment of the present invention is schematic, and is only a logical function division, and the actual implementation may have another division manner.
  • FIG. 9 is a schematic structural diagram of another video encoding apparatus according to an embodiment of the present invention.
  • the network includes a network card 901, a memory 902, a processor 903, and a bus 904.
  • the network card 901 is configured with multiple communication interfaces, and the terminal collects or receives video through the communication interface, and performs video encoding and decoding.
  • Memory 902 is used to store program instructions.
  • Network card 901, memory 902, and processor 903 communicate over bus 904.
  • the processor 903 is configured to perform a discrete cosine transform on the prediction residual of the current prediction mode according to the program instructions stored in the memory 902 to obtain transform coefficients.
  • the prediction residual is the difference between the pixel values of the original image and the predicted image, and the predicted image is predicted according to an inter prediction mode or an intra prediction mode, and the inter prediction mode is predicted by using the temporally adjacent previous frame image.
  • the intra prediction mode is to predict the current frame image using the current intraframe spatial correlation.
  • the transform coefficients are quantized to obtain quantized coefficients.
  • the quantized coefficients are inverse quantized to obtain inverse quantized coefficients. Obtaining the loss of the current prediction mode according to the difference between the transform coefficient and the inverse quantization coefficient True value.
  • the current prediction mode cost value is obtained according to the code rate value and the distortion value of the current prediction mode.
  • the prediction mode with the lowest generation value among the multiple prediction modes is selected as the optimal prediction mode.
  • the inverse discrete cosine transform is performed on the inverse quantized coefficients of the optimal prediction mode to obtain a reconstructed residual.
  • the memory 902 may be a storage device or a collective name of a plurality of storage elements, and is used to store information such as programs and data required to run the conference server.
  • the memory 902 may include a random access memory (RAM), a flash memory, a read only memory (ROM), an erasable programmable read only memory (EPROM), and an electric memory.
  • RAM random access memory
  • ROM read only memory
  • EPROM erasable programmable read only memory
  • EEPROM erasable programmable read only memory
  • register a hard disk, a removable hard disk, a compact disk (CD-ROM), a flash memory, or any other form of storage medium known in the art. Or a combination of multiple storage media.
  • the processor 903 can be a CPU, a general-purpose processor, a DSP, an Application-Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, and a transistor logic. Device, hardware component, or any combination thereof. It is possible to implement or carry out the various illustrative logical blocks, units and circuits described in connection with the present disclosure.
  • the processor may also be a combination of computing functions, for example, including one or more microprocessor combinations, a combination of a DSP and a microprocessor, and the like.
  • the bus 904 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus.
  • ISA Industry Standard Architecture
  • PCI Peripheral Component
  • EISA Extended Industry Standard Architecture
  • the bus 904 can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in Figure 9, but it does not mean that there is only one bus or one type of bus.
  • the processor 903 calculates a current prediction mode cost value by a rate distortion optimization function for weighing the code rate value and the distortion value according to the code rate value and the distortion value of the current prediction mode.
  • the processor 903 performs the discrete cosine transform on the prediction residual of the current prediction mode to obtain transform coefficients, including:
  • the discrete residual cosine transform is performed on the prediction residual of the current prediction mode by the discrete cosine transform matrix to obtain transform coefficients.
  • the discrete cosine transform matrix is orthogonally reversible.
  • the integer discrete cosine transform matrix is non-orthogonally reversible.
  • the integer discrete cosine transform coefficients are multiplied to obtain transform coefficients.
  • bus 904 can be used to connect the various units in FIG.
  • the processor 903 can be used to perform the functions of the units 801-807, and the processor 902 can be used to store the data of each unit of 801-807.
  • the video coding method and apparatus provided by the embodiments of the present invention adopt the frequency domain distortion estimation method, which has the advantages of low complexity and low power consumption compared with the existing spatial domain distortion calculation method, and has high reliability.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

本发明实施例涉及视频编码方法及装置。该方法包括：对当前预测模式的预测残差进行离散余弦变换，得到变换系数；对变换系数进行量化，得到量化系数；对量化系数进行反量化，得到反量化系数；根据变换系数和反量化系数之差得到所述当前预测模式的失真值。根据当前预测模式的码率值和失真值得到当前预测模式代价值；选取多种预测模式中代价值最小的预测模式为最优预测模式；对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。本发明实施例根据变换系数和反量化系数之差得到失真值，只需对最优预测模式进行一次反变换得到重建残差。本发明实施例具有低复杂度、低功耗的优势，且可靠性高。

Description

视频编码方法及装置 技术领域
本发明涉及图像处理技术领域，尤其涉及一种视频编码方法及装置。
背景技术
数字视频压缩格式H.264标准和高效率视频编码(High Efficiency Video Coding，简称HEVC)标准是目前使用的两种视频编码方案，其中，HEVC标准又可称为H.265标准。
图1为现有技术视频编码框架图。在H.264和HEVC标准中，采用的整体编码框架都是预测、变换、量化、熵编码。预测是为了压缩图像。其中，预测部分是通过去除视频内容的空间相关性和时间相关性为当前编码图像获得预测值，以及得到原始图像与预测图像的残差。对残差进行变换量化。对预测残差的量化系数(quantized coefficients)以及预测模式等信息进行熵编码形成码流输出。
其中，预测模式可分为帧间预测和帧内预测，帧间预测为利用时间上相邻的前一帧预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。帧间预测模式可利用视频相邻帧信息预测，压缩率更高。帧内预测模式的压缩率较帧间预测模式低，但可去除当前帧内部的相邻块的空间冗余度。
预测的准确性对编码性能起到了决定性的作用。在编码过程中，每种预测类型下都有多种预测模式，每种预测模式具有不同的预测值和预测残差，每种预测模式熵编码后的码率(rate，简称R)和重建视频后的失真(distortion，简称D)不同。通过率失真优化(Rate Distortion Optimization，简称RDO)计算每种模式的代价，选取代价函数最小的模式为最优模式。RDO利用R和 D两个量设计代价函数计算每种预测模式的代价值J，如公式(1)所示：
J＝D+λR               (1)
现有技术输出最优预测模式的编码数据，并保存最优预测模式的重建残差和重建视频。现有技术对预测残差进行变换、量化、反量化和反变换操作得到重建残差，根据预测残差和重建残差之差得到失真D。其中，预测残差为空域数据，变换操作将空域数据转换为频域数据。故现有技术又称为空域失真计算方法。
现有技术需要对每种预测模式的预测残差进行变换、量化、反量化及反变换，才能选出最优预测模式。现有技术复杂度高，其中，变换和反变换的复杂度随着变换大小成指数增长，HEVC标准的变换大小相比H.264增大，同时，HEVC标准中的预测模式的数目也相比于H.264标准成倍增长，因此，现有技术会导致HEVC标准的复杂度剧增。
现有技术在硬件实现中，变换和反变换是功耗较高的模块，现有技术对每种预测模式的预测残差都进行变换和反变换，功耗过高。
发明内容
本发明实施例涉及一种视频编码方法及装置，解决现有技术复杂度高，硬件实现功耗大的问题。
第一方面，本发明实施例提供了一种视频编码方法，该方法包括：对当前预测模式的预测残差进行离散余弦变换，得到变换系数，其中，预测图像为根据帧间预测模式或帧内预测模式预测得到，帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。对变换系数进行量化，得到量化系数。对量化系数进行反量化，得到反量化系数。根据变换系数和反量化系数之差得到当前预测模式的失真值。根据当前预测模式的码率值和失真值得到当前预测模式代价值。选取多种预测模式中代价值最小的预测模式为最优预测模式。对最优预测模 式的反量化系数进行反离散余弦变换，得到重建残差。
具体地，离散余弦变换将空域数据转换成频域数据。变换系数和反量化系数属于频域数据。本发明实施例采用频域失真估算方法，相比现有空域失真计算方法，无需对每种预测模式进行离散余弦变换和反离散余弦变换，即可计算失真，进一步选取最优预测模式。复杂度低、功耗低。
在一种可能的设计中，根据当前预测模式的码率值和失真值得到当前预测模式代价值，包括：根据当前预测模式的码率值和失真值通过率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
在一种可能的设计中，对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：通过离散余弦变换矩阵对当前预测模式的预测残差进行离散余弦变换，得到变换系数，离散余弦变换矩阵正交可逆。
具体地，本发明实施例提供的方法可以应用到HEVC/H.265标准。
在一种可能的设计中，对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：通过整数离散余弦变换矩阵对当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，整数离散余弦变换矩阵非正交可逆；对整数离散余弦变换系数进行点乘，得到变换系数。
具体地，本发明实施例提供的方法可以应用到H.264标准。
第二方面，本发明实施例提供了一种视频编码失真估算装置，该装置包括：离散余弦变换单元，用于对当前预测模式的预测残差进行离散余弦变换，得到变换系数，其中，预测残差为原始图像与预测图像的像素值之差，预测图像为根据帧间预测模式或帧内预测模式预测得到，帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。量化单元，用于对变换系数进行量化，得到量化系数。反量化单元，用于对量化系数进行反量化，得到反量化系数。失真值计算单元，用于根据变换系数和反量化系数之差得到当前预测模式的失真值。代价值计算单元，用于根据当前预测模式的码率值和失真值得到当前预测模 式代价值。最优预测模式选取单元，用于选取多种预测模式中代价值最小的预测模式为最优预测模式。重建残差单元，用于对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
具体地，预测模式包括多种，每种预测模式的预测图像不同，对当前预测模式的量化系数进行熵编码，得到熵编码信息，根据熵编码信息确定码率值，其中，当前预测模式为多种预测模式中的任一种。
在一种可能的设计中，代价值计算单元具体用于：根据当前预测模式的码率值和失真值通过率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
在一种可能的设计中，离散余弦变换单元具体用于：通过离散余弦变换矩阵对当前预测模式的预测残差进行离散余弦变换，得到变换系数，离散余弦变换矩阵正交可逆。
在一种可能的设计中，离散余弦变换单元具体用于：通过整数离散余弦变换矩阵对当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，整数离散余弦变换矩阵非正交可逆；对整数离散余弦变换系数进行点乘，得到变换系数。
第三方面，本发明实施例提供了一种视频编码失真估算装置，该装置包括：
存储器，用于存储程序指令。处理器，用于根据存储器中存储的程序指令执行以下操作：对当前预测模式的预测残差进行离散余弦变换，得到变换系数；其中，预测残差为原始图像与预测图像的像素值之差，预测图像为根据帧间预测模式或帧内预测模式预测得到，帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。对变换系数进行量化，得到量化系数。对量化系数进行反量化，得到反量化系数。根据变换系数和反量化系数之差得到当前预测模式的失真值。根据当前预测模式的码率值和失真值得到当前预测模式代价值。选 取多种预测模式中代价值最小的预测模式为最优预测模式。对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
在一种可能的设计中，处理器执行根据当前预测模式的码率值和失真值得到当前预测模式代价值，包括：根据当前预测模式的码率值和失真值通过率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
在一种可能的设计中，处理器执行对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：通过离散余弦变换矩阵对当前预测模式的预测残差进行离散余弦变换，得到变换系数，离散余弦变换矩阵正交可逆。
在一种可能的设计中，处理器执行对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：通过整数离散余弦变换矩阵对当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，整数离散余弦变换矩阵非正交可逆。对整数离散余弦变换系数进行点乘，得到变换系数。
第四方面，本发明实施例提供了一种计算机存储介质，用于存储有程序，程序执行的步骤包括上述第一方面的步骤。
本发明实施例提供一种视频编码方法及装置，利用频域失真计算法得到失真值，并根据率失真优化函数选出最有预测模式，对最优预测模式的反量化系数进行一次反变换得到重建残差。相比现有技术，需要每种预测模式的预测残差进行变换、量化、反量化和反变换计算每种预测模式的失真值以及得到重建残差，本发明实施例具有低复杂度、低功耗的优势，且可靠性高。
附图说明
图1为现有技术视频编码框架图；
图2为本发明实施例提供的一种视频编码方法流程示意图；
图3为本发明实施例提供的又一种视频编码方法流程示意图；
图4为现有技术视频编码方法实现流程图；
图5为本发明实施例提供的视频编码方法实现流程图；
图6为本发明实施例提供的一种采用本发明的失真估算方法计算得到的失真与现有技术计算得到的失真的对比示意图；
图7为本发明实施例提供的又一种采用本发明的失真估算方法计算得到的失真与现有技术计算得到的失真的对比示意图；
图8为本发明实施例提供的一种视频编码失真估算装置结构示意图；
图9为本发明实施例提供的又一种视频编码失真估算装置结构示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
需要说明的是，通常，在视频编码中，变换操作通常为离散余弦变换(Discrete Cosine Transformation，简称DCT)。
可以理解的是，本申请中涉及的变换系数、量化系数、反量化系数其英文分别为：transform coefficients、quantized coefficients、dequantized coefficients。以下不再重复说明。
图2为本发明实施例提供的一种视频编码方法流程示意图，参照图2，该方法包括：
步骤201，对当前预测模式的预测残差进行DCT变换，得到变换系数，其中，该预测残差为原始图像与预测图像的像素值之差，预测图像为根据原始图像的空间相关性和时间相关性按照当前预测模式预测得到。
其中，预测图像为根据帧间预测模式或帧内预测模式预测得到，帧间预 测模式为利用时间上相邻的前一帧图像预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。
具体地，通过DCT变换矩阵对当前预测模式的预测残差进行DCT变换，得到变换系数，该DCT变换矩阵正交可逆。
具体地，通过整数DCT变换矩阵对当前预测模式的预测残差进行整数DCT变换，得到整数DCT变换系数，该整数DCT变换矩阵非正交可逆；对整数DCT系数进行点乘，得到变换系数。
需要说明的是，预测残差为空域数据。DCT变换将空域数据转换为频域数据。量化对数据进行压缩。视频编码失真由量化操作引入，DCT变换属于无损变换。以下以变换大小为4×4为例，说明本发明实施例提供的视频编码方法适用于HEVC标准和H.264标准。
第一方面，对于HEVC标准，DCT变换包括以下步骤：
步骤201a，为实现DCT变换，先对整数DCT变换矩阵A进行精度提升，对A中的每个矩阵元素同乘以128(27)，并对每个矩阵元素近似取整，得到DCT变换矩阵C：
Figure PCTCN2016090884-appb-000001
其中，
Figure PCTCN2016090884-appb-000002
步骤201b，设预测残差为X，通过DCT变换矩阵C对预测残差X进行DCT变换，得到变换系数Y，如公式(2)：
Y＝(CXCT)＞＞9             (2)
其中，公式(2)表明通过DCT变换矩阵C对预测残差X进行运算，并右移9位，即除以29。
具体地，X为空域数据，Y为频域数据。
步骤201c，由于DCT变换矩阵C为正交可逆矩阵，因此，对Y进行反变换是无损的，如公式(3)：
Y′＝(C-1(Y＜＜9)(C-1)T)＝X             (3)
可以理解的是，根据公式(2)和公式(3)，在HEVC标准中，DCT变换为无损变换，整个编码过程不会由于变换引入误差，同时，由于编码的流程为：预测、变换、量化和熵编码。故在HEVC标准中，只有量化引入了误差，因此HEVC可以利用量化前的变换系数和反量化后的反量化系数估算失真D。
第二方面，对于H.264标准，DCT变换包括整数DCT变换和点乘。
整数DCT变换又可命名为核变换，其整数DCT变换矩阵Cf为：
Figure PCTCN2016090884-appb-000003
可以理解的是，在H.264标准中，整数DCT变换矩阵Cf是非正交可逆的。
需要说明的是，本领域技术人员可知，H.264标准中，将编码方法实际应用到硬件模块时，变换模块包括核变换，量化模块包括点乘和量化。根据整数DCT变换矩阵Cf非正交可逆可知，若在H.264标准中，直接采用核变换后的核变换系数作为变换系数，无法满足公式(2)和公式(3)所述的无损变换。故需要将点乘归到变换模块，核变换和点乘为完整的DCT变换。故，H.264标准同样可以利用量化前的变换系数和反量化后的反量化系数估算失真D。
可以理解的是，这里对核变换系数进行点乘后的点乘系数为步骤201中所述的变换系数。
可以理解的是，本发明实施例中提及的核变换等同整数DCT变换，本发 明实施例中提及的点乘系数等同变换系数。其命名根据实际进行的运算而定，并不用于限定各运算步骤。
可以理解的是，本发明实施例中提及的变换均为DCT变换的简称。
需要说明的是，本发明实施例仅以变换大小4×4为例，说明本发明实施例提供的视频编码方法适用于HEVC标准和H.264标准，对于其他变换大小的视频编码，本发明实施例同样适用。
步骤202，对所述变换系数进行量化，得到量化系数。
具体地，变换系数属于频域数据，量化压缩后的量化系数还属于频域数据。
步骤203，对所述量化系数进行反量化，得到反量化系数。
具体地，反量化是为了解压缩。进一步利用量化前的变换系数和反量化后的反量化系数估算失真。
步骤204，根据所述变换系数和所述反量化系数之差估算所述当前预测模式的失真值。
具体地，对量化系数进行反量化操作，得到相应的反量化系数Q′，直接利用该反量化系数Q′与量化前的变换系数T的误差估算视频编码失真D′。具体计算方法如公式(4)所示：
Figure PCTCN2016090884-appb-000004
其中，blocksize表示当前块的大小，(i,j)表示当前块的坐标值。
需要说明的是，反DCT变换将频域数据转换成空域数据。现有技术利用预测残差和重建残差两个空域数据计算编码失真。本发明实施例利用变换系数和反量化系数两个频域数据估算失真。因此，本发明实施例中提供的失真计算方法又称为频域失真估算法。
步骤205，根据当前预测模式的码率值和失真值得到当前预测模式代价值。
根据当前预测模式的码率值和失真值通过如公式(1)所述的率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
步骤206，选取多种预测模式中代价值最小的预测模式为最优预测模式。
步骤207，对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
本发明实施例，采用频域失真计算法，相比于现有空域失真计算法，本发明无需对每种预测模式都进行反变换操作即可估算每种预测模式的失真。本发明只对最优预测模式进行一次反离散余弦变换，得到重建残差并保存。另外，将最优预测模式下的预测图像与重建残差之和保存为重建视频。本发明大大降低了视频编码复杂度和功耗。
图3为本发明实施例提供的又一种视频编码方法流程示意图，参照图3，该方法包括：
采用图2所示的频域失真估算方法，通过频域误差得到每种预测模式的失真值D。
具体地，对当前预测模式的量化系数进行熵编码，得到熵编码信息，根据熵编码信息确定码率值R，其中，当前预测模式为多种预测模式中的任一种。根据当前预测模式的码率值R和失真值D得到当前预测模式代价值。
具体地，通过率失真优化函数计算每种预测模式的代价值J。率失真优化函数用于权衡码率值R和失真值D。
需要说明的是，预测模式包括多种，每种预测模式的预测图像不同，因此，每种预测模式的预测残差X不同。因此，每种预测模式的码率值R和失真值D不同，则每种预测模式的代价值J不同。
具体地，选取多种预测模式中代价值J最小的预测模式为最优预测模式；对最优预测模式的反量化系数进行反DCT变换，得到重建残差。
具体地，重建残差与最优预测模式的预测值相加组成重建视频。
具体地，对最优预测模式的模式信息和量化系数熵编码并输出。
本发明实施例，仅需对最优预测模式进行反变换操作得到重建残差和重建视频。本发明实施例采用频域失真估算法，无需对每种预测模式进行DCT变换和反DCT变换即可得到失真，本发明实施例仅需对最优预测模式进行一次反DCT变换，得到重建残差，具有低功耗低复杂度的优势。
图4和图5分别为现有技术和本发明实施例提供的视频编码方法实现流程图。图4和图5分别仅以5种预测模式并行实现的情况为例，以说明本发明和现有技术的区别。
如图4所示，现有技术需对每种预测模式的预测残差进行变换、量化、反量化、反变换操作，得到每种预测模式的重建残差。对每种预测模式获得的预测值与原始值之间的残差进行变换、量化，对获得的量化系数进行熵编码以获得当前块的编码比特数(码率)R，同时，对量化系数进行反量化反变换，以获得解码(重建)残差，利用反变换后重建残差与变换前的预测残差求平方和得到失真D，具体计算方法如公式(5)所示：
Figure PCTCN2016090884-appb-000005
其中，Porg(i,j)表示原始块对应位置(i,j)处的像素值，prec(i,j)表示重建块对应位置(i,j)处的像素值，ppred(i,j)表示预测块对应位置(i,j)处的像素值。
根据预测残差和重建残差计算每种预测模式的失真值D。如图4中的D11—D15所示。并利用R和D按照公式(1)计算得到当前模式下的率失真代价。
本发明主要涉及到预测过程中的率失真优化技术选择最优预测模式获 得最优预测值。尤其涉及到率失真优化技术中的失真D的获取。
如图5所示，码率R的获取方式与图4现有技术所示方案一致。在本发明中，对量化系数进行反量化操作，得到相应的反量化系数Q′，直接利用该反量化系数Q′与量化前的变换系数T的误差估算D，如图5中的D21—D25所示。具体计算方法如公式(3)所示。
本发明实施例通过设计频域的失真估计算法来取代原来利用空域的失真计算方法，以使得编码端无需对所有预测模式进行反变换操作，仅仅需要对选择出的最优模式进行完整的编码策略。本发明实施例提供的视频编码方法具有低复杂度、低功耗的优势。
以下通过图6和图7提供的采用本发明的失真估算方法计算得到的失真与现有技术计算得到的失真的对比示意图，说明本发明实施例提供的频域失真估算方法的可靠性。本发明实施例在对一帧视频分块处理后通过采用现有技术精确计算D和采用本发明的频域失真估算方法所得D进行对比。具体如下：
图6为本发明实施例提供的一种采用本发明的失真估算方法计算得到的失真与现有技术计算得到的失真的对比示意图。参照图6，该图为H.264标准中，由于图像分辨率为1920×1080，按分块4×4进行处理，选取其中一些视频块的失真值进行对比，以说明本发明实施例。
如图6所示，每个点的横坐标表示采用H.264标准中的现有空域失真计算方法得到的当前块的D值，其纵坐标表示采用本发明实施例提供的频域失真估算方法得到的当前块的D′值。
图7为本发明实施例提供的又一种采用本发明的失真估算方法计算得到的失真与现有技术计算得到的失真的对比示意图。
如图7所示，每个点的横坐标表示采用HEVC标准中的现有空域失真计算方法得到的当前块的D值，其纵坐标表示采用本发明实施例提供的频域失真估算方法得到的当前块的D′值。
具体地，根据图6或图7中的选取的视频块点可见，图6或图7中点基本都在直线y＝x附近，表示采用本发明实施例提供的空域失真估算方法与现有空域失真计算方法两种失真计算方法得到的失真值D基本都相近。由此可见，本发明实施例提供的频域失真估算方法与现有技术计算的失真值非常接近。
需要说明的是，图6和图7中示意出了部分视频块的失真值计算对比，本发明实施例经过大量的计算以及示图结果表明，采用本发明实施例提供的失真估算方法和现有技术对任一视频块的失真计算，均满足图6和图7所示的规律。在此不予赘述。故本发明实施例提供的频域失真估算方法，其可靠性较高。
可以理解的是，对比图6和图7可知，本发明实施例提供的频域失真估算法得到的失真值与HEVC标准的计算结果较H.264标准的计算结果更接近。由于HEVC标准相比H.264更复杂，进一步验证，本发明实施例提供的失真估算方法可以很好的应用到变换大小以及预测模式更多的HEVC标准中。
本发明实施例通过设计频域的失真估计算法来取代原来利用空域的失真计算方法，以使得编码端无需对所有预测模式进行反变换操作，仅仅需要对选择出的最优模式进行完整的编码策略，且采用本发明实施例估算的失真值非常接近采用现有空域失真计算方法得到的失真值。因此，本发明实施例具有低复杂度、低功耗的优势，且可靠性高。
图8为本发明实施例提供的一种视频编码装置结构示意图；参照图8，该装置包括：
离散余弦变换单元801，用于对当前预测模式的预测残差进行离散余弦变换，得到变换系数，其中，预测残差为原始图像与预测图像的像素值残差，预测图像为根据原始图像的空间相关性和时间相关性按照当前预测模式预测 得到。
具体地，预测图像为根据帧间预测模式或帧内预测模式预测得到，所述帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，所述帧内预测模式为利用当前帧内空间相关性预测当前帧图像。
具体地，离散余弦变换单元801具体用于：在HEVC标准下，通过离散余弦变换矩阵对当前预测模式的预测残差进行离散余弦变换，得到变换系数。离散余弦变换矩阵正交可逆。或，在H.264标准下，通过整数离散余弦变换矩阵对当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，整数离散余弦变换矩阵非正交可逆。对整数离散余弦变换系数进行点乘，得到变换系数。
量化单元802，用于对变换系数进行量化，得到量化系数。
反量化单元803，用于对量化系数进行反量化，得到反量化系数。
失真值计算单元804，用于根据变换系数和反量化系数之差估算当前预测模式的失真值。
代价值计算单元805，用于根据当前预测模式的码率值和失真值得到当前预测模式代价值。
具体地，代价值计算单元805具体用于：对当前预测模式的量化系数进行熵编码，得到熵编码信息，根据熵编码信息确定码率值，其中，当前预测模式为多种预测模式中的任一种。根据当前预测模式的码率值和失真值通过率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
最优预测模式选取单元806，用于选取多种预测模式中代价值最小的预测模式为最优预测模式。
重建残差单元807，用于对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
具体各单元的工作流程可参见上述方法实施例的介绍，在此不予赘述。
可以理解的是，801—807各单元等是为了实现上述方法实施例的功能，其包含了执行各个功能相应的硬件结构和/或软件模块。本领域技术人员应该很容易意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，本发明能够以硬件或硬件和计算机软件的结合形式来实现。某个功能究竟以硬件还是计算机软件驱动硬件的方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
本发明实施例可以根据上述方法实施例对801—807等进行功能模块的划分，例如，可以对应各个功能划分各个功能模块，也可以将两个或两个以上的功能集成在一个处理模块中。上述集成的模块既可以采用硬件的形式实现，也可以采用软件功能模块的形式实现。需要说明的是，本发明实施例中对模块的划分是示意性的，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。
图9为本发明实施例提供的又一种视频编码装置结构示意图，如图9所示，包括：网卡901、存储器902、处理器903和总线904。
具体地，网卡901配置多个通信接口，终端通过通信接口采集或者接收视频，进行视频编解码。存储器902用于存储程序指令。网卡901、存储器902和处理器903通过总线904通信。
在一个示例中，处理器903，用于根据存储器902中存储的程序指令执行以下操作：对当前预测模式的预测残差进行离散余弦变换，得到变换系数。其中，预测残差为原始图像与预测图像的像素值之差，预测图像为根据帧间预测模式或帧内预测模式预测得到，帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，帧内预测模式为利用当前帧内空间相关性预测当前帧图像。对变换系数进行量化，得到量化系数。对量化系数进行反量化，得到反量化系数。根据变换系数和反量化系数之差得到所述当前预测模式的失 真值。根据所述当前预测模式的码率值和失真值得到当前预测模式代价值。选取多种预测模式中代价值最小的预测模式为最优预测模式。对最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
存储器902可以是一个存储装置，也可以是多个存储元件的统称，且用于存储运行会议服务器所需的程序以及数据等信息。且存储器902可以包括随机存取存储器(Random Access Memory，简称RAM)、闪存、只读存储器(Read Only Memory，简称ROM)、可擦除可编程只读存储器(Erasable Programmable ROM，简称EPROM)、电可擦可编程只读存储器(Electrically EPROM，简称EEPROM)、寄存器、硬盘、移动硬盘、只读光盘(CD-ROM)、闪存(Flash)或者本领域熟知的任何其它形式的存储介质等中的一个或多个存储介质的组合。
处理器903可以是CPU，通用处理器，DSP，专用集成电路(Application-Specific Integrated Circuit，简称ASIC)，现场可编程门阵列(Field Programmable Gate Array，简称FPGA)或者其他可编程逻辑器件、晶体管逻辑器件、硬件部件或者其任意组合。其可以实现或执行结合本发明公开内容所描述的各种示例性的逻辑方框，单元和电路。所述处理器也可以是实现计算功能的组合，例如包含一个或多个微处理器组合，DSP和微处理器的组合等等。
总线904可以是工业标准体系结构(Industry Standard Architecture，简称ISA)总线、外部设备互连(Peripheral Component，简称PCI)总线或扩展工业标准体系结构(Extended Industry Standard Architecture，简称EISA)总线等。该总线904可以分为地址总线、数据总线、控制总线等。为便于表示，图9中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
在一个示例中，处理器903根据当前预测模式的码率值和失真值通过率失真优化函数计算当前预测模式代价值，率失真优化函数用于权衡码率值和失真值。
在一个示例中，处理器903执行所述对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：
通过离散余弦变换矩阵对当前预测模式的预测残差进行离散余弦变换，得到变换系数。其中，离散余弦变换矩阵正交可逆。或，通过整数离散余弦变换矩阵对当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数。其中，整数离散余弦变换矩阵非正交可逆。对整数离散余弦变换系数进行点乘，得到变换系数。
进一步，总线904可用于连接图8中的各单元。处理器903可用于执行801—807各单元的功能，处理器902可用于存储801—807各单元的数据。
本发明实施例提供的视频编码方法及装置，采用频域失真估算方法，相比现有空域失真计算方法具有低复杂度、低功耗的优势，且可靠性高。
专业人员应该还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分步骤是可以通过程序来指令处理器完成，所述的程序可以存储于计算机可读存储介质中，所述存储介质是非短暂性(英文：non-transitory)介质，例如随机存取存储器，只读存储器，快闪存储器，硬盘，固态硬盘，磁带(英文：magnetic tape)，软盘(英文：floppy disk)，光盘(英文：optical disc)及其任意组合。
以上所述，仅为本发明较佳的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可 轻易想到的变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应该以权利要求的保护范围为准。

Claims (13)

  1. 一种视频编码方法，其特征在于，所述方法包括：
    对当前预测模式的预测残差进行离散余弦变换，得到变换系数，其中，所述预测残差为原始图像与预测图像的像素值之差，所述预测图像为根据帧间预测模式或帧内预测模式预测得到，所述帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，所述帧内预测模式为利用当前帧内空间相关性预测当前帧图像；
    对所述变换系数进行量化，得到量化系数；
    对所述量化系数进行反量化，得到反量化系数；
    根据所述变换系数和所述反量化系数之差得到所述当前预测模式的失真值；
    根据所述当前预测模式的码率值和失真值得到所述当前预测模式代价值；
    选取所述多种预测模式中代价值最小的预测模式为最优预测模式；
    对所述最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
  2. 如权利要求1所述的方法，其特征在于，所述根据所述当前预测模式的码率值和失真值得到所述当前预测模式代价值，包括：
    根据所述当前预测模式的码率值和失真值通过率失真优化函数计算所述当前预测模式代价值，所述率失真优化函数用于权衡所述码率值和所述失真值。
  3. 如权利要求1所述的方法，其特征在于，所述对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：
    通过离散余弦变换矩阵对所述当前预测模式的预测残差进行离散余弦变换，得到所述变换系数，所述离散余弦变换矩阵正交可逆。
  4. 如权利要求1所述的方法，其特征在于，所述对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：
    通过整数离散余弦变换矩阵对所述当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，所述整数离散余弦变换矩阵非正交可逆；
    对所述整数离散余弦变换系数进行点乘，得到所述变换系数。
  5. 一种视频编码装置，其特征在于，所述装置包括：
    离散余弦变换单元，用于对当前预测模式的预测残差进行离散余弦变换，得到变换系数；其中，所述预测残差为原始图像与预测图像的像素值之差，所述预测图像为根据帧间预测模式或帧内预测模式预测得到，所述帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，所述帧内预测模式为利用当前帧内空间相关性预测当前帧图像；
    量化单元，用于对所述变换系数进行量化，得到量化系数；
    反量化单元，用于对所述量化系数进行反量化，得到反量化系数；
    失真值计算单元，用于根据所述变换系数和所述反量化系数之差得到所述当前预测模式的失真值；
    代价值计算单元，用于根据所述当前预测模式的码率值和失真值得到所述当前预测模式代价值；
    最优预测模式选取单元，用于选取所述多种预测模式中代价值最小的预测模式为最优预测模式；
    重建残差单元，用于对所述最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
  6. 如权利要求5所述的装置，其特征在于，所述代价值计算单元具体用于：
    根据所述当前预测模式的码率值和失真值通过率失真优化函数计算所述当前预测模式代价值，所述率失真优化函数用于权衡所述码率值和所述失真值。
  7. 如权利要求5所述的装置，其特征在于，所述离散余弦变换单元具体用于：
    通过离散余弦变换矩阵对所述当前预测模式的预测残差进行离散余弦变换，得到所述变换系数，所述离散余弦变换矩阵正交可逆。
  8. 如权利要求5所述的装置，其特征在于，所述离散余弦变换单元具体用于：
    通过整数离散余弦变换矩阵对所述当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，所述整数离散余弦变换矩阵非正交可逆；
    对所述整数离散余弦变换系数进行点乘，得到所述变换系数。
  9. 一种视频编码装置，其特征在于，所述装置包括：
    存储器，用于存储程序指令；
    处理器，用于根据所述存储器中存储的程序指令执行以下操作：
    对当前预测模式的预测残差进行离散余弦变换，得到变换系数；其中，所述预测残差为原始图像与预测图像的像素值之差，所述预测图像为根据帧间预测模式或帧内预测模式预测得到，所述帧间预测模式为利用时间上相邻的前一帧图像预测下一帧图像，所述帧内预测模式为利用当前帧内空间相关性预测当前帧图像；
    对所述变换系数进行量化，得到量化系数；
    对所述量化系数进行反量化，得到反量化系数；
    根据所述变换系数和所述反量化系数之差得到所述当前预测模式的失真值；
    根据所述当前预测模式的码率值和失真值得到所述当前预测模式代价值；
    选取所述多种预测模式中代价值最小的预测模式为最优预测模式；
    对所述最优预测模式的反量化系数进行反离散余弦变换，得到重建残差。
  10. 如权利要求9所述的装置，其特征在于，所述处理器执行所述根据所述当前预测模式的码率值和失真值得到所述当前预测模式代价值，包括：
    根据所述当前预测模式的码率值和失真值通过率失真优化函数计算所述当前预测模式代价值，所述率失真优化函数用于权衡所述码率值和所述失真值。
  11. 如权利要求9所述的装置，其特征在于，所述处理器执行所述对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：
    通过离散余弦变换矩阵对所述当前预测模式的预测残差进行离散余弦变换，得到所述变换系数，所述离散余弦变换矩阵正交可逆。
  12. 如权利要求9所述的装置，其特征在于，所述处理器执行所述对当前预测模式的预测残差进行离散余弦变换，得到变换系数，包括：
    通过整数离散余弦变换矩阵对所述当前预测模式的预测残差进行整数离散余弦变换，得到整数离散余弦变换系数，所述整数离散余弦变换矩阵非正交可逆；
    对所述整数离散余弦变换系数进行点乘，得到所述变换系数。
  13. 一种计算机存储介质，其特征在于，所述计算机存储介质存储有程序，所述程序执行的步骤包括如权利要求1—4中任一项所述的步骤。
PCT/CN2016/090884 2016-07-21 2016-07-21 视频编码方法及装置 WO2018014301A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2016/090884 WO2018014301A1 (zh) 2016-07-21 2016-07-21 视频编码方法及装置
CN201680054222.9A CN108028938A (zh) 2016-07-21 2016-07-21 视频编码方法及装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2016/090884 WO2018014301A1 (zh) 2016-07-21 2016-07-21 视频编码方法及装置

Publications (1)

Publication Number Publication Date
WO2018014301A1 true WO2018014301A1 (zh) 2018-01-25

Family

ID=60992787

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/090884 WO2018014301A1 (zh) 2016-07-21 2016-07-21 视频编码方法及装置

Country Status (2)

Country Link
CN (1) CN108028938A (zh)
WO (1) WO2018014301A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111757104A (zh) * 2020-06-29 2020-10-09 北京百度网讯科技有限公司 一种图像编码方法、装置、电子设备及存储介质
CN112866695A (zh) * 2021-01-07 2021-05-28 上海富瀚微电子股份有限公司 一种视频编码器
CN113225556A (zh) * 2021-05-31 2021-08-06 杭州当虹科技股份有限公司 视频编码方法
CN113301347A (zh) * 2021-05-08 2021-08-24 广东工业大学 一种hevc高清视频编码的优化方法
CN113592966A (zh) * 2021-08-04 2021-11-02 上海哔哩哔哩科技有限公司 图像处理方法及装置、电子设备和存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006262075A (ja) * 2005-03-17 2006-09-28 Pioneer Electronic Corp 画像符号化装置、画像符号化方法、および画像符号化用プログラム
CN101415121A (zh) * 2007-10-15 2009-04-22 华为技术有限公司 一种自适应的帧预测的方法及装置
US20120183051A1 (en) * 2011-01-19 2012-07-19 Samsung Electronics Co., Ltd. Method for video encoding mode selection and video encoding apparatus performing the same
CN102647593A (zh) * 2012-04-18 2012-08-22 北京大学 一种avs帧内模式决策方法和装置
CN103501437A (zh) * 2013-09-29 2014-01-08 北京航空航天大学 一种基于分形和h.264的高光谱图像压缩方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006262075A (ja) * 2005-03-17 2006-09-28 Pioneer Electronic Corp 画像符号化装置、画像符号化方法、および画像符号化用プログラム
CN101415121A (zh) * 2007-10-15 2009-04-22 华为技术有限公司 一种自适应的帧预测的方法及装置
US20120183051A1 (en) * 2011-01-19 2012-07-19 Samsung Electronics Co., Ltd. Method for video encoding mode selection and video encoding apparatus performing the same
CN102647593A (zh) * 2012-04-18 2012-08-22 北京大学 一种avs帧内模式决策方法和装置
CN103501437A (zh) * 2013-09-29 2014-01-08 北京航空航天大学 一种基于分形和h.264的高光谱图像压缩方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHANG YANG ET AL.: "Fast distortion and rate estimation algorithm for H.264 video coding", JOURNAL OF TSINGHUA UNIVERSITY SCIENCE AND TECHNOLOGY, vol. 49, no. 7, 15 July 2009 (2009-07-15), pages 961 - 964, ISSN: 1000-0054 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111757104A (zh) * 2020-06-29 2020-10-09 北京百度网讯科技有限公司 一种图像编码方法、装置、电子设备及存储介质
CN111757104B (zh) * 2020-06-29 2022-11-18 北京百度网讯科技有限公司 一种图像编码方法、装置、电子设备及存储介质
CN112866695A (zh) * 2021-01-07 2021-05-28 上海富瀚微电子股份有限公司 一种视频编码器
CN113301347A (zh) * 2021-05-08 2021-08-24 广东工业大学 一种hevc高清视频编码的优化方法
CN113225556A (zh) * 2021-05-31 2021-08-06 杭州当虹科技股份有限公司 视频编码方法
CN113592966A (zh) * 2021-08-04 2021-11-02 上海哔哩哔哩科技有限公司 图像处理方法及装置、电子设备和存储介质
CN113592966B (zh) * 2021-08-04 2024-04-30 上海哔哩哔哩科技有限公司 图像处理方法及装置、电子设备和存储介质

Also Published As

Publication number Publication date
CN108028938A (zh) 2018-05-11

Similar Documents

Publication Publication Date Title
WO2016138779A1 (zh) 帧内编解码方法、编码器和解码器
WO2018014301A1 (zh) 视频编码方法及装置
TWI469644B (zh) 使用辭典編碼視訊之方法
JP5406222B2 (ja) 連続的な動き推定を利用した映像符号化並びに復号化方法及び装置
JP4987086B2 (ja) 画像符号化方法及び復号方法、それらの装置、それらのプログラム並びにプログラムを記録した記録媒体
US10085028B2 (en) Method and device for reducing a computational load in high efficiency video coding
US10009611B2 (en) Visual quality measure for real-time video processing
MX2014008778A (es) Decodificador de video, codificador de video, metodo de decodificacion de video, y metodo de codificacion de video.
JP2013502140A (ja) 動きベクトルの正確度の調節を利用した映像符号化、復号化方法及び装置
US9014499B2 (en) Distributed source coding using prediction modes obtained from side information
WO2019148906A1 (zh) 视频编码方法、计算机设备和存储介质
JP2009290498A (ja) 画像符号化装置及び画像符号化方法
JP2006511113A (ja) 選択されたマクロブロックにおける動き推定のスキップを伴うビデオエンコーディング
BR112013032716B1 (pt) método de decodificação de vídeo e método de codificação de vídeo para compensação de movimento
JP2008502261A (ja) 圧縮コーディングおよび電力消費量を含む費用関数を使用して画像をメモリに格納する方法
KR20110101443A (ko) 움직임 벡터 해상도 조합을 이용한 움직임 벡터 부호화/복호화 방법 및 장치와 그를 이용한 영상 부호화/복호화 방법 및 장치
US8792549B2 (en) Decoder-derived geometric transformations for motion compensated inter prediction
JP6564315B2 (ja) 符号化装置、復号装置、及びプログラム
KR101035746B1 (ko) 동영상 인코더와 동영상 디코더에서의 분산적 움직임 예측 방법
JP5468383B2 (ja) ビデオストリームの圧縮を最適化するための方法および装置
WO2022061563A1 (zh) 视频编码方法、装置及计算机可读存储介质
WO2019150435A1 (ja) 映像符号化装置、映像符号化方法、映像復号装置、映像復号方法、及び映像符号化システム
CN108900842B (zh) Y数据压缩处理方法、装置、设备及WebP压缩***
JP2018186419A (ja) トランスコーダ、トランスコード方法、及びトランスコードプログラム
JP4799377B2 (ja) 画像符号化装置

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: 16909228

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: 16909228

Country of ref document: EP

Kind code of ref document: A1