WO2011110088A1 - 一种视频预测编码方法和装置 - Google Patents

一种视频预测编码方法和装置 Download PDF

Info

Publication number
WO2011110088A1
WO2011110088A1 PCT/CN2011/071611 CN2011071611W WO2011110088A1 WO 2011110088 A1 WO2011110088 A1 WO 2011110088A1 CN 2011071611 W CN2011071611 W CN 2011071611W WO 2011110088 A1 WO2011110088 A1 WO 2011110088A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
reconstructed
value
pixels
block
Prior art date
Application number
PCT/CN2011/071611
Other languages
English (en)
French (fr)
Inventor
陶品
吴文婷
肖谋
温江涛
谷沉沉
吕静
盛馥钟
Original Assignee
清华大学
腾讯科技（深圳）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 清华大学, 腾讯科技（深圳）有限公司 filed Critical 清华大学
Priority to MX2012010518A priority Critical patent/MX2012010518A/es
Priority to BR112012022951-6A priority patent/BR112012022951B1/pt
Priority to SG2012065314A priority patent/SG183888A1/en
Priority to RU2012142523/08A priority patent/RU2536366C2/ru
Priority to CA2797569A priority patent/CA2797569C/en
Publication of WO2011110088A1 publication Critical patent/WO2011110088A1/zh
Priority to US13/608,064 priority patent/US8654845B2/en

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/59—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
    • H04N19/436—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation using parallelised computational arrangements
    • 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
    • H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding

Definitions

  • Video predictive coding method and device
  • the present invention relates to the field of video processing technologies, and more particularly, to a video prediction encoding method and apparatus. Background of the invention
  • video is generally divided into I frames, P frames, and B frames for encoding.
  • the I frame is an intra-coded frame.
  • the encoding process only the information in the frame can be used for prediction.
  • decoding only the code stream of the frame can be used to decode and reconstruct the frame image.
  • the encoding and decoding process of the I frame is independent, independent of the information of the previous and succeeding frames.
  • the P frame and the B frame can also perform interframe prediction.
  • the interframe prediction technique needs to search for a block that is most similar to the current macroblock in the reference frame as a prediction, and at the time of decoding.
  • decoding is performed by finding a prediction block from the decoded reference frame according to the motion vector information.
  • the encoding and decoding process of P frame and B frame are not independent, and the time complexity is high. However, the time correlation between frames and the spatial correlation within the frame are utilized, and efficient compression can be realized.
  • I frames are only predicted by the spatial correlation in the frame. The encoding process is independent and the complexity is low, but the compression efficiency of I frames is far less than that of P frames and B frames.
  • Intra prediction is currently mainly used in the encoding process of I frames.
  • a P frame or a B frame when the interframe motion search technique is difficult to search for an accurate prediction block, there are also a small number of macros.
  • the block uses intra prediction.
  • intra prediction uses a spatial prediction algorithm. Each pixel of the coded block is predicted using a pixel value of 128 instead of using the information of the coded frame itself. For most sequences, this method predicts a large residual, so the I frame compression efficiency of this intra prediction method is ⁇ .
  • the intra-coded block uses a DC/AC prediction algorithm in the frequency domain. The coding block first performs DCT transform to the frequency domain, and uses the DC/AC coefficients of the adjacent block to predict the coefficients of the current block. With the MPEG-4 method, the compression of I frames is improved.
  • an intra prediction algorithm for direction prediction is employed.
  • the algorithm predicts from a certain direction for each 16x16, 8x8 or 4x4 pixel block using its upper, left, upper left and upper right coded blocks.
  • the direction prediction technique greatly improves the compression efficiency of the I frame, but in the prediction process of the block as the prediction unit, only the correlation of the pixels between the blocks is utilized, and the correlation between adjacent pixels in the block is used. , still not fully utilized. For the whole block, pixels far from the predicted pixel in the block are difficult to obtain accurate prediction, so the intra prediction residual is large, especially for blocks with complex texture information, it is difficult to achieve good prediction effect, which directly leads to The compression efficiency of intraframe coding is low, which in turn affects the compression efficiency of the entire video sequence.
  • Embodiments of the present invention provide a video predictive coding method to improve compression efficiency of video coding.
  • Embodiments of the present invention also provide a video predictive coding apparatus to improve video coding. Compression efficiency.
  • a video predictive coding method comprising:
  • the main set pixel reconstruction value and the residual set pixel reconstruction value are combined to obtain a reconstructed value of the pixel block.
  • a video predictive coding apparatus comprising: a pixel dividing unit, a main set pixel encoding unit, a residual set pixel encoding unit, and a pixel block reconstructed value combining unit, wherein:
  • a pixel dividing unit configured to extract a pixel block from a current frame, and divide the pixel block into a main set pixel and a residual set pixel;
  • a main set pixel coding unit configured to encode the main set pixel, output a main set pixel code stream, and obtain a main set pixel reconstruction value
  • a residual pixel encoding unit configured to interpolate the reconstructed pixel value including the primary set pixel reconstructed value as a predicted value of the residual set pixel, and perform intra prediction encoding on the remaining set pixel, and output Remainder pixel code stream, and obtain a residual set pixel reconstruction value;
  • a pixel block reconstruction value combining unit configured to combine the main set pixel reconstructed value and the residual set pixel reconstructed value in a manner corresponding to the pixel division to obtain a reconstructed value of the pixel block.
  • the pixel block is first taken out from the current frame, and the pixel block is divided into a main set pixel and a residual set pixel, and then the main set image is Encoding, outputting the main set pixel code stream, and obtaining the main set pixel reconstruction value, and interpolating the reconstructed pixel values including at least the main set pixel reconstruction value, and performing intra prediction encoding on the residual set pixel
  • the residual pixel code stream is output, and the residual pixel reconstruction value is obtained.
  • the main set pixel reconstruction value and the residual set pixel reconstruction value are combined to obtain a reconstructed value of the pixel block.
  • the I frame encoding of the embodiment of the present invention for H.264/AVC is decomposed into even macroblocks and odd macroblocks by horizontal slicing sampling, which can reduce the code stream overhead of coding intra prediction mode flag information, thereby enabling Further improve the compression efficiency.
  • the fixed prediction mode saves the code stream overhead of the prediction mode flag information; and for even macroblocks, the improved intra prediction mode predictive coding method enables the code stream overhead of the intra prediction mode flag information to be larger. Limit savings.
  • the embodiments of the present invention have lower implementation complexity.
  • the calculation of the 6-tap interpolation filter uses only 19x16 pixels, of which 16x16 pixels are from the current 32x16 blocks belonging to even macroblocks.
  • the pixels, 3x16 pixels from the left reconstructed neighbor macroblock do not need to cache other rows of macroblocks during the encoding process, thereby greatly saving memory usage and improving the cache hit rate.
  • this low-memory-occupancy solution is easy to implement on a chip.
  • embodiments of the present invention are well suited for implementation of highly parallel computing, enabling fast encoding compression of high definition video sources.
  • FIG. 1 is a flowchart of a video predictive coding method according to an embodiment of the present invention
  • FIG. 2 is a schematic view showing sampling of an interlaced spacer according to an embodiment of the present invention.
  • 3 is a schematic diagram of a checkerboard sampling according to an embodiment of the present invention
  • 4 is a schematic diagram of sampling of a chessboard of a 2 ⁇ 1 pixel block by a sampling unit according to an embodiment of the present invention
  • FIG. 5 is a flowchart of encoding an I frame luminance component according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of macroblock decomposition according to an embodiment of the present invention.
  • FIG. 7 is a schematic diagram of selection of 4 ⁇ 4 luma block prediction pixels according to an embodiment of the present invention
  • FIG. 8 is a schematic diagram of even macroblock interpolation as an odd macroblock prediction value according to an embodiment of the present invention
  • FIG. 9 is a schematic diagram of prediction of an even macroblock intra prediction mode according to an embodiment of the present invention.
  • FIG. 10 is a schematic diagram showing an overall frame structure of a video predictive coding apparatus according to an embodiment of the present invention.
  • FIG. 11 is a schematic structural diagram of a video predictive coding apparatus according to an embodiment of the present invention.
  • FIG. 12 is a comparison diagram of rate distortion curves of a BigShips sequence according to an embodiment of the present invention. Mode for carrying out the invention
  • Defects of the present invention provide a video predictive coding method based on adjacent pixel correlation to achieve the purpose of overcoming the above drawbacks and improving coding efficiency.
  • FIG. 1 is a flowchart of a video predictive coding method according to an embodiment of the present invention.
  • the method includes:
  • Step 101 Extract a pixel block from the current frame, and divide the pixel block into a main set pixel and a residual set pixel.
  • an M*N-sized pixel block may be taken from the current encoded frame, and the M*N-sized pixel is referred to as a primary block, where M is the width of the pixel block and N is the height of the pixel block.
  • M*N The pixel block of the size may be a luma block or a chroma block, which is not limited in the embodiment of the present invention.
  • some pixels in the M*N-sized pixel block are extracted according to a certain sampling rule (the number of the pixels is set as the main pixel); the remaining pixels are recorded as the residual pixels. That is, a pixel is sampled from the original block of M*N size to form a main set pixel, and the remaining pixels are composed into a residual set pixel.
  • M should be less than or equal to the frame width of the current frame, and N is less than or equal to the current frame.
  • the frame height is high, and P ⁇ M * N (that is, the main set pixel size must be smaller than the original block size;).
  • the pixel block may be divided into a main set pixel and a residual set pixel according to an interlaced block sampling method or according to a checkerboard sampling mode (or a checkerboard sampling method).
  • the pixels represented by the gray squares constitute the main set pixels. If the main set pixels are arranged in the order of the initial position into the * « size block, that is, the width of the main set pixels is, the height of the main set pixels is ", then, n) M, N The relationship between them is as shown in equation (1).
  • FIG. 4 is a schematic diagram of a checkerboard sampling of a 2x1 pixel block by a sampling unit according to an embodiment of the present invention. As shown in FIG. 3, pixels represented by gray squares (or white squares) are sampled to form a main set pixel.
  • the main set pixels may constitute a coding block according to the coding requirements, and the remaining set pixels are composed of the remaining pixels after the main set pixels are removed from the original block, and therefore, the remaining set pixel size It may be the same as or different from the main set pixel, and the coding block composed of the remaining set pixels does not necessarily have the same shape and number as the main set pixel.
  • Step 102 Encode the main set pixel, output the main set pixel code stream, and obtain the main set pixel reconstruction value.
  • the prediction method of the main set pixels differs depending on the currently encoded frame type. If the current frame is an I frame, the main set pixel can be predicted by a similar H.264/AVC direction prediction algorithm; if the current frame is a P frame or a B frame, the main set pixel can adopt a H.264/AVC direction prediction algorithm or frame. Inter-sport search technology for prediction.
  • the type of the current frame is first determined, and when the type of the current frame is an I frame, the H.264/AVC direction prediction algorithm is used to encode the main set pixel, if the current frame type is a P frame or a B frame.
  • the main set of pixels may be encoded using a H.264/AVC direction prediction algorithm or an interframe motion search algorithm.
  • H.264/AVC is currently the most efficient video coding standard.
  • intra-coded blocks use directional prediction techniques for intra prediction.
  • Direction prediction uses different prediction modes for blocks of different sizes.
  • the block size is 16x16 and 4x4, of which 16x16 luma block (corresponding macroblock type is I16MB) has 4 prediction modes, 4x4 luma block (corresponding macroblock type is I4MB), and there are 9 prediction modes.
  • the block size is 8x8 and there are 4 prediction modes.
  • the 16x16 luma block (or 8x8 chroma block) has four prediction modes, namely horizontal prediction, vertical prediction, DC prediction and plane prediction.
  • the predicted value of each row of pixels is equal, and the predicted value is the left adjacent The boundary value of the reconstructed block in the row; in the vertical prediction, the predicted value of each column of pixels is equal, and the predicted value is the boundary pixel value of the adjacent reconstructed block in the column; when DC prediction, the entire block is used
  • the predicted value is 16 (or 8) boundary pixels of the adjacent reconstructed block on the top of the block, 16 (or 8) boundary pixels of the left adjacent reconstructed block, and the upper left corner
  • the neighboring pixels are averaged. If some of the pixels do not exist (the current macroblock is located at the upper or left boundary of the image), the predicted value is obtained by averaging the existing pixels.
  • the left side of the current block is used.
  • the boundary pixels of the adjacent reconstructed block above are predicted by a linear function for each pixel in the current block.
  • the corresponding prediction modes are vertical prediction (mode number 0), horizontal prediction (mode number 1), diagonal lower left prediction (mode number 3), and diagonal lower right prediction ( Mode number 4), vertical right prediction (mode number 5), horizontal downward prediction (mode number 6), vertical left prediction (mode number 7), and horizontal left prediction (mode number 8), together with DC Prediction (Mode No. 2) - A total of 9 prediction modes, each of which predicts the pixels of the current block using the reconstructed pixels in the corresponding direction of the mode.
  • Step 103 Interpolating the reconstructed pixel values including the primary set pixel reconstructed value to obtain a predicted value of the residual set pixel, performing intra prediction encoding on the remaining set pixel, and outputting the residual set pixel code stream And get the residual set pixel reconstruction value.
  • the reconstructed pixel value may be either the main set pixel reconstruction value itself, or the main set pixel reconstruction value and the reconstructed value of other pixel blocks except the pixel block.
  • the reconstructed pixel value includes only the main set pixel reconstruction.
  • only the main set pixel reconstruction value is interpolated to obtain the predicted value of the residual pixel, and then the residual set pixel is intra-predictively encoded, the residual set pixel code stream is output, and the residual set pixel reconstructed value is obtained.
  • all of the reconstructed pixel values including the main set pixel reconstruction values and the reconstructed pixel values of other pixel blocks are commonly used for interpolation.
  • the value obtains the predicted value of the residual set pixel, performs intra prediction encoding, outputs the residual set pixel code stream, and obtains the residual set pixel reconstructed value.
  • each pixel in the remaining set of pixels is classified according to a position of each pixel in the remaining set of pixels, wherein pixels with the same degree of adjacent information are placed in the same category; and then, for the remaining set of pixels of each classification Interpolating all reconstructed pixel values including the main set pixel reconstructed value, the reconstructed residual set pixel reconstructed value, and the reconstructed pixel values of other pixel blocks, respectively, to obtain a predicted value, and then layering The intra prediction coding of the structure.
  • the residual pixels are predicted using reconstructed values of all reconstructed pixel values, particularly the main set pixels.
  • the remaining pixels can be reclassified and a hierarchical structure is adopted. Make predictions.
  • the remaining pixels are mainly classified according to the position of the remaining pixels. The pixel position determines the degree to which the adjacent information of the pixel is known, and the remaining pixels having the same degree of adjacent information are placed in the same category.
  • the same type of pixels adopt a unified prediction mode, and different types of residual pixels adopt different prediction modes.
  • the underlay prediction i.e., the prediction of the post-coded pixels
  • the higher-level prediction that is, the prediction of the first-coded pixel
  • different types of pixels may select different quantization precision according to the accuracy of the prediction, and adopt higher quantization precision for the higher-level prediction residual data.
  • the predicted value of the residual pixel is mainly obtained by interpolating the reconstructed pixel value by using an interpolation filter.
  • an interpolation filter For hierarchical prediction, multiple filters can be designed based on the residual pixel of each layer. There are many different filter designs. You can directly copy the reconstructed neighboring pixel values, you can also use the H.264/AVC 6-tap filter, you can also design other types of filters, or you can use a variety of filters to interpolate and select the most. Good prediction results.
  • Step 104 Combine the main set pixel reconstruction value and the residual set pixel reconstruction value to obtain an image The reconstructed value of the prime block.
  • the reconstructed value of the main set pixel obtained in step 102 and the reconstructed value of the residual set pixel obtained in step 103 are recombined to obtain the reconstructed value of the original block, which is the inverse of the sampling mode in step 101.
  • the encoding process of the embodiment of the present invention will be described below by taking the I frame luminance component encoding of the video image as an example.
  • FIG. 5 is a flow chart of encoding an I frame luminance component according to an embodiment of the present invention.
  • the implementation flow of the present invention is described by taking a 32x16 luminance block having a width of 32 and a height of 16 (where 32 corresponds to the aforementioned M, 16 corresponds to the aforementioned N) as an example.
  • 32x16 luminance block having a width of 32 and a height of 16 (where 32 corresponds to the aforementioned M, 16 corresponds to the aforementioned N) as an example.
  • 32x16 luminance block having a width of 32 and a height of 16 (where 32 corresponds to the aforementioned M, 16 corresponds to the aforementioned N) as an example.
  • 32x16 luminance block is for illustrative purposes only and is not intended to limit the scope of the invention.
  • the method includes:
  • Step 501 Extract 32x16 luma blocks (corresponding to two natural macroblocks consecutive in the horizontal direction) at the position of the current encoded frame, and perform horizontal decomposition on the 32x16 luma block, wherein the even columns constitute an even macroblock, and the odd columns constitute An odd macroblock corresponding to an even macroblock.
  • the even columns of the 32x16 luma block (columns numbered 2, 4, 6 16) are extracted to form an even macroblock, and the odd columns of the 32x16 luma block (numbered 1, 3, 5)
  • the column of 15 is extracted to form an odd macroblock, as shown in Fig. 6.
  • Step 502 Determine whether the current coding position reaches the right boundary macroblock of the frame, whether there is only one single macroblock remaining and cannot extract the luminance block of 32 ⁇ 16 , if yes, execute step 503, otherwise step 504 is performed.
  • Step 503 Encode the individual macroblocks by using a direction prediction algorithm, and perform step 507. Specifically, for a single macroblock remaining on the right boundary, a prediction method is used to perform prediction, and the prediction residual is transformed, quantized, and entropy encoded to output the block code stream, and the amount is simultaneously The inverse coefficient is inverse-transformed to obtain a reconstructed block of the block.
  • Step 504 The dual macroblock is predicted by using a direction prediction algorithm, and the prediction residual is transformed, quantized, and entropy encoded to output the block code stream, and the quantized coefficients are inverse quantized and inversely transformed to obtain a reconstructed block of the block.
  • the even macroblock still has four prediction modes (ie, horizontal prediction, vertical prediction, DC prediction, and planar prediction) and I4MB for I16MB (a macroblock type that performs intra prediction encoding on a 16x16 macroblock as a whole).
  • 9 prediction modes (dividing 16x16 macroblocks into 16 4x4 blocks, intra-predicted macroblock types for each block) (ie, vertical prediction (mode number 0), horizontal prediction (mode number 1) , diagonal lower left prediction (mode number 3), diagonal lower right prediction (mode number 4), vertical right prediction (mode number 5), horizontal downward prediction (mode number 6), vertical left prediction (Mode No. 7), Horizontal Left Prediction (Mode No. 8), and DC Prediction (Mode No.
  • the prediction method and the position of the input pixel are consistent with the prediction method of the 16x16 luma block in the prior art direction prediction algorithm. If the current macroblock is decomposed into 16 4x4 blocks for prediction, although the prediction mode remains at nine, the position of the pixel used for prediction changes. As shown in Fig. 6, the gray dots represent the pixels that make up the even macroblock, the white dots represent the pixels that make up the odd macroblock, and the gray dots in the rectangle frame constitute the currently encoded 4x4 block, which is used for prediction.
  • the pixel is AM, wherein the pixel AD is from an adjacent even macroblock above the current macroblock, and the pixel IM is from the adjacent even macroblock to the left of the current macroblock, which is performed with the prior art direction prediction algorithm for the 4x4 luma block.
  • the prediction of the predicted pixel point is consistent, but the pixel EH position is changed, and the pixel EH is derived from the nearest natural neighboring block at the upper right of the current macroblock.
  • Step 505 Predicting the odd macroblock by using the reconstructed pixel of the even macroblock, where the reconstructed even macroblock is interpolated by using a 6-tap filter of H.264/AVC to obtain a predicted value of the odd macroblock. Transforming, quantizing, and entropy encoding the prediction residual of the odd macroblock to output the code stream, The inverse quantized inverse transform of the quantized coefficients yields a reconstructed block of the block.
  • the odd macroblock is predicted using the reconstructed pixel block of the even macroblock.
  • the even macroblock is reconstructed, the adjacent pixels around the odd macroblock have been reconstructed.
  • the 6-tap interpolation filter can be used to interpolate the reconstructed pixel values of the even macroblock to obtain the prediction of the current odd macroblock. value.
  • the origin of the gray indicates the pixel that has been reconstructed in the even macroblock, and the origin of white indicates the pixel to be encoded of the current odd macroblock.
  • the current odd macroblock to be encoded pixel X its predicted value can be pre-expressed by X.
  • the reconstructed value of the reconstructed pixels A, B, C, D, E, and F located around the pixel X can be used to perform interpolation as the predicted value X 3 ⁇ 4 of the pixel X to be encoded of the current odd macroblock by applying the following equation (2).
  • round is a rounding rounding function
  • D is "the reconstructed value of the closest reconstructed pixel D to the right of pixel X;
  • step 505 each pixel of the odd macroblock is separately predicted by its neighboring pixels, making full use of the horizontal correlation between the pixels, and the prediction result is more accurate than the direction prediction. Especially for blocks with complex textures, the prediction effect of this prediction method is significantly improved, which also directly leads to an improvement in coding performance.
  • the application of step 505 also brings another advantage, that is, the intraframe coding method of the embodiment of the present invention is significantly less complicated than the H.264/AVC time complexity. Due to the application of step 505, nearly half or half of the currently encoded I frames are predicted by a single 6-tap interpolation filter, and no high complexity mode decision is required. This greatly reduces the computational complexity.
  • the conventional intra coding uses the prediction modes of the block U and the block Lo to predict the prediction mode of the current block (number 1), and the use of step 505 causes the left adjacent macroblock of the current even macroblock to be changed.
  • the prediction mode of the current block (number 1) is predicted by using the prediction modes of the block Le and the block U located in the nearest even macroblock on the left side of the current even macroblock, and this also belongs to this. Also included are blocks of numbers 5, 9, and 13 in the current even macroblock and blocks in the left boundary position among other even macroblocks.
  • Step 506 Combine the reconstructed block of the even macroblock obtained in step 504 with the reconstructed block of the odd macroblock obtained in step 505 to obtain a reconstructed block of the original 32x16 luma block, where each column of the even macroblock corresponds to 32x16 For even columns in a block, each column of an odd macroblock corresponds to an odd column in a 32x16 block.
  • Step 507 Determine whether the current frame is encoded. If the encoding is completed, the process ends. Otherwise, go to step 501 to continue encoding the next luma block.
  • an embodiment of the present invention also proposes a video predictive encoding device.
  • Figure 10 is a diagram showing the overall frame structure of a video predictive encoding apparatus according to an embodiment of the present invention.
  • the apparatus includes a pixel dividing unit 1001 and a main set pixel encoding unit. 1002, the remainder set pixel encoding unit 1003 and the pixel block reconstruction value combining unit 1004, wherein: the pixel dividing unit 1001 is configured to extract a pixel block from the current frame, and divide the pixel block into a main set pixel and a residual set pixel;
  • the main set pixel encoding unit 1002 is configured to encode the main set pixel, output a main set pixel code stream, and obtain a main set pixel reconstruction value;
  • the residual pixel encoding unit 1003 is configured to perform interpolation on the reconstructed pixel value including the primary set pixel reconstructed value as a predicted value of the residual set pixel, and perform intra prediction encoding on the remaining set pixel, Outputting a residual set of pixel code streams, and obtaining a residual set pixel reconstruction value;
  • the pixel block reconstruction value combining unit 1004 is configured to combine the main set pixel reconstruction value and the residual set pixel reconstruction value in a manner corresponding to the pixel division to obtain a reconstructed value of the pixel block.
  • the main set pixel reconstruction value obtaining unit 1002 is configured to determine the type of the current frame, and when the current frame type is an I frame, the H.264/AVC direction prediction algorithm is used for encoding, when the current frame is used.
  • the type is P frame or B frame
  • encoding is performed using a H.264/AVC direction prediction algorithm or an interframe motion search algorithm.
  • the residual set pixel reconstruction value obtaining unit 1003 is configured to classify each pixel in the residual set pixel according to the position of each pixel in the residual set pixel, where the adjacent information is known to have the same degree of pixels In the same category, and for the residual pixels of each classification, the interpolation filter is used to respectively include the reconstructed values of the main set pixel reconstructed value, the reconstructed residual set pixel reconstructed value, and the reconstructed pixel values of other pixel blocks. All reconstructed pixel values are interpolated to obtain predicted values, and then intra-predictive coding of the hierarchical structure is performed.
  • the interpolation filter included in the residual set pixel reconstruction value obtaining unit 1003 may be a 6-tap filter of H.264/AVC, or may be another filter. When a 6-tap filter is used, a 6-tap interpolation filter can be used to interpolate the reconstructed pixels as a residual set of pixel prediction values for each classification.
  • FIG. 11 is a schematic structural diagram of a video predictive encoding apparatus according to an embodiment of the present invention. In Figure 11, the apparatus is used for predictive coding of an M*N block of pixels.
  • the apparatus includes a sampling unit, configured to extract a block of M*N size from a current encoded frame, and extract a partial pixel group in the block as a main set according to a certain sampling rule; the remaining pixels are recorded as The remainder.
  • a sampling unit configured to extract a block of M*N size from a current encoded frame, and extract a partial pixel group in the block as a main set according to a certain sampling rule; the remaining pixels are recorded as The remainder.
  • the main set pixel can be predicted by using a prediction mode suitable for the frame type, and the prediction residual is transformed, quantized and entropy encoded to output the main set pixel code stream, and the quantized coefficient is inverse quantized.
  • the reconstructed pixel values including the reconstructed values of the main set pixels are first interpolated as the predicted values of the residual set pixels, and then the prediction residuals are transformed, quantized, and entropy encoded to output the residual set pixel codes.
  • the stream at the same time, inversely quantizes and inverse transforms the quantized coefficients, and adds the predicted values of the remaining pixels to obtain the reconstructed values of the residual pixels.
  • the main set pixel reconstruction value and the residual set pixel reconstruction value are combined in a manner corresponding to pixel division to obtain a reconstructed value of the pixel block.
  • the circle indicates that the two data streams are added.
  • the plus sign means that the data stream is positive, and the minus sign means that the data stream is negative.
  • the sampling unit extracts the M*N-sized block from the current encoded frame, and extracts a part of the pixel group in the block as a main set pixel according to a certain sampling rule; the remaining pixels are recorded as the remaining set pixels.
  • the prediction method suitable for the frame type is used for prediction.
  • the main set pixel prediction residual is subtracted from the main set pixel original value to obtain the main set pixel prediction residual, and then the node A is sequentially operated downward to transform, quantize and entropy the main set pixel prediction residual.
  • node A calculates the prediction residual coefficient after transform and quantization, and needs to undergo inverse quantization and inverse transformation in turn. To get a difference value.
  • the node B adds the difference value to the predicted value of the main set pixel to obtain a reconstructed value of the main set pixel.
  • the reconstructed pixel value including the main set pixel reconstruction value is first interpolated as the predicted value of the residual set pixel, and then the residual value is subtracted from the residual set pixel original value at the node D.
  • the pixel prediction value is collected to obtain a residual pixel prediction residual, and then the residual pixel prediction residual is subjected to transformation, quantization and entropy coding in turn, and then the residual pixel code stream is output.
  • the node D calculates the prediction residual coefficient after the transform and quantization, and needs to undergo inverse quantization and inverse change in turn to obtain a difference value, and then computes the difference value and the predicted value of the residual pixel at the node C to obtain the remaining value. Sets the reconstructed value of the pixel.
  • the embodiment of the present invention reduces the code stream overhead of the intra prediction mode flag information, which is another reason for the improvement of compression efficiency.
  • the fixed prediction mode saves the code stream overhead of the prediction mode flag information; and for even macroblocks, the improved intra prediction mode predictive coding method enables the code stream overhead of the intra prediction mode flag information to be larger. Limit savings.
  • the intra coding scheme in the embodiment of the present invention uses the reconstructed pixel block of the coded even macroblock for the odd macroblock to perform predictive coding using a unified 6-tap interpolation filter, so that the currently encoded I In the frame, nearly half or half of the pixels do not need to perform high-complexity mode decision calculation, so the computational complexity is greatly reduced compared to the H.264/AVC intra prediction algorithm.
  • the technical solution of the present invention has lower implementation complexity.
  • the calculation of the 6-tap interpolation filter uses only 19x16 pixels, of which 16x16 pixels are from the pixels belonging to the even macroblock in the current 32x16 block, 3x16 pixels.
  • the reconstructed neighbor macroblock from the left side does not need to cache macroblocks of other rows during the encoding process, thereby greatly saving memory usage and improving the cache hit rate.
  • this low-memory-consuming solution is easy to implement on a chip.
  • the technical solution of the present invention is suitable for the implementation of highly parallel computing, so that high-definition video source can be quickly encoded and compressed.
  • the reference software adopts the reference software JM15.0 of the H.264/AVC standard, and can obtain an average of 0.39 dB on a sequence with a resolution of 720p. PSNR gain or 7.31% code rate savings.
  • FIG. 12 is a graph showing a rate distortion curve of a BigShips sequence in accordance with an embodiment of the present invention.
  • Figure 12 shows the rate distortion curve comparison on a BigShips sequence with a resolution of 720p and a frame rate of 30fps.
  • the curve above the graph is a rate distortion curve on a BigShips sequence with a resolution of 720p and a frame rate of 30fps according to an embodiment of the present invention
  • the curve below the image is a BigShips sequence with a resolution of 720p and a frame rate of 30fps.
  • embodiments of the present invention can achieve a peak signal to noise ratio (PSNR) gain of 0.79 dB or a code rate savings of 15%.
  • PSNR peak signal to noise ratio

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computing Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Description

一种视频预测编码方法和装置
技术领域
本发明涉及视频处理技术领域， 更具体地， 本发明涉及一种视频预 测编码方法和装置。 发明背景
目前， 网络应用逐渐向多媒体业务方向发展。 视频作为多媒体业务 的重要组成部分， 已经成为信息传播的主要载体之一。 视频的广泛应用 对视频的质量提出了新的要求， 现在各种视频的分辨率也越来越大。 同 时， 视频数据量的增长速度也远远超过存储设备和网络带宽的发展速 度， 因此视频压缩编码技术受到了长期广泛的关注。
在现有的视频压缩编码技术中， 一般将视频分为 I帧、 P帧和 B帧 进行编码。 I 帧是帧内编码帧， 在编码过程中只能利用帧内的信息进行 预测， 而解码时也只需要利用本帧的码流即可解码重构该帧图像。 I 帧 的编解码过程均是独立的， 与前后帧的信息无关。 P帧和 B帧在编码过 程中， 除了采用帧内预测技术， 还可以进行帧间预测， 帧间预测技术需 要在参考帧中搜索出与当前宏块最相似的块作为预测， 而在解码时， 根 据运动向量信息从已解码的参考帧中找到预测块实现解码。 P帧和 B帧 的编解码过程均不独立， 时间复杂度高， 但利用了帧间的时间相关性和 帧内的空间相关性， 能实现高效压缩。 与 P帧和 B帧的编码过程相比， I 帧只利用帧内的空间相关性进行预测， 编码过程独立， 复杂度较低， 但是 I帧的压缩效率远远不如 P帧和 B帧。
帧内预测目前主要应用在 I帧的编码过程中。 不过， 在 P帧或 B帧 中， 当帧间运动搜索技术难以搜索出准确的预测块时， 也会有少量的宏 块采用帧内预测。
目前， 在 MPEG-2标准中， 帧内预测采用筒单的空域预测算法。 编 码块的每个像素都采用像素值 128进行预测， 而不是利用编码帧本身的 信息进行预测。 对大多数序列而言， 这种方法预测残差较大， 因此采用 这种帧内预测方法的 I帧压缩效率^ ^氐。 在 MPEG-4标准中， 帧内编码 块采用了频域的 DC/AC预测算法， 编码块首先进行 DCT变换到频域， 利用相邻块的 DC/AC系数对当前块的系数进行预测， 利用 MPEG-4方 法， I帧的压缩有所提高。 在 H.264/AVC中， 采用了方向预测的帧内预 测算法。 该算法对于每个 16x16、 8x8或者 4x4的像素块， 利用其上边、 左边、 左上和右上的已编码块从一定的方向进行预测。 方向预测技术较 大幅度的提高了 I帧的压缩效率， 但是在这种以块为预测单元的预测过 程中只利用了块间的像素的相关性， 对于块内相邻像素之间的相关性， 仍然没有加以充分利用。 对于整个块而言， 块内离预测像素较远的像素 难以得到准确的预测， 因此帧内预测残差较大， 特别是对于具有复杂纹 理信息的块难以达到良好的预测效果， 从而直接导致了帧内编码的压缩 效率较低， 进而影响整个视频序列的压缩效率。
然而， 内容相同的视频图像， 分辨率越高， 像素之间的相关性也就 越大， 但现有的帧内编码技术并没有充分利用这种空间相关性。 针对这 个缺点， 研究能充分利用像素间相关性的预测算法， 提高视频编码的压 缩效率， 是具有重要意义的。 发明内容
本发明实施方式提供了一种视频预测编码方法， 以提高视频编码的 压缩效率。
本发明实施方式还提供一种视频预测编码装置， 以提高视频编码的 压缩效率。
本发明实施方式的技术方案如下：
一种视频预测编码方法， 该方法包括：
从当前帧中取出像素块，并将该像素块划分为主集像素和余集像素； 对所述主集像素进行编码， 输出主集像素码流， 并得到主集像素重 构值；
对包括所述主集像素重构值在内的已重构像素值进行插值以作为余 集像素的预测值， 再对所述余集像素进行帧内预测编码， 输出余集像素 码流， 并得到余集像素重构值；
对所述主集像素重构值和余集像素重构值进行组合， 得到所述像素 块的重构值。
一种视频预测编码装置， 该装置包括像素划分单元、 主集像素编码 单元、 余集像素编码单元和像素块重构值组合单元， 其中：
像素划分单元， 用于从当前帧中取出像素块， 并将该像素块划分为 主集像素和余集像素；
主集像素编码单元， 用于对所述主集像素进行编码， 输出主集像素 码流， 并得到主集像素重构值；
余集像素编码单元， 用于对包括所述主集像素重构值在内的已重构 像素值进行插值作为余集像素的预测值， 再对所述余集像素进行帧内预 测编码， 输出余集像素码流， 并得到余集像素重构值；
像素块重构值组合单元， 用于对所述主集像素重构值和余集像素重 构值按照与像素划分相对应的方式进行组合， 得到所述像素块的重构 值。
从上述技术方案可以看出， 在本发明实施方式中， 首先从当前帧中 取出像素块， 并将该像素块划分为主集像素和余集像素， 然后对主集像 素进行编码， 输出主集像素码流， 并得到主集像素重构值， 并对至少包 括主集像素重构值在内的已重构像素值进行插值， 对余集像素进行帧内 预测编码， 输出余集像素码流， 并得到余集像素重构值， 最后对主集像 素重构值和余集像素重构值进行组合，得到像素块的重构值。 由此可见， 本发明实施方式充分利用了帧内像素间的相关性， 提高了预测准确度， 减小了预测残差， 因而能够直接降低残差的码流开销，提高了压缩效率。
而且， 将本发明实施方式用于 H.264/AVC的 I帧编码， 采用水平隔 列抽样分解为偶宏块和奇宏块， 可以降低编码帧内预测模式标志信息的 码流开销， 从而能够进一步提高压缩效率。 对于奇宏块， 固定的预测模 式节约了预测模式标志信息的码流开销； 而对于偶宏块， 改进的帧内预 测模式预测编码方法使帧内预测模式标志信息的码流开销能得到较大 限度的节约。
不仅与此， 本发明实施方式具有较低的实现复杂度。 利用已编码偶 宏块的重构像素块对奇宏块进行预测编码时， 6抽头插值滤波器的计算 只利用了 19x16个像素点， 其中有 16x16个像素来自于当前 32x16块中 属于偶宏块的像素， 3x16个像素来自于左边的已重构邻居宏块， 编码过 程中不需要緩存其他行的宏块， 因而极大地节省了内存占用， 提高了高 速緩存的命中率。 尤其是对于数据量很大的高清视频源， 这种低内存占 用的方案易于芯片实现。 还有， 本发明实施方式非常适合高度并行计算 的实现， 从而能对高清视频源进行快速的编码压缩。 附图简要说明
图 1为根据本发明实施方式的视频预测编码方法流程图；
图 2为根据本发明实施方式的隔行隔列抽样示意图；
图 3为根据本发明实施方式的棋盘抽样示意图； 图 4为根据本发明实施方式的抽样单元为 2x1像素块的类棋盘抽样 示意图；
图 5为根据本发明实施方式的 I帧亮度分量编码流程图；
图 6为根据本发明实施方式的宏块分解示意图；
图 7为根据本发明实施方式的 4x4亮度块预测像素的选择示意图； 图 8为根据本发明实施方式的偶宏块插值作为奇宏块预测值的示意 图；
图 9为根据本发明实施方式的偶宏块帧内预测模式的预测示意图； 图 10 为根据本发明实施方式的视频预测编码装置总体框架结构示 意图；
图 11为根据本发明实施方式的视频预测编码装置结构示意图； 图 12为根据本发明实施方式的 BigShips序列的率失真曲线对比图。 实施本发明的方式
为使本发明实施方式的目的、 技术方案和优点更加清楚， 下面将结 合附图对本发明实施方式作进一步地详细描述。 的缺陷， 本发明实施方式提出一种基于相邻像素相关性的视频预测编码 方法， 以实现克服上述缺陷和提高编码效率的目的。
图 1为根据本发明实施方式的视频预测编码方法流程图。
如图 1所示， 该方法包括：
步骤 101 : 从当前帧中取出像素块， 并将该像素块划分为主集像素 和余集像素。
比如， 可以从当前编码帧中取出 M*N大小的像素块， 将这 M*N大 小的像素称为原块， 其中 M为像素块的宽， N为像素块的高。 这 M*N 大小的像素块既可以是亮度块， 还可以是色度块， 本发明实施方式对此 并无限定。 然后， 按某种取样规则抽取该 M*N大小的像素块内的部分 像素（设定这部分像素点的个数为 这部分像素称为主集像素； 余下 的像素记为余集像素。 也就是， 从 M*N大小的原块中抽样出 个像素 点组成主集像素， 并将其余像素组成余集像素。 在这里， M应该小于或 者等于当前帧的帧宽， N小于或等于当前帧的帧高， 而且 P<M*N (即 要求主集像素大小必须小于原块大小；)。
更具体地， 可以根据隔行隔列抽样方式或根据棋盘抽样方式（或者 类棋盘抽样方式）将该像素块划分为主集像素和余集像素。
下面对抽样方式进行详细说明。
隔行隔列抽样根据行抽样周期（记为 7>, 7>为非 0整数， Tr<N), 列抽样周期（记为 ；, 7；为非 0整数， TC<M, Tr - Tc≠1 ), 首行主集像 素在像素块中的行号（记为 r, r= l,..., Tr)和首列主集像素在像素块中 的列号 （记为 c, c= l,..., Tc) 的不同表现为不同的抽样方式。
图 2为根据本发明实施方式的隔行隔列抽样示意图，其中参数 c = 3, r = 4, Tc = 4, Tr=5。 灰色方块代表的像素组成主集像素， 若将主集像 素按初始位置顺序排列成 *«大小的块， 即主集像素的宽为 , 主集像 素的高为《, 则 、 n ) M、 N之间的关系如式（1)。
Figure imgf000008_0001
(类）棋盘抽样包括棋盘抽样和由棋盘抽样转化出的各种抽样方法。 棋盘抽样时， 每个被抽样出像素的上、 下、 左、 右相邻像素均未被抽样。 若被抽样出单元不再是一个像素点， 而是固定大小的像素块， 但仍然采 用棋盘抽样的规则， 称这类抽样为类棋盘抽样。 图 3为根据本发明实施 方式的棋盘抽样示意图； 图 4为根据本发明实施方式的抽样单元为 2x1 像素块的类棋盘抽样示意图。 如图 3所示， 灰色方块（或白色方块）代 表的像素被抽样出组成主集像素。
注意到， 主集像素的大小根据抽样方式的不同而不同， 主集像素根 据编码需要可以构成编码块， 余集像素由原块中除去主集像素后余下的 像素组成， 因此， 余集像素大小可以与主集像素相同或者不同， 余集像 素组成的编码块在形状、 数量上并非一定与主集像素一致。
步骤 102: 对主集像素进行编码， 输出主集像素码流， 并得到主集 像素重构值。
在这里， 主集像素的预测方法根据当前编码帧类型的不同而不同。 若当前帧为 I帧， 主集像素可以采用类似 H.264/AVC方向预测算法进行 预测；若当前帧为 P帧或 B帧，主集像素可以采用类似 H.264/AVC方向 预测算法或者帧间运动搜索技术进行预测。
具体包括， 首先判断当前帧的类型， 并当当前帧的类型为 I帧时， 采用类似 H.264/AVC方向预测算法对主集像素进行编码，如果当前帧的 类型为 P帧或 B帧时，可以采用类似 H.264/AVC方向预测算法或者帧间 运动搜索算法对所述主集像素进行编码。
更具体地， H.264/AVC 是目前压缩效率最高的视频编码标准， 在 H.264/AVC中， 帧内编码的块采用方向预测技术进行帧内预测。 方向预 测对不同尺寸的块采用不同的预测模式。对于亮度分量，块尺寸有 16x16 和 4x4两种， 其中 16x16亮度块（对应的宏块类型为 I16MB )有 4种预 测模式， 4x4亮度块（对应的宏块类型为 I4MB ), 有 9种预测模式； 对 于色度分量， 块尺寸为 8x8, 有 4种预测模式。 16x16亮度块（或 8x8 色度块）有 4种预测模式， 分别为水平预测、 竖直预测、 DC预测和平 面预测。 水平预测时， 每行像素的预测值相等， 预测值即为左边相邻已 重构块在该行的边界像素值； 竖直预测时， 每列像素的预测值相等， 预 测值即为上边相邻已重构块在该列的边界像素值； DC预测时， 整个块 采用相同的预测值， 预测值由该块上边相邻已重构块的 16个（或 8个） 边界像素、 左边相邻已重构块的 16个（或 8个）边界像素和左上角的 相邻像素取平均得到， 如果这些像素中有部分像素不存在（当前宏块位 于图像上边界或者左边界时）， 则由存在的像素取平均得出预测值； 平 面预测时， 用当前块左边和上边相邻已重构块的边界像素通过一个线性 函数实现对当前块中每个像素的预测。对于 4x4亮度块进行方向预测时， 对应的预测模式分别为竖直预测（模式编号 0 )、水平预测（模式编号 1 )、 对角线左下预测 （模式编号 3 )、 对角线右下预测 （模式编号 4 )、 竖直 向右预测 （模式编号 5 )、 水平向下预测 （模式编号 6 )、 竖直向左预测 (模式编号 7 )、 和水平向左预测（模式编号 8 ), 连同 DC预测（模式编 号 2 )—共 9中预测模式， 每种预测模式利用该模式对应方向上的已重 构像素对当前块的像素进行预测。
步骤 103: 对包括所述主集像素重构值在内的已重构像素值进行插 值得到余集像素的预测值， 再对所述余集像素进行帧内预测编码， 输出 余集像素码流， 并得到余集像素重构值。
在这里， 已重构像素值既可以就是主集像素重构值自身， 也可以包 括主集像素重构值以及除了该像素块之外的其它像素块的重构值。 比 如， 当前编码的像素块为第一个执行编码的像素块时， 则没有除了该像 素块之外的其它像素块的重构值， 此时已重构像素值就只包括主集像素 重构值， 则只对主集像素重构值进行插值得到余集像素的预测值， 再对 余集像素进行帧内预测编码， 输出余集像素码流， 并得到余集像素重构 值。 当存在其它像素块的已重构像素值时， 则共同利用包括主集像素重 构值以及其它像素块的已重构像素值在内的所有已重构像素值进行插 值得到余集像素的预测值， 再进行帧内预测编码， 输出余集像素码流， 并得到余集像素重构值。
优选的， 根据余集像素中各像素所在的位置对所述余集像素中各像 素进行分类，其中将相邻信息已知程度相同的像素放在同一类别； 然后， 对于各个分类的余集像素， 分别对包括主集像素重构值、 已重构的余集 像素重构值以及其它像素块的已重构像素值在内的所有已重构像素值 进行插值得到预测值， 再进行分层结构的帧内预测编码。
具体地， 余集像素利用所有已重构像素值特别是主集像素的重构值 进行预测。 根据余集像素的大小， 特别是当余集像素大小远大于主集像 素时（比如， 余集像素大小是主集像素的 10倍时）， 可以对余集像素再 分类， 并采用分层结构进行预测。 余集像素主要根据余集像素所在的位 置进行分类， 像素位置决定像素的相邻信息已知的程度， 将相邻信息已 知程度相同的余集像素放在同一类别。 同类别的像素采用统一的预测方 式， 不同类别的余集像素采用不同预测方式。 余集像素在采用分层结构 预测编码时， 底层预测 （即后编码像素的预测）可以利用高层预测 （即 先编码像素的预测）编码后的结果， 较底层的预测具有较高的准确性。 因此不同类别的像素在后续的量化过程中， 可以根据预测的准确程度选 择不同的量化精度， 对较高层的预测残差数据采用较高的量化精度。
在具体实施中， 余集像素的预测值主要采用插值滤波器对已重构像 素值进行插值得到。 对于分层结构预测， 可以根据每层余集像素的情况 设计多种滤波器。 滤波器设计可以有多种。 既可以直接复制已重构的相 邻像素值， 也可以利用 H.264/AVC的 6抽头滤波器， 还可以设计其他类 型的滤波器， 也可以用多种滤波器分别插值并从中并选择最佳的预测结 果。
步骤 104: 对主集像素重构值和余集像素重构值进行组合， 得到像 素块的重构值。
在这里， 将步骤 102得到的主集像素的重构值和步骤 103中得到的 余集像素的重构值重新组合得到原块的重构值， 组合方式为步骤 101中 取样方式的逆过程。
以上详细描述了本发明实施方式的编码实现流程。
下面以视频图像的 I帧亮度分量编码为实例说明本发明实施方式的 编码过程。
图 5为根据本发明实施方式的 I帧亮度分量编码流程图。 在这个实 例中， 以宽为 32, 高为 16的 32x16亮度块（此处的 32即对应于前述的 M, 16对应于前述的 N )为实例对本发明实现流程进行说明。 本领域技 术人员可以意识到， 以 32x16亮度块为实例进行说明仅是阐述性目的， 并不用于限定本发明的实施范围。
如图 5所示， 该方法包括：
步骤 501 : 在当前编码帧的位置取出 32x16亮度块（相当于水平方 向上连续的两个自然宏块）， 并对该 32x16 亮度块进行水平分解， 其中 由偶数列构成偶宏块， 奇数列构成与偶宏块对应的奇宏块。
更具体地， 将 32x16亮度块的偶数列 （编号为 2、 4、 6 16的 列）抽取出构成偶宏块，将 32x16亮度块的奇数列（编号为 1、 3、 5
15的列）抽取出构成奇宏块， 具体示意如图 6所示。
步骤 502: 判断当前编码位置是否到达该帧的右边界宏块， 是否只 剩一个单独的宏块而无法取出 32χ16的亮度块，如果是则执行步骤 503 , 否则执行步骤 504。
步骤 503:对单独宏块采用方向预测算法进行编码，并执行步骤 507。 具体地， 对于右边界剩余的单独的一个宏块， 采用方向预测算法进行预 测， 并对预测残差进行变换、 量化和熵编码以输出该块码流， 同时对量 化系数进行反量化反变换得到该块的重构块。
步骤 504: 对偶宏块采用方向预测算法进行预测， 并对预测残差进 行变换、 量化和熵编码以输出该块码流， 同时对量化系数进行反量化反 变换得到该块的重构块。
更具体地， 该偶宏块仍然要对 I16MB (对 16x16宏块整体做帧内预 测编码的宏块类型 ) 的 4种预测模式（即水平预测、 竖直预测、 DC预 测和平面预测）和 I4MB (将 16x16宏块分为 16个 4x4的块， 对每个块 分别做帧内预测的宏块类型） 的 9种预测模式（即竖直预测 （模式编号 0 )、 水平预测 （模式编号 1 )、 对角线左下预测 （模式编号 3 )、 对角线 右下预测 （模式编号 4 )、 竖直向右预测 （模式编号 5 )、 水平向下预测 (模式编号 6 )、 竖直向左预测 （模式编号 7 )、 水平向左预测 （模式编 号 8 )和 DC预测 （模式编号 2 ) )进行模式判决， 以编码计算选择最优 的预测模式。 当前偶宏块作为 16x16的块进行预测时， 预测方法和输入 像素的位置， 与现有技术中方向预测算法对 16x16亮度块进行预测的方 法是一致的。 若当前宏块分解为 16个 4x4块进行预测， 虽然预测模式 保持为 9种， 但用来预测的像素的位置有所改变。 如图 6所示， 灰色的 圓点表示组成偶宏块的像素点， 白色的圓点表示组成奇宏块的像素点， 长方形框中的灰色圓点组成当前编码的 4x4 块， 用来预测的像素点为 A-M, 其中像素 A-D来自于当前宏块上方的相邻偶宏块， 像素 I-M来自 于当前宏块左边的相邻偶宏块， 这与现有技术中方向预测算法对 4x4亮 度块进行预测时预测像素点的选择是一致的，但是像素 E-H位置有所改 变， 像素 E-H来自于当前宏块右上方最近的自然相邻块。
步骤 505: 利用偶宏块的重构像素对奇宏块进行预测， 这里利用 H.264/AVC的 6抽头滤波器对重构出的偶宏块进行插值， 得到奇宏块的 预测值。 对奇宏块的预测残差进行变换、 量化和熵编码以输出码流， 同 时对量化系数进行反量化反变换得到该块的重构块。
更具体地， 利用偶宏块的重构像素块对奇宏块进行预测。 当偶宏块 重构后， 即奇宏块左右的相邻像素已被重构， 在这里， 可以应用 6抽头 插值滤波器对偶宏块的重构像素值进行插值， 得到当前奇宏块的预测 值。 如图 7所示， 灰色的原点表示偶宏块中已经被重构出的像素， 白色 的原点表示当前奇宏块的待编码像素。
对于当前奇宏块待编码像素 X, 其预测值可以 X预表示。 可以用位 于像素 X左右的已重构像素 A、 B、 C、 D、 E和 F的重构值， 应用下式 ( 2 )进行插值作为当前奇宏块待编码像素 X的预测值 X ¾。
X 领 = ound( (A 重构一 5B 重构 +20G 重构 + 20D 重构一 5E 重构 + F 重构) 1 32 ) ( 2 );
其中 round为四舍五入取整函数；
C 重构为像素 X左边最接近的已重构像素 C的重构值；
B 重构为像素 X左边第二接近的已重构像素 B的重构值；
A 重构为像素 X左边第三接近的已重构像素 A的重构值；
D 重《为像素 X右边最接近的已重构像素 D的重构值；
E 重构为像素 X右边第二接近的已重构像素 E的重构值；
F 重构为像素 X右边第三接近的已重构像素 F的重构值。
在步骤 505中， 奇宏块的每个像素均由其相邻像素单独预测， 充分 利用了像素间的水平相关性， 预测结果较方向预测更加准确。 尤其是对 于具有复杂纹理的块， 这种预测方式的预测效果显著提高， 这也直接带 来编码性能的提高。 步骤 505的应用还会带来另一个优势， 即本发明实 施方式的帧内编码方法较 H.264/AVC 时间复杂度显著降低。 由于步骤 505的应用， 当前编码的 I帧中， 有接近一半或者一半的像素是用筒单 统一的 6抽头插值滤波器进行预测的，不需要进行高复杂度的模式判决， 这极大地降低了计算复杂度。
在现有技术中， 在 H.264/AVC视频编码标准中， 对于采用 I4MB类 型的编码的宏块， 需要在码流中记录每个 4x4块的预测模式。 为了节约 这部分码流开销，需要对这些预测模式进行预测。 由于步骤 505的采用， 奇宏块无论采用哪种宏块类型（ I4MB或 I16MB ), 只有固定的一种预测 模式， 因此不需要在码流中记录。 而对于偶宏块， 当采用 I4MB宏块类 型时， 仍然需要记录每个 4x4块的预测模式。 为了更准确对 I4MB类型 的偶宏块的预测模式进行预测， 用来预测的块的位置有所改变。 如图 9 所示， 传统的帧内编码利用块 U和块 Lo的预测模式对当前块 (编号 1) 的预测模式进行预测， 而步骤 505的使用使当前偶宏块的左边相邻宏块 变为采用固定预测模式的奇宏块， 因此这里采用位于当前偶宏块左边最 近的偶宏块中的块 Le和块 U的预测模式对当前块（编号 1 ) 的预测模 式进行预测， 同样属于这种情况的还有当前偶宏块中编号为 5、 9、 13 的块和其他偶宏块中处于左边界位置的块。
步骤 506: 将步骤 504得到的偶宏块的重构块和步骤 505中得到的 奇宏块的重构块进行组合得到原 32x16亮度块的重构块， 其中偶宏块的 每一列对应为 32x16块中的偶数列， 奇宏块的每一列对应为 32x16块中 的奇数列。
步骤 507: 判断当前帧是否编码完成， 若编码完成， 流程结束， 否 则跳转至步骤 501继续编码下一个亮度块。
基于上述详细描述， 本发明实施方式还提出了一种视频预测编码装 置。
图 10 为根据本发明实施方式的视频预测编码装置总体框架结构示 意图。
如图 10所示， 该装置包括像素划分单元 1001、 主集像素编码单元 1002、 余集像素编码单元 1003和像素块重构值组合单元 1004, 其中： 像素划分单元 1001 , 用于从当前帧中取出像素块， 并将该像素块划 分为主集像素和余集像素；
主集像素编码单元 1002, 用于对所述主集像素进行编码， 输出主集 像素码流， 并得到主集像素重构值；
余集像素编码单元 1003 ,用于对包括所述主集像素重构值在内的已 重构像素值进行插值作为余集像素的预测值， 再对所述余集像素进行帧 内预测编码， 输出余集像素码流， 并得到余集像素重构值；
像素块重构值组合单元 1004,用于对所述主集像素重构值和余集像 素重构值按照与像素划分相对应的方式进行组合 , 得到所述像素块的重 构值。
在一个实施方式中， 主集像素重构值获取单元 1002, 用于判断当前 帧的类型， 并当当前帧的类型为 I帧时， 采用 H.264/AVC方向预测算法 进行编码， 当当前帧的类型为 P帧或 B帧时，采用类似 H.264/AVC方向 预测算法或者帧间运动搜索算法进行编码。
在一个实施方式中， 余集像素重构值获取单元 1003 , 用于根据余集 像素中各像素所在的位置对余集像素中各像素进行分类， 其中将相邻信 息已知程度相同的像素放在同一类别， 并且对于各个分类的余集像素， 利用插值滤波器分别对包括主集像素重构值、 已重构的余集像素重构值 以及其它像素块的已重构像素值在内的所有已重构像素值进行插值得 到预测值， 再进行分层结构的帧内预测编码。
其中， 余集像素重构值获取单元 1003 包括的插值滤波器可以是 H.264/AVC的 6抽头滤波器， 也可以是其他滤波器。 当采用 6抽头滤波 器时， 6抽头插值滤波器， 可以用于对已重构像素进行插值作为各个分 类的余集像素预测值。 图 11为根据本发明实施方式的视频预测编码装置结构示意图。在图 11中， 该装置用于对一个 M*N像素块进行预测编码。
如图 11所示，该装置包括抽样单元，用于从当前编码帧中取出 M*N 大小的块， 并按照某种取样规则抽取该块内的部分像素组为主集； 余下 的像素记为余集。
根据当前帧的类型的不同， 主集像素可以采用适合帧类型的预测方 式进行预测， 并对预测残差进行变换、 量化和熵编码以输出主集像素码 流， 同时对量化系数进行反量化、 反变换， 再加上主集像素的预测值得 到主集像素的重构值。
对于余集像素， 先对包括主集像素重构值在内的已重构像素值进行 插值作为余集像素的预测值， 再对预测残差进行变换、 量化和熵编码以 输出余集像素码流， 同时对量化系数进行反量化、 反变换， 再加上余集 像素的预测值得到余集像素的重构值。
最后对所述主集像素重构值和余集像素重构值按照与像素划分相对 应的方式进行组合， 得到所述像素块的重构值。 如图 11 所示， 圓圏表 示对两个数据流相加， 加号是指这个数据流取正， 减号是指这个数据流 取负。
下面对图 11进行更加详细的说明。首先，抽样单元从当前编码帧中 取出 M*N大小的块， 并按照某种取样规则抽取该块内的部分像素组为 主集像素； 余下的像素记为余集像素。
针对主集像素， 根据当前帧的类型的不同， 采用适合帧类型的预测 方式进行预测。 在节点 A处， 从主集像素原始值中减去主集像素预测值 得到主集像素预测残差， 然后从节点 A向下依次运算， 对主集像素预测 残差进行变换、 量化和熵编码以输出主集像素码流。 其中， 节点 A向下 运算经过变换量化后的预测残差系数， 还需要依次经历反量化和反变 化， 得到一个差分值。 再在节点 B将这个差分值与主集像素的预测值相 加， 得到主集像素的重构值。
类似地， 针对余集像素， 首先对包括主集像素重构值在内的已重构 像素值进行插值作为余集像素的预测值，再在节点 D处从余集像素原始 值中减去余集像素预测值， 得到余集像素预测残差， 然后余集像素预测 残差再依次经过变换、 量化和熵编码后， 输出余集像素码流。 其中， 节 点 D向下运算经过变换量化后的预测残差系数，还需要依次经历反量化 和反变化， 得到一个差分值， 再在节点 C将差分值与余集像素的预测值 运算， 得到余集像素的重构值。
如图 5实例说明， 对于每一个 I帧应用本发明实施方式以后， 由于 有一半（或接近一半） 的像素采用了一种筒单的、 固定的、 有效的插值 滤波算法进行预测， 充分利用了帧内像素间的水平相关性， 提高预测准 确度， 减小预测残差， 因而直接降低了残差的码流开销， 提高了压缩效 率。
同时， 本发明实施方式降低了帧内预测模式标志信息的码流开销， 这也是压缩效率提高的另一个原因。 对于奇宏块， 固定的预测模式节约 了预测模式标志信息的码流开销； 而对于偶宏块， 改进的帧内预测模式 预测编码方法使帧内预测模式标志信息的码流开销能得到较大限度的 节约。
而且， 本发明实施方式实例中的帧内编码方案对奇宏块利用已编码 偶宏块的重构像素块，采用筒单统一的 6抽头插值滤波器进行预测编码， 这样， 在当前编码的 I帧中， 有接近一半或者一半的像素不需要进行高 复杂度的模式判决计算， 因此相对于 H.264/AVC的帧内预测算法， 计算 复杂度得到极大的降低。
另外， 本发明技术方案具有较低的实现复杂度。 利用已编码偶宏块 的重构像素块对奇宏块进行预测编码时， 6抽头插值滤波器的计算只利 用了 19x16个像素点， 其中有 16x16个像素来自于当前 32x16块中属于 偶宏块的像素， 3x16个像素来自于左边的已重构邻居宏块， 编码过程中 不需要緩存其他行的宏块， 因而极大地节省了内存占用， 提高高速緩存 的命中率。 尤其是对于数据量很大的高清视频源， 这种低内存占用的方 案易于芯片实现。
不仅与此， 本发明技术方案适合高度并行计算的实现， 从而能对高 清视频源进行快速的编码压缩。
将本发明实施方式的方法应用到 H.264/AVC标准的基本档次， 参考 软件采用 H.264/AVC标准的参考软件 JM15.0, 在分辨率为 720p的序列 上能获得平均为 0.39dB的 PSNR增益或 7.31%的码率节省。
图 12为根据本发明实施方式的 BigShips序列的率失真曲线对比图。 图 12给出了在分辨率为 720p帧率为 30fps的 BigShips序列上率失真曲 线对比。 其中： 图形上方的曲线为应用本发明实施方式在分辨率为 720p 帧率为 30fps 的 BigShips序列上率失真曲线； 图像下方的曲线为应用 H.264在分辨率为 720p帧率为 30fps的 BigShips序列上率失真曲线。 在 这个序列上， 本发明实施方式能获得 0.79dB的峰值信噪比（PSNR )增 益或 15%的码率节省。
以上所述仅为本发明的较佳实施方式而已， 并不用以限制本发明， 凡在本发明的精神和原则之内， 所做的任何修改、 等同替换、 改进等， 均应包含在本发明方式保护的范围之内。

Claims

权利要求书
1、 一种视频预测编码方法， 其特征在于， 该方法包括：
从当前帧中取出像素块，并将该像素块划分为主集像素和余集像素； 对所述主集像素进行编码， 输出主集像素码流， 并得到主集像素重 构值；
对包括所述主集像素重构值在内的已重构像素值进行插值以作为所 述余集像素的预测值， 再对所述余集像素进行帧内预测编码， 输出余集 像素码流， 并得到余集像素重构值；
对所述主集像素重构值和余集像素重构值进行组合， 得到所述像素 块的重构值。
2、根据权利要求 1所述的方法， 其特征在于， 所述将该像素块划分 为主集像素和余集像素包括： 根据隔行隔列抽样方式、 棋盘抽样方式或 类棋盘抽样方式将该像素块划分为主集像素和余集像素。
3、根据权利要求 1所述的方法， 其特征在于， 所述对主集像素进行 编码包括：
判断所述当前帧的类型， 并当所述当前帧的类型为 I 帧时， 采用 H.264/AVC方向预测算法对所述主集像素进行编码， 当所述当前帧的类 型为 P帧或 B帧时，采用类似 H.264/AVC方向预测算法或者帧间运动搜 索算法对所述主集像素进行编码。
4、根据权利要求 1所述的方法， 其特征在于， 所述已重构像素值为 所述主集像素重构值。
5、根据权利要求 1所述的方法， 其特征在于， 所述主集像素重构值 包括所述主集像素重构值以及除了该像素块之外的其它像素块的重构 值。
6、 根据权利要求 1-5中任一项所述的方法， 其特征在于， 所述对包 括主集像素重构值在内的已重构像素值进行插值作为所述余集像素的 预测值， 再对余集像素进行帧内预测编码包括：
根据所述余集像素中各像素所在的位置对所述余集像素中各像素进 行分类， 其中将相邻信息已知程度相同的像素放在同一类别；
对于各个分类的余集像素， 分别对包括主集像素重构值、 已重构的 余集像素重构值以及其它像素块的已重构像素值在内的所有已重构像 素值进行插值得到预测值， 再进行分层结构的帧内预测编码。
7、 根据权利要求 6所述的方法， 其特征在于， 对包括主集像素重 构值、 已重构的余集像素重构值以及其它像素块的已重构像素值在内 的所有已重构像素值进行插值得到预测值，再进行分层结构的帧内预 测编码包括：
利用 H.264/AVC的 6抽头滤波器， 对所有已重构像素进行插值作 为各个分类的余集像素预测值。
8、 根据权利要求 7所述的方法， 其特征在于， 所述主集像素为偶 宏块， 余集像素为奇宏块； 对于奇宏块的任意像素 X, 其预测值 Χ ¾为：
X 领 = ound( (A 重构一 5B 重构 +20G 重构 + 20D 重构一 5E 重构 + F 重构) I 32 ); 其中 round为四舍五入取整函数；
C 重《为像素 X左边最接近的已重构像素 C的重构值；
B 重构为像素 X左边第二接近的已重构像素 B的重构值；
A 重构为像素 X左边第三接近的已重构像素 A的重构值；
D 重《为像素 X右边最接近的已重构像素 D的重构值；
E 重构为像素 X右边第二接近的已重构像素 E的重构值；
F 重构为像素 X右边第三接近的已重构像素 F的重构值。
9、根据权利要求 1所述的方法， 其特征在于， 所述从当前帧中取出 像素块， 并将该像素块划分为主集像素和余集像素为： 根据隔行隔列抽 样方式将该像素块划分为主集像素和余集像素， 其中：
主集像素的宽为 , 主集像素的高为《, 且 、 《具有：
M-c N-r
m + l,n + 1
T T
Μ为该像素块的宽； N为该像素块的高；
7；为列抽样周期； ；为行抽样周期；7；为非 0整数， Tr N 7；为非
0 数， TC<M, Tr · Tc≠l;
r为首行主集像素在像素块中的行号， r= l,— , Tr; c为首列主集像 素在像素块中的列号， c= l,..., rc。
10、 根据权利要求 1-9中任一项所述的方法， 其特征在于， 所述像 素块为亮度块或色度块。
11、 一种视频预测编码装置， 其特征在于， 该装置包括像素划分单 元、 主集像素编码单元、 余集像素编码单元和像素块重构值组合单元， 其中：
像素划分单元， 用于从当前帧中取出像素块， 并将该像素块划分为 主集像素和余集像素；
主集像素编码单元， 用于对所述主集像素进行编码， 输出主集像素 码流， 并得到主集像素重构值；
余集像素编码单元， 用于对包括所述主集像素重构值在内的已重构 像素值进行插值作为所述余集像素的预测值， 再对所述余集像素进行帧 内预测编码， 输出余集像素码流， 并得到余集像素重构值；
像素块重构值组合单元， 用于对所述主集像素重构值和余集像素重 构值进行组合， 得到所述像素块的重构值。
12、 根据权利要求 11所述的视频预测编码装置， 其特征在于， 所述像素划分单元， 用于根据隔行隔列抽样方式或根据棋盘抽样方 式将该像素块划分为主集像素和余集像素。
13、 根据权利要求 11所述的视频预测编码装置， 其特征在于， 所述主集像素编码单元， 用于判断所述当前帧的类型， 并当所述当 前帧的类型为 I帧时， 采用 H.264/AVC方向预测算法进行编码， 当所述 当前帧的类型为 P帧或 B帧时，采用类似 H.264/AVC方向预测算法或者 帧间运动搜索算法进行编码。
14、 根据权利要求 11所述的视频预测编码装置， 其特征在于， 所述余集像素编码单元， 用于根据所述余集像素中各像素所在的位 置对所述余集像素中各像素进行分类， 其中将相邻信息已知程度相同的 像素放在同一类别， 并且对于各个分类的余集像素， 对包括主集像素重 构值、 已重构的余集像素重构值以及其它像素块的已重构像素值在内的 所有已重构像素值进行插值得到预测值， 再进行分层结构的帧内预测编 码。
15、 根据权利要求 11-14 中任一项所述的视频预测编码装置， 其特 征在于，所述余集像素重构值获取单元包括 H.264/AVC的 6抽头滤波器； 所述 H.264/AVC的 6抽头滤波器，对所有已重构像素进行插值作为 各个分类的余集像素预测值。
16、根据权利要求 15所述的视频预测编码装置， 其特征在于， 所述 主集像素为偶宏块， 余集像素为奇宏块； \
对于奇宏块的任意像素 X, 其预测值 Χ ¾为：
X ¾ — ound( (A 重构一 5B 重构 +20G 重构 + 20D 重构一 5E 重构 + F 重构) I 32 ); 其中 round为四舍五入取整函数；
C 重构为像素 X左边最接近的已重构像素 C的重构值； B重构为像素 X左边第二接近的已重构像素 B的重构值；
A重构为像素 X左边第三接近的已重构像素 A的重构值；
D重《为像素 X右边最接近的已重构像素 D的重构值；
E重构为像素 X右边第二接近的已重构像素 E的重构值；
F重构为像素 X右边第三接近的已重构像素 F的重构值。
17、根据权利要求 11所述的视频预测编码装置， 其特征在于， 所述 像素划分单元， 用于根据隔行隔列抽样方式将该像素块划分为主集像素 和余集像素， 其中：
主集像素的宽为 , 主集像素的高为《, 且 、 《具有：
Figure imgf000024_0001
其中函数 L」为向下取整函数；
M为该像素块的宽； N为该像素块的高；
7；为列抽样周期； ；为行抽样周期；7；为非 0整数， Tr<N; 7；为非
0整数， TC<M, Tr · Tc≠l;
r为首行主集像素在像素块中的行号， r=l，. ··, Tr; c为首列主集像 素在像素块中的列号， c=l,...,rc。
PCT/CN2011/071611 2010-03-12 2011-03-08 一种视频预测编码方法和装置 WO2011110088A1 (zh)

Priority Applications (6)

Application Number Priority Date Filing Date Title
MX2012010518A MX2012010518A (es) 2010-03-12 2011-03-08 Metodo y dispositivo para codificacion predictiva de video.
BR112012022951-6A BR112012022951B1 (pt) 2010-03-12 2011-03-08 Método e dispositivo para codificação preditiva de vídeo
SG2012065314A SG183888A1 (en) 2010-03-12 2011-03-08 Method and device for video predictive encoding
RU2012142523/08A RU2536366C2 (ru) 2010-03-12 2011-03-08 Способ и устройство кодирования видео с предсказанием
CA2797569A CA2797569C (en) 2010-03-12 2011-03-08 Method and device for video predictive encoding
US13/608,064 US8654845B2 (en) 2010-03-12 2012-09-10 Method and device for video predictive encoding

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2010101239505A CN101783957B (zh) 2010-03-12 2010-03-12 一种视频预测编码方法和装置
CN201010123950.5 2010-03-12

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US13/608,064 Continuation US8654845B2 (en) 2010-03-12 2012-09-10 Method and device for video predictive encoding

Publications (1)

Publication Number Publication Date
WO2011110088A1 true WO2011110088A1 (zh) 2011-09-15

Family

ID=42523743

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/071611 WO2011110088A1 (zh) 2010-03-12 2011-03-08 一种视频预测编码方法和装置

Country Status (8)

Country Link
US (1) US8654845B2 (zh)
CN (1) CN101783957B (zh)
BR (1) BR112012022951B1 (zh)
CA (1) CA2797569C (zh)
MX (1) MX2012010518A (zh)
RU (1) RU2536366C2 (zh)
SG (1) SG183888A1 (zh)
WO (1) WO2011110088A1 (zh)

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102792690B (zh) * 2010-03-09 2016-06-15 汤姆森特许公司 基于分类环路滤波器的方法和装置
CN101783957B (zh) * 2010-03-12 2012-04-18 清华大学 一种视频预测编码方法和装置
MX338309B (es) * 2010-09-03 2016-04-12 Sony Corp Dispositivo de codificacion, metodo de codificacion, dispositivo de descodificacion, y metodo de descodificacion.
GB2486726B (en) * 2010-12-23 2017-11-29 British Broadcasting Corp Compression of pictures
KR20120140181A (ko) 2011-06-20 2012-12-28 한국전자통신연구원 화면내 예측 블록 경계 필터링을 이용한 부호화/복호화 방법 및 그 장치
CN107105305B (zh) 2011-06-24 2020-04-03 Lg 电子株式会社 图像信息编码和解码方法
JP2013126182A (ja) * 2011-12-15 2013-06-24 Samsung Electronics Co Ltd 撮像装置及び画像処理方法
CN103248885B (zh) * 2012-02-14 2018-01-26 乐金电子(中国)研究开发中心有限公司 帧内图像预测编解码方法及视频编解码器
CN103260019B (zh) * 2012-02-16 2018-09-07 乐金电子(中国)研究开发中心有限公司 帧内图像预测编解码方法及视频编解码器
KR102349788B1 (ko) * 2015-01-13 2022-01-11 인텔렉추얼디스커버리 주식회사 영상의 부호화/복호화 방법 및 장치
CN104702962B (zh) * 2015-03-03 2019-04-16 华为技术有限公司 帧内编解码方法、编码器和解码器
CN106851288B (zh) * 2017-02-27 2020-09-15 北京奇艺世纪科技有限公司 一种帧内预测编码方法及装置
CN107277508B (zh) * 2017-07-25 2020-04-24 哈尔滨工业大学 一种采用自适应模式选择的像素级两向帧内预测方法
WO2019191891A1 (zh) * 2018-04-02 2019-10-10 北京大学 用于视频处理的方法和设备
CN108737836A (zh) * 2018-06-13 2018-11-02 北京奇艺世纪科技有限公司 一种帧间预测编码方法、装置及电子设备
CN109547795B (zh) * 2018-10-26 2021-02-12 上海九吾尊易信息科技有限公司 视频编码方法及装置
EP3886435A4 (en) * 2019-06-25 2022-01-26 Guangdong Oppo Mobile Telecommunications Corp., Ltd. MAPPING PROCESS, ENCODER, DECODER AND COMPUTER STORAGE MEDIA

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1929611A (zh) * 2005-09-06 2007-03-14 三星电子株式会社 用于视频帧内预测编码和解码的方法和装置
CN1929612A (zh) * 2005-09-06 2007-03-14 三星电子株式会社 用于视频内预测编码和解码的方法和装置
US20070206872A1 (en) * 2006-03-03 2007-09-06 Samsung Electronics Co., Ltd. Method of and apparatus for video intraprediction encoding/decoding
CN101783957A (zh) * 2010-03-12 2010-07-21 清华大学 一种视频预测编码方法和装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7336712B1 (en) * 1998-09-02 2008-02-26 Koninklijke Philips Electronics N.V. Video signal transmission
KR100626419B1 (ko) * 2001-01-03 2006-09-20 노키아 코포레이션 비디오 전송에서 비트 스트림들간의 교환
US6920175B2 (en) 2001-01-03 2005-07-19 Nokia Corporation Video coding architecture and methods for using same
GB0414421D0 (en) * 2004-06-28 2004-07-28 Nokia Corp Authenticating users
US8630346B2 (en) * 2007-02-20 2014-01-14 Samsung Electronics Co., Ltd System and method for introducing virtual zero motion vector candidates in areas of a video sequence involving overlays
US8611435B2 (en) * 2008-12-22 2013-12-17 Qualcomm, Incorporated Combined scheme for interpolation filtering, in-loop filtering and post-loop filtering in video coding
US20100226437A1 (en) * 2009-03-06 2010-09-09 Sony Corporation, A Japanese Corporation Reduced-resolution decoding of avc bit streams for transcoding or display at lower resolution
JP5169978B2 (ja) * 2009-04-24 2013-03-27 ソニー株式会社 画像処理装置および方法
US20120183041A1 (en) * 2011-01-14 2012-07-19 Sony Corporation Interpolation filter for intra prediction of hevc

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1929611A (zh) * 2005-09-06 2007-03-14 三星电子株式会社 用于视频帧内预测编码和解码的方法和装置
CN1929612A (zh) * 2005-09-06 2007-03-14 三星电子株式会社 用于视频内预测编码和解码的方法和装置
US20070206872A1 (en) * 2006-03-03 2007-09-06 Samsung Electronics Co., Ltd. Method of and apparatus for video intraprediction encoding/decoding
CN101783957A (zh) * 2010-03-12 2010-07-21 清华大学 一种视频预测编码方法和装置

Also Published As

Publication number Publication date
CA2797569C (en) 2016-01-19
CN101783957B (zh) 2012-04-18
RU2536366C2 (ru) 2014-12-20
RU2012142523A (ru) 2014-05-10
US20130051468A1 (en) 2013-02-28
MX2012010518A (es) 2013-01-18
US8654845B2 (en) 2014-02-18
CA2797569A1 (en) 2011-09-15
CN101783957A (zh) 2010-07-21
BR112012022951B1 (pt) 2022-02-22
BR112012022951A2 (pt) 2017-03-21
SG183888A1 (en) 2012-10-30

Similar Documents

Publication Publication Date Title
WO2011110088A1 (zh) 一种视频预测编码方法和装置
CN101252686B (zh) 基于交织预测的视频帧内无损编解码方法及系统
JP5048146B2 (ja) Ａｃ予測状態情報のためのビットプレーンコーディングおよびデコーディング
EP1457056B1 (en) Skip macroblock coding
JP3118237B1 (ja) 画像予測復号化方法
US9571861B2 (en) Decoder, encoder, method for decoding and encoding, and data stream for a picture, using separate filters
RU2404537C2 (ru) Устройство для кодирования динамических изображений, устройство для декодирования динамических изображений, способ кодирования динамических изображений и способ декодирования динамических изображений
JP5401009B2 (ja) 映像のイントラ予測符号化、復号化方法及び装置
KR101624649B1 (ko) 계층적인 부호화 블록 패턴 정보를 이용한 비디오 부호화 방법 및 장치, 비디오 복호화 방법 및 장치
KR100739714B1 (ko) 인트라 예측 모드 결정 방법 및 장치
TWI711297B (zh) 利用量化係數分量與圖框間預測資訊解碼視訊資料的方法
KR101701342B1 (ko) 적응적인 루프 필터링을 이용한 비디오의 부호화 방법 및 장치, 비디오 복호화 방법 및 장치
KR20110018189A (ko) 영상의 부호화 방법 및 장치, 그 복호화 방법 및 장치
CN101247525B (zh) 一种提高图像帧内编码速率的方法
KR20050045746A (ko) 계층 구조의 가변 블록 크기를 이용한 움직임 추정 방법및 장치
WO2007055158A1 (ja) 動画像符号化方法、動画像復号化方法および装置
WO2009052697A1 (en) A dual prediction video encoding and decoding method and a device
KR20110065089A (ko) 영상의 부호화 방법 및 장치, 그 복호화 방법 및 장치
US20070133689A1 (en) Low-cost motion estimation apparatus and method thereof
CN103442228A (zh) 从h.264/avc标准到hevc标准的快速帧内转码方法及其转码器
KR100984650B1 (ko) Ｈ．２６４ 및 다른 변환 코딩된 정보에 대한 매우효율적인 부분 디코딩을 가능하게 하는 비디오 인코딩 방법
EP1841235A1 (en) Video compression by adaptive 2D transformation in spatial and temporal direction
CN110913232B (zh) 一种tu划分模式的选择方法及装置、可读存储介质
Wang et al. Overview of the second generation avs video coding standard (avs2)
CN100574443C (zh) 一种确定增强层帧内预测模式的方法和编解码设备

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: 11752831

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2797569

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: MX/A/2012/010518

Country of ref document: MX

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 1201004667

Country of ref document: TH

Ref document number: 7878/CHENP/2012

Country of ref document: IN

WWE Wipo information: entry into national phase

Ref document number: 2012142523

Country of ref document: RU

122 Ep: pct application non-entry in european phase

Ref document number: 11752831

Country of ref document: EP

Kind code of ref document: A1

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112012022951

Country of ref document: BR

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 15-11-2013)

122 Ep: pct application non-entry in european phase

Ref document number: 11752831

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 112012022951

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20120911