WO2012065448A1

WO2012065448A1 - Intra frame prediction method for video encoding

Info

Publication number: WO2012065448A1
Application number: PCT/CN2011/076354
Authority: WO
Inventors: 舒倩
Original assignee: 深圳市融创天下科技股份有限公司
Priority date: 2010-11-15
Filing date: 2011-06-25
Publication date: 2012-05-24
Also published as: CN102469308A; CN102469308B

Abstract

The present invention discloses an intra frame prediction method for video encoding, and the method includes the following steps: with regard to luminance prediction, adopting different luminance prediction modes according to picture features, relevancy between encoding pixels and adjacent encoded pixels, and next adjacent encoded pixel information; with regard to chrominance prediction, adopting sub-block based chrominance intra frame prediction, which uses different pixels to predict different encoding sub-blocks according to relevancy between pixels in a current sub-block and adjacent encoded pixels. By adopting the intra frame prediction method in the present invention, prediction accuracy and encoding effect can be improved. The intra frame prediction method of the present invention is an integrated intra frame prediction system, whereas each of the prediction modes can also be utilized independently.

Description

Intra prediction method for video coding

Technical field

The present invention relates to the field of video coding, and in particular, to a video coding intra prediction method.

Background technique

In video coding, intra prediction is typically used to eliminate spatial redundancy of the image, and interframe prediction is used to eliminate temporal redundancy. The intra-predicted frame is also used as the reference frame for inter-frame prediction, so its coding performance is particularly important. Based on the YUV420 format video source, conventional intra prediction is divided into luminance component prediction and chrominance component prediction. Since the video source has different image features, there are flat textures. Simple areas also have texture details. Based on this actual situation, the intra prediction of conventional luminance is also divided into 16x16 luma prediction suitable for relatively flat regions and 4x4 luma prediction mode suitable for texture information. At the same time, since the human eye is sensitive to luminance information, it has relatively weak sensitivity to chrominance, and chromaticity prediction design is generally simple. At present, the conventional spatial domain intra prediction method only uses the nearest neighbor pixel encoded in the current block for prediction, and performs intra prediction. On the one hand, it only considers the correlation between the pixel of the current coding block and its encoded neighboring pixels, and ignores The correlation between the predicted pixel and the encoded neighboring pixel will decrease as the distance is further away. On the other hand, it does not consider the image features, which makes it impossible to eliminate the spatial redundancy when the video source has extremely similar regions, thus restricting the optimization of the compression performance.

Summary of the invention

An object of the present invention is to provide a video coding intra prediction method, which aims to solve the problem that the correlation between the coded pixel and the adjacent coded pixel is reduced as the distance is far away. The characteristics, and the image features are not considered, so that when the video source has extremely similar regions, the spatial redundancy cannot be eliminated to the maximum extent, which restricts the optimization implementation of the compression performance.

The method of the embodiment of the present invention is implemented by the video coding intra prediction method, where the method includes:

Different brightness predictions are used according to image features, degree of association of coded pixels with adjacent coded pixels, and secondary neighbor coded pixel information;

Sub-block-based chroma intra prediction is used to perform chromaticity prediction using different pixels for different coded sub-blocks according to the degree of association between the pixels of the current coded sub-block and the adjacent coded pixels. The prediction of the 16x16 luma block is specifically: using the frame ftH_Skip, V_Skip mode, combining the image feature and the degree of association between the coded pixel and the adjacent coded pixel, using the first fade mode and the first progressive DC mode prediction.

The prediction of the 4×4 luma block is specifically performed according to the degree of association between the encoded pixel and the encoded neighboring pixel and the encoded sub-neighboring pixel information, wherein different 4×4 brightness is adopted according to different prediction directions. Block prediction mode.

The chrominance information of the current coded macroblock is evenly divided into four sub-blocks, and different pixels are used for prediction according to the degree of association between the pixels of the current coded sub-block and the adjacent coded pixels, including the following prediction modes: Sub-block based progressive DC mode, second fade mode, sub-block based upper mode, and sub-block based left mode.

Advantageous Effects of Invention: Embodiments of the present invention provide a new intra prediction method that combines image features, the degree of association of coded pixels with adjacent coded pixels, and utilizes secondary neighbor coded pixel information. Among them, the intra 11_8^ and V_Skip modes combine image characteristics and extend the Skip mode to intra prediction. The advantage is obvious in video sources with more similar image regions. In other brightness prediction modes, the degree of correlation between the coded pixels and the adjacent coded pixels is considered, and the secondary adjacent coded pixel information is utilized to make the prediction more accurate.

A sub-block based chrominance prediction mode is proposed for chrominance prediction. The method takes into account: On the one hand, the human eye is relatively insensitive to chrominance information, and on the other hand, due to the currently used YUV420 format coding source, the sampling point spacing is large in chrominance, so it is necessary to update the chrominance prediction in time. The predicted pixel value. In combination with the above features, the embodiment of the present invention proposes a sub-block based chroma intra prediction mode, which is first. First, the current coded chroma block is evenly divided into 4 sub-blocks; considering that the correlation between the pixels of the current coded sub-block and the adjacent coded pixels is further away from the distance, the correlation is also reduced, and the different coded sub-blocks are Use different prediction pixels for prediction to improve coding. The intra prediction in the embodiment of the present invention is both a complete intra prediction system, and each of the prediction modes can be used independently.

DRAWINGS

no.

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. For the convenience of description, only the parts related to the embodiments of the present invention are shown. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The embodiment of the present invention proposes a new intra prediction method, which combines image features, the degree of association between coded pixels and adjacent coded pixels, and utilizes secondary neighbor coded pixel information. Among them, the intra-frame H_Skip and V_Skip modes combine image characteristics and extend the Skip mode to intra prediction. The advantage is obvious in video sources with more similar image regions. In other brightness prediction modes, the degree of correlation between the coded pixels and the adjacent coded pixels is considered, and the secondary adjacent coded pixel information is utilized, so that the prediction is more accurate.

A sub-block based chrominance prediction mode is proposed for chrominance prediction. The method takes into account: On the one hand, the human eye is relatively insensitive to chrominance information, and on the other hand, due to the currently used YUV420 format coding source, the sampling point spacing is large in chrominance, so it is necessary to update the chrominance prediction in time. The predicted pixel value. In combination with the above features, the embodiment of the present invention proposes a sub-block-based chroma intra prediction mode, which firstly divides the current coded chroma block into 4 sub-blocks uniformly; considering the pixel of the current coded sub-block and the adjacent coded pixel The correlation degree decreases with the distance away from the distance, and different prediction pixels are used for prediction of different coding sub-blocks, thereby improving the coding effect. The intra prediction in the embodiment of the present invention is both a complete intra prediction system, and each of the prediction modes can be used independently. The embodiment of the invention provides a video coding intra prediction method, which comprises two parts: luminance prediction and chrominance prediction. For the luminance component, there are two types of prediction: 16x16 luma block and 4x4 luma block. In general, 4x4 luma blocks are used for prediction of macroblocks with more spatial detail information, and 16x16 luma blocks are used for flatter regions. Prediction; Chroma prediction is independent of luma prediction, and the two chroma components U, V use the same prediction mode, so an 8x8 chroma block can be used as the basic unit of intra prediction encoding.

The specific implementation methods of luminance prediction (16x16 luma block, 4x4 luma block) and chroma prediction (8x8 chroma block) are further described below.

Sl, brightness prediction

S11. The prediction of the 16x16 luma block (the corresponding luma block in the macroblock) is specifically as follows: combining the image features, adopting the frame ftH_Skip, V_Skip mode, and approximating the macroblock for the image feature; combining the image feature with the encoded pixel and the adjacent encoded pixel The degree of association, using the first gradient mode and the first progressive DC mode for prediction;

The compression ratio of intraframe coding is much lower than the compression ratio of interframe coding. On the one hand, the intraframe correlation is lower than the interframe correlation, and the other is because the redundancy of information still existing in the frame is not sufficient. compression. To describe the 16x16 luma intra prediction, the luminance information of the currently encoded macroblock MB and the predicted pixels required for intra prediction are labeled as follows:

Where MB is the current coded macroblock,

A0~A15 are: the pixel value of the first row of the upper side of the current coded macroblock, that is, the nearest neighbor pixel value of the currently coded macroblock;

B0~B15 are: the pixel value of the first column on the left side of the current coded macroblock, that is, the nearest neighbor pixel value of the currently coded macroblock;

C0~C15 are: the pixel value of the second row of the upper side of the current coded macroblock, that is, the next adjacent pixel value of the currently coded macroblock;

D0~D15 are: the pixel value of the second column on the left side of the current coded macroblock, that is, the pixel value of the next adjacent column of the currently coded macroblock;

X is: the pixel value of the upper left corner of the current coded macroblock;

S111, intraframe H_Skip, V_Skip mode:

Intra-frame H_Skip, V_Skip mode: The pixels of the current coded macroblock are directly represented by the left and upper coded macroblock information of the current coded macroblock.

The specific implementation is as follows:

If (left subblock is encoded) Pr W ] = Mb _left [i][j]

If (upper subblock is encoded) Pr ί3⁄43⁄4·][ | = Mb _up [i] [j]

Pr ed[i] [j]: the predicted value of the current coded pixel, where the pixel is located at the current macroblock;

Mb _left : the left-coded macroblock of the currently encoded macroblock;

b _up : the upper coded macroblock of the current coded macroblock;

The 16x16 luma prediction mode is generally applicable to flat regions, and there are often macroblocks with extreme approximation in the horizontal and vertical directions in the image sequence. This similarity allows us to further compress the similarity in the frame, and at a low bit rate. Next, or from the sensitivity of human vision is feasible. In combination with the above characteristics, the embodiment of the present invention proposes an intra-sky Skip mode in the horizontal and vertical directions, that is, an intraframe. H_Skip, V_Skip mode, this mode can further compress the redundancy of the intra-frame information without affecting the visual effect and the overall code rate is insufficiently allocated.

5112, first gradient mode:

First Gradient Mode: The pixels of the current coded macroblock are represented by a linear function of the pixel mean of the current coded macroblock, among the coded macroblocks on the upper and left sides.

The specific implementation is as follows:

If (upper and left subblocks are encoded)

Mean _ U = mean(A0, Al, A2,..., A15)

Mean _L = mean(fi0, Bl, B2,..., B15)

Pr ed[i][j] = ((16 - j) * mean _ [/ + (16 _ i) * mean _ L) / (32 -i-j) mean: find the mean

Pved[i][j]: the predicted value of the current encoded pixel,

i : the pixel is in the row of the current macroblock;

j : the pixel is in the column of the current macroblock;

In the extreme approximation region, the above mode can be used for intra prediction, and for the macroblock having the gradation feature, the gradation mode of the embodiment of the present invention can be used for compression of redundant information.

5113, first progressive DC mode:

First progressive DC mode: The pixels of the current coded macroblock are represented by the mean of the two-row and two-column pixel variations in the upper and left coded macroblocks closest to the current coded macroblock.

The specific implementation is as follows:

Sum _ du = s draw (2 *A0-C0,2*A1- l,2 *A2-C2,...,2*A15-C15)

Sum _dl = swm(2 * BO - DO, 2 *B1- Dl,2 *B2- D2,...,2 *B15-D15)

If (upper and left subblocks are encoded) Pre<i[ ] [knife = mean(sum _ du, sum _ dl) else if (only the upper sub-block is encoded) Pr ed[i][j] = mean(sum_du) else if (the left sub-block has coding)

Pr ed[i][j] = mean(sum_dl) else

Pr^[ ][;] = 128

Sum: summation

Pved[i][j]: the predicted value of the current encoded pixel, i : the pixel is in the row of the current macroblock; j : the pixel is in the column of the current macroblock; mean: seeking the mean 〇

S12. For the prediction of the 4x4 luma block, the correlation degree between the coded pixel and the encoded neighboring pixel and the encoded sub-neighbor pixel information are respectively considered for prediction, and nine modes are proposed according to different prediction directions.

When the texture is more complicated, it is necessary to adopt intra prediction in the form of a sub-block. Considering that the correlation between pixels is closely related to its position, in the embodiment of the present invention, in the luminance 4x4 and chrominance prediction, in order to describe the luminance frame of 4x4. For intra prediction, the luminance information of the current coded 4x4 sub-block and the predicted pixels required for intra prediction are labeled as follows:

X2 XI A1 B1 C1 D1 E1 F1 G1 HI

X A B C D E F G H

11 I

J1 J current encoding

K1 K 4x4 sub-block

L1 L ABCDEFGH is: the first row of the pixel value of the current coded macroblock, that is, the nearest neighbor row pixel value of the current coded macroblock;

A1B1 CI Dl El Fl Gl HI is: the pixel value of the second row of the upper side of the current coded macroblock, that is, the pixel value of the next adjacent row of the currently coded macroblock;

IJKL is: the pixel value of the first column on the left side of the current coded macroblock, that is, the nearest neighbor pixel value of the currently coded macroblock;

I1J1K1L1 is: the pixel value of the second column on the left side of the current coded macroblock, that is, the pixel value of the next adjacent column of the currently coded macroblock;

X is: the pixel value of the upper left corner of the current coded macroblock;

XI is: X upper side pixel value;

X2 is: XI left pixel value.

S121, second progressive DC mode

The second progressive DC mode: that is, the pixel of the current coded sub-block is represented by the mean value of the pixel change of each of the two rows and two columns of the current coded macroblock in the upper and left coded sub-blocks. The specific implementation is as follows:

Sum _ du4 = s m(A *2-Al, B* 2-B1, C*2-C1, D*2-D1)

s m_dl4 = s m(I*2-n, J*2-Jl, K*2-Kl, L*2-Ll) if (upper and left sub-blocks are encoded)

Ρτβύ?[ ][ '] = mean(sum _ du4, sum _ dl4) else if (only the upper sub-block is encoded)

P^ed[i][j] = mean(sum _du4) else if (the left subblock is encoded)

P^ed[i][j] = mean(sum _ dl4) Else

Pr ed[i] [j] = 12S

Sum: Qiukou

P ed[i] [j] : the predicted value of the current encoded pixel,

: the pixel is in the row of the current macroblock;

j : the pixel is in the column of the current macroblock;

Mean: Find the mean.

S122, weighted Up mode

Weighted Up mode: The mode predicts the correlation degree information of the pixel of the current coded sub-block and its encoded neighboring pixel and the coded secondary neighboring pixel, that is, the pixel of the current coded sub-block uses the nearest neighbor and the next of the coded sub-block of the upper side. A linear function of adjacent coded pixels is represented.

The specific implementation is as follows:

If (the upper sub-block is encoded)

Pr ecl[i] [j]: the predicted value of the current coded pixel,

j : The pixel is in the column of the current macroblock.

S123, weighted Left mode

Weighted Left mode: This mode considers the pixel of the current encoded sub-block with its encoded neighboring pixels, The correlation degree information of the coded secondary neighboring pixels is predicted, that is, the pixels of the current coded sub-block are represented by a linear function of the nearest neighboring and next neighboring coded pixels in the left coded sub-block.

The specific implementation is as follows:

If (the left subblock is encoded)

Pred[i][j]: the predicted value of the current encoded pixel,

i : The pixel is in the row of the current macroblock.

S124, the first weighting mode

First Weighting Mode: This mode predicts the degree of correlation of the pixels of the current coded sub-block with its encoded neighboring pixels, and also reduces the correlation as the distance is further removed. In view of this, the first weighting mode, that is, the pixel of the current coded sub-block is represented by a linear function of the nearest neighbored coded pixel in the upper and upper right side coded sub-blocks, after the coded pixel is farther away from the coded pixel, The nearest encoded coded pixel of the current coded sub-block is replaced with the predicted value of the coded pixel of the current coded sub-block. The specific implementation is as follows:

The upper and upper right subblocks are encoded)

Pved [0] [0] = (A + C + 2*B + 2) / 4;

Pved [0] [1] = (B + D + 2*C + 2) / 4;

Pved [0] [2] = (C + E + 2*D + 2) / 4; ■ / (ζ + [6]dui9 +[8]du + [e] [z P ^a - ^J d) = [ε][ε] I (z +[8]*ι + ([e] [zi

[Z] [ε] P ⁹ 3⁄4

■ / (Z + [S] dais; +[ άm^ _ή :z + [£] [ΐ] P ⁹ 3⁄4)= [£] [Z] P ^9J d

■ / {z + ]du + [e] [T] Ρ ³ 1⁄4^Ζ + [z] [τ] P ⁹ 3⁄4)= [Z] [Z] P ^9J d ^: /(ζ+[ε] [ΐ ] [ΐ] ) +[ΐ] [ΐ] P ^3j d)= [ΐ] [z] p^d

■ /(Z+[Z [ΐ] P ^dJ d +([ΐ] [ΐ] ^3⁄4)*2+[0] [ΐ] Ρ^ά)= [0] [2] P^ ■ / (Z + [Tjdais +[0] α *2 + [G] [0] P ⁹ 3⁄4)= [C] [T] P ^9J d

■ I {z +[ο]*κ + [e] [o] P ³ ^Z + [z [o] P ⁹ 3⁄4)= [Z] [T] p^d

■ / (Z+ [Z]

[6] du

: / (Z+ [9] ώι +^ [g] du + [ α ) ₌ [§] α

■ / (Z+ [ε]

[g] drai : / (Z+ [Z] [ΐ] du + [o] dd ) = [ ] drai

: /( +Η*ε+.) = [ ]*κ

+3) = [0] drai

■ /(Z+[ [0] P ^dJ d +([2] [0] P^d)^Z+[l [0] Ρ^ά)= [ΐ] [ΐ] P^d ■ /(Z+[Z ] [0] P ^dJ d + ([T] [0]

■ / (z + + ή + a) = [ε] [o] ^dj d Pred[i][j]: the predicted value of the current encoded pixel,

Temp[]: is the intermediate variable of the current mode, which is used to replace the nearest neighbor of the current coded sub-block with the predicted value of the coded pixel of the current sub-block directly after the distance between the coded pixel and the coded pixel is changed. A clear description of the function of the pixel, and the middle of the setting is i: the pixel is located in the row of the current macroblock;

j : The pixel is in the column of the current macroblock.

S125, second weighting mode

Second Weighting Mode: This mode considers the correlation degree information of the pixels of the current coded sub-block and its encoded neighboring pixels, and also reduces the correlation as the distance is farther away. In view of this, the second weighting mode, that is, the pixel of the current coding sub-block is represented by a linear function of the nearest adjacent coded pixel in the upper and upper right encoded sub-blocks, after the distance between the coded pixel and the coded pixel is changed, The nearest encoded coded pixel of the current coded sub-block is replaced with the predicted value of the coded pixel of the current coded sub-block. The specific implementation is as follows:

Side subblocks are encoded)

Pred [0] [0] = (A + B + 1) / ¹ 2;

Pred [0] [1] = (B + C + 1) / ¹ 2;

Fred [0] [2] = (C + D + 1) / ¹ 2;

Pred [0] [3] = (D + E + 1) / ¹ 2;

Pred [1] [0] = (Pred [0] [0] + PreJ [0] [1] + 1)

Pred [1] [1] = (Pred [0] [1] + PreJ [0] [2] + 1)

Pred [1] [2] = (Pred [0] [2] + Prej ] [0] [3] + 1) Fred [1] [3] = (D + 2*E + F + 2) / 4;

Fred [2] [0] = (Pved [1] [0] + Pved [1] [1] + l) / 2;

Fred [2] [1] = (Pved [1] [1] + Pved [1] [2] + l) / 2;

PreJ [2] [2] = (Pved [1] [2] + PreJ [1] [3] + l) / 2;

PreJ [2] [3] = (Fred [1] [3]*4 +D + 2*E + F+ 4) /8;

Fred [3] [0] = (Pved [2] [0] + Pved] [2] [1] + l) / 2;

PreJ [3] [1] = (PreJ [2] [1] + Pved [2] [2] + 1) / 2;

Fred [3] [2] = (PreJ [2] [2] + Pved [2] [3] + l) / 2;

PreJ [3] [3] = (Pred*8 + (G+F ) *3 + E+ H+ 8) /16;

Pr _e J[ ][7]: the predicted value of the current coded pixel,

Temp port: the intermediate variable of the current mode;

i : the pixel is in the row of the current macroblock;

j : The pixel is in the column of the current macroblock.

S126, third weighting mode

Third Weighting Mode: This mode predicts the degree of correlation of the pixels of the current coded sub-block with its encoded neighboring pixels, and also reduces the correlation as the distance is further removed. In view of this, the third weighting mode, that is, the pixel of the current coded sub-block is represented by a linear function of the nearest neighbored coded pixel in the left coded sub-block, and after the distance between the coded pixel and the coded pixel is changed, the current The predicted value of the encoded pixel of the encoded sub-block replaces the nearest encoded encoded pixel of the current encoded sub-block. The specific implementation is as follows:

Side subblocks are encoded)

Fred [0] [0] = (I + J + 1) / 2;

Fred [0] [1] = (I + 2*J + K + 2) / 4; Pred [1] [0] = (J + K + 1) / 2;

Pred [0] [2] = (J + K +Pred [1] [0]*2+2) / 4;

Pred [1] [1] = (J + 2*K + L + 2) / 4;

Pred [2] [1] = (K + 2*L + L + 2) / 4;

Pred [0] [3] = (Pr -cd [0] [1] + 2*PreJ [1] [1] + Pred [2] [1] + 2) / '

Pred [2] [0] = (K + L + 1) / 2;

Pred [1] [2] = (Pr ■ed [1] [1] + PreJ [2] [1] + l) / 2;

Pred [3] [1] = L;

Pred [1] [3] = (Pi 'ed [1] [1] + 2*PreJ [2] [1] + Vved [3] [1] + 2) /

Pred [3] [0] = Pred [2] [2] = Pred [2] [3] =Pred [3] [2] =Pred [3] [3] =

Pred[i][j]: the predicted value of the current encoded pixel,

i : the pixel is in the row of the current macroblock;

j : The pixel is in the column of the current macroblock.

S127, fourth weighting mode

Fourth Weighting Mode: This mode predicts the correlation degree information of the pixels of the current coded sub-block and its encoded adjacent and neighboring pixels. In view of this, the fourth weighting mode, i.e., the pixel of the current coded sub-block, is represented by a linear function of the nearest neighboring, next neighboring coded pixels in the upper, left, and upper left encoded sub-blocks. The specific implementation is as follows:

The upper and left sub-blocks are encoded)

Pred [3] [0] = (J1+ + 2*K + 2) / 3;

Pred [2] [0] = Pved [3] [1] = (11 + 2*J + 2) / 3;

Pred [1] [0] = Pved [2] [1] = Pred [3] [2] = (X2+ 2*1 + 2) / 3;

Pred [0] [0] Two Pred [1] [1] = Pred [2] [2] = Pred [3] [3] = (I + 2*X + A + 2) / 4;

Fred [0] [1] = Pved [1] [2] = Fred [2] [3] = (X1+ 2*A + 2) / 3; Fred [0] [2] =Pred [1] [3] = (Al + 2*B + 2) / 3;

Fred [0] [3] = (Bl + 2*C + 2) / 3;

Pved[i][j]: the predicted value of the current encoded pixel,

i : the pixel is in the row of the current macroblock;

j : The pixel is in the column of the current macroblock.

S128, fifth weighting mode

Fifth weighting mode: This mode predicts the correlation degree information of the pixels of the current coded sub-block and its encoded adjacent and neighboring pixels. In view of this, the fifth weighting mode, i.e., the pixel of the current coded sub-block, is represented by a linear function of the nearest neighboring, next neighboring coded pixels in the upper, left, and upper left encoded sub-blocks. The specific implementation is as follows:

If (upper and left subblocks are encoded)

Pred [0] [0] = Pred [2] [1] = (X + A*2 +X1+ 2) / 4;

Pred [0] [1] = Pred [2] [2] = (A + B*2 + A1+ 2) / 4

Pred [0] [2] = Pred [2] [3] = (B + C + 1) / 2;

Pred [0] [3] = (C + D + 1) / 2;

Pred [1] [0] = Pred [3] [1] = (I + 2*X + XI + 2) /

Pred [1] [1] = Pred [3] [2] = (Al + 2*A + X + 2) 1

Pred [1] [2] = Pred [3] [3] = (Bl + 2*B + A + 2) 1

Pred [1] [3] = (CI + 2*C + B + 2) 14;

Pred [2] [0] = (X + 2*1 + J + 2) 14;

Pred [3] [0] = (I + 2*J + K + 2) 14; Pred[i][j]: the predicted value of the current encoded pixel,

i: the pixel is located in the row of the current macroblock;

j: The pixel is in the column of the current macroblock.

S129, sixth weighting mode

Sixth Weighting Mode: This mode predicts the correlation degree information of the pixels of the current coded sub-block and its encoded neighboring and neighboring pixels. In view of this, the sixth weighting mode, i.e., the pixel of the current coded sub-block, is represented by a linear function of the nearest neighboring, next neighboring coded pixels in the upper, left, and upper left encoded sub-blocks. The specific implementation is as follows:

Side and left, U sub-blocks are encoded)

Pred [0] [0] = (X + I + 1) / 2

Fred [1] [2] = (X + 1*3 + 2) / 4

Pred [0] [1] = (3*X + A + 2) / 4;

Pred [1] [3] = (X + A*3 + 2) / 4;

Pred [0] [2] = (Xl+X + A + B + 2) 1

Pred [0] [3] = (A1+ A + B + C + 2)

Pred [1] [0] = (Il+I + J *2+ 2) / 4

Pred [2] [2] = (I + J + 1) / 2;

Pred [1] [1] = (X2 +1*3 + 2) / 4;

Pred [2] [3] = (X2 + I + 1) /2;

Pred [2] [0] = (J + K*2 + Jl+2) 14

Pred [3] [2] = (J + K +1) 12;

Pred [2] [1] = (11 + 2* J + J1+ 2) 4

Pred [3] [3] = (11 + J + J1+ 1) 13 Fr ed [3] [0] = (K + L*2 + K1 + 2) / 4 ;

Fr ed [3] [1] = (Jl + 2* K + Kl + 2) / 4 ;

Pved[i][j] : the predicted value of the current encoded pixel,

i : the pixel is in the row of the current macroblock;

j : The pixel is in the column of the current macroblock.

S2, chroma intra prediction

S21, firstly, the current coded chroma block (the corresponding chroma block in the macroblock) is evenly divided into 4 sub-blocks; taking into account the distance information of the pixel of the current coded chroma block and its encoded adjacent coded pixel Far from this correlation, the correlation is also reduced, and different coding sub-blocks are used to predict different coding sub-blocks, thereby improving the coding effect.

On the one hand, the human eye is relatively insensitive to chrominance information. On the other hand, since the YUV420 format encoding source currently used has a large sampling point spacing in chroma, it is necessary to update the required prediction pixel value in chrominance prediction. In combination with the above features, the embodiment of the present invention proposes a sub-block based chroma intra prediction mode. To describe the 8x8 chroma intra prediction, the chrominance information of the currently coded macroblock MB and the predicted pixels required for intra prediction are labeled as follows:

Blockll, Blockl2, Block21, and Block22 are: 4 sub-blocks that are uniformly divided into the current coded chroma block, called a chroma sub-block;

A B C D E F G H is: the first row of pixel values on the upper side of the current coded chroma block; IJKLMNOP is: the first column of pixel values on the left side of the current coded chroma block;

X is: the pixel value of the upper left corner of the current coded chroma block.

S211, sub-block based progressive DC mode

Sub-block-based progressive DC mode: The upper left sub-block pixel of the current coded chroma block is described using the mean of the row and column pixels closest to the current coded chroma block in the coded chroma block on the upper and left sides. The other three sub-blocks are respectively represented by the mean of the row and column pixels closest to the current coded chroma sub-block and the mean of the coded sub-blocks closest to the current coded chroma block.

The specific implementation is as follows:

a) Blockll

If (upper and left subblocks are encoded)

Pr eJ _n [i][j] = mean(A, B, C, D, I, J, K, L)

Else if (only the upper sub-block is encoded)

Pr ed _n [i] [j] = mean(A, B, C, D)

Else if (only the left subblock is encoded)

PreJnfiJfj] = mean(I,J ,K,L)

Else

Pr [i][j] = 128

Pved[i][j]: the predicted value of the current encoded pixel,

i : the pixel is in the row of the current chroma block;

j : the pixel is in the column of the current chroma block; Mean: seeking mean 〇b) Blockl2

If (the upper sub-block is encoded)

Pre<i ₁₂ [i][j] = mean(E, F,G,H ,Pred _n ) else

Pred ₁₂ [i][i] = Pred _n

Pved[i][j]: the predicted value of the current coded pixel, i : the pixel is in the row of the current chroma block; j : the pixel is in the column of the current chroma block; mean: seeking the mean 〇 c)Block21

If (the left subblock is encoded)

Pre<i ₂₁ [i][j] = mean(M , Ν, Ο, Ρ, Pr ed _n ) else

Pred ₂₁ [i][i] = Pred _n

Pved[i][j]: the predicted value of the current encoded pixel, i : the pixel is in the row of the current chroma block; j : the pixel is in the column of the current chroma block; mean: seeking the mean 〇 c) Block22

Pre<i ₂₂ [i][j] = mean(Fr ed ₁₂ , Pr ed ₂₁ )

Pved[i][j]: the predicted value of the current coded pixel, i : the pixel is located in the row of the current chroma block; j : the pixel is in the column of the current chroma block;

Mean: seeking mean 〇

5212, second gradient mode

In the 8x8 chromaticity prediction, for the macroblock having the gradation characteristic, the gradation mode of the embodiment of the present invention can be used for the compression of the redundant information. The second gradation mode: that is, the pixel of the current coded chrominance block is represented by a linear function of the pixel mean value closest to the current chrominance block among the coded chrominance blocks on the upper side and the left side.

The specific implementation is as follows:

Mean _u = mean(A, B, C, D, E, F, G, H)

Mean _ 1 = mean(I, J, K, L, M, N, O, P)

BlocklK Blockl2, Block21, Block22

Pre<i[i][j] = ((8 - j) * mean _ u + (8 - i) * mean _ h)/(16 - i - j)

Pved[i][j]: the predicted value of the current encoded pixel,

i : the pixel is in the row of the current chroma block;

j : the pixel is in the column of the current chroma block;

Mean: seeking mean 〇

5213, sub-block based upper mode

On the one hand, the human eye is relatively insensitive to the chrominance information. On the other hand, the sampling point spacing is large on the chromaticity, and the prediction is too rough to introduce a great prediction error. Based on this, the embodiment of the present invention proposes a sub-block based upper side mode, that is, the pixel of the current coded chroma block is directly represented by the nearest neighbor sub-block information of the upper coded chroma block. The specific implementation is as follows: UP ₁₂

Lefln Blockll Blockl2

Lefi _2i Block21 Block22

Blockll Blockl2 Block21 Block22 is: 4 sub-blocks that evenly divide the current coded chroma block, that is, each size is 4x4;

UP _N UP ₁₂ is: the upper two sub-blocks of the current coded chroma block, the size is 4x4;

L _e ft _u UP _l2 "k: two sub-blocks on the left side of the current coded chroma block, all of which are 4x4;

If (upper side coded)

Pr _e J ₁₁ [i][j] = C/P ₁₁ [i][j] ;

Pr^ ₁₂ [i][j] = t/F ₁₂ [i][j];

Pr _e J ₂₁ [i][j] = C/P _n [3][j] ;

Pr _e J ₂₂ [i][j] = C/P ₁₂ [3][j];

Pr^/ _U [i][j]: the pixel prediction value of the current coded chroma sub-block Blockll;

Pr^ ₁₂ [i][j]: the pixel prediction value of the current coded chroma sub-block Blockl2;

Pr^ ₂₁ [i][j]: the pixel prediction value of the current coded chroma sub-block Block21;

Pr^ ₂₂ [i][j]: the pixel prediction value of the current coded chroma sub-block Block22;

UP _n vm: a pixel prediction value of the upper sub-block of the current coded chroma sub-block Blockl;

UP ₁₂ im: the pixel prediction value of the upper sub-block of the current coding chroma sub-block Blockl2; C/P _n [3][j]: the pixel prediction value of the 4th row in the upper subblock of the current coding chroma subblock Blockll; C/P ₁₂ [3][j]: the current coding chroma subblock Blockl2 The pixel prediction value of the 4th row in the side sub-block; i : the pixel is located in the row of the current chroma block;

j: The pixel is in the column of the current chroma block.

S214, sub-block based left mode

On the one hand, the human eye is relatively insensitive to chrominance information. On the other hand, the sampling point spacing is large on the chromaticity, and the prediction is too rough to introduce a large prediction error. Based on this, the embodiment of the present invention proposes, based on the left block mode of the sub-block, that is, the pixel of the current coded chroma block is directly represented by the nearest neighbor sub-block information of the left coded chroma block. The specific implementation is as follows:

If (coded on the left)

Pr^ _n [i][j] = L^ _n [i][j] Prt3⁄4? ₂₁ [i][j] = Le i ₂₁ [i][j]

Prt3⁄4? ₁₂ [i][j] = Le i _n [i][3] PreJ ₂₂ [i][j] = Left ₂₁ [i][3]

Pr^ _n [i][j]: the pixel prediction value of the current coded chroma sub-block Blockll;

Pred ₁₂ [m: pixel prediction value of the current coded chroma sub-block Blockl2;

Pr _e J ₂₁ [i][j] : the pixel prediction value of the current coded chroma sub-block Block 21;

Pr ₂₂ [i][j]: the pixel prediction value of the current coded chroma sub-block Block22;

Left _n [m: the pixel prediction value of the left subblock of the current coded chroma subblock Blockll;

Left ₂₁ [i][}: the pixel prediction value of the left sub-block of the current coded chroma sub-block Block21;

L^ _n [i][3]: the pixel prediction value of the fourth column in the left subblock of the current coded chroma subblock Blockll; L^ ₂₁ [i][3]: the current coded chroma subblock Block21 left side The pixel prediction value of the fourth column in the block; i : the pixel is located in the row of the current chroma block;

j: The pixel is in the column of the current chroma block. It will be understood by those skilled in the art that all or part of the steps of the foregoing embodiments may be implemented by a program instruction related hardware, and the program may be stored in a computer readable storage medium. The storage medium may be a ROM, a RAM, a magnetic disk, an optical disk, or the like.

The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. Within the scope.

Claims

Rights request

A video coding intra prediction method, the method comprising: using different brightness prediction according to an image feature, a degree of association between a coded pixel and a neighboring coded pixel, and a second adjacent coded pixel information;

Sub-block-based chroma intra prediction is used to perform chromaticity prediction using different pixels for different coded sub-blocks according to the degree of association between the pixels of the current coded sub-block and the adjacent coded pixels.

2. The video coding intra prediction method according to claim 1, wherein the luminance prediction comprises prediction of a 16x16 luma block and prediction of a 4x4 luma block.

3. The video coding intra prediction method according to claim 2, wherein:

The prediction of the 16x16 luma block is specifically: using an intra 11_8^, V_Skip mode, combining the image feature and the degree of association between the coded pixel and the adjacent coded pixel, using the first fade mode and the first progressive DC The mode is predicted.

4. The video coding intra prediction method according to claim 3, wherein:

The H_Skip, V_Skip mode is: the pixels of the currently coded macroblock are represented by the coded macroblock information on the left and upper sides of the pixels of the current coded macroblock.

5. The video coding intra prediction method according to claim 4, wherein:

The first gradation mode is that the pixels of the currently coded macroblock are represented by a linear function of the pixel mean of the upper and left coded macroblocks closest to the currently coded macroblock.

6. The video coding intra prediction method according to claim 4, wherein:

The first progressive DC mode is that the pixels of the current coded macroblock are represented by the mean value of the pixel variations of the two rows and two columns of the current coded macroblock, which are used in the upper and left coded macroblocks.

7. The video coding intra prediction method according to claim 2, wherein:

The prediction of the 4×4 luma block is specifically: performing prediction according to the degree of association between the encoded pixel and the encoded neighboring pixel and the encoded secondary neighboring pixel information, where according to different predictions The direction uses different 4x4 luma block prediction modes.

The video coding intra prediction method according to claim 7, wherein the different 4x4 luma block prediction mode comprises a second progressive DC mode, a weighted Up mode, a weighted Left mode, a first weighting mode, and a second The two weighting mode, the third weighting mode, the fourth weighting mode, the fifth weighting mode, and the sixth weighting mode.

The video coding intra prediction method according to claim 8, wherein the second progressive DC mode is: the pixels of the current coding sub-block are used in the upper and left encoded sub-blocks, and the closest The mean of the pixel variations of the two rows and two columns of the current coded macroblock is represented.

The video coding intra prediction method according to claim 8, wherein the weighted Up mode is: the pixels of the current coded sub-block use the most adjacent and second adjacent coded pixels in the upper coded sub-block. The linear function is represented.

The video coding intra prediction method according to claim 8, wherein the weighted Left mode is: the pixels of the current coded sub-block use the most adjacent and next adjacent coded pixels in the left coded sub-block. The linear function is represented.

The video coding intra prediction method according to claim 8, wherein the first weighting mode and the second weighting mode are: the coded upper and upper right sides of the pixel of the current coding sub-block are used. A linear function of the nearest neighboring coded pixel in the sub-block indicates that after the coded pixel is farther away from the coded pixel, the predicted value of the coded pixel of the current coded sub-block is directly used to replace the nearest neighbor of the current coded block. Encoded pixels.

The video coding intra prediction method according to claim 8, wherein the third weighting mode is: the pixels of the current coded sub-block use the nearest adjacent coded pixel of the coded sub-block on the left side. The linear function indicates that after the coded pixel is farther away from the coded pixel, the nearest code of the current coded sub-block is replaced by the predicted value of the coded pixel of the current coded sub-block. Pixel.

The video coding intra prediction method according to claim 8, wherein the fourth weighting mode, the fifth weighting mode, and the sixth weighting mode are: using pixels of a current coding sub-block A linear function of the nearest neighboring, next neighboring coded pixels in the encoded sub-blocks of the side, left side, and upper left side is represented.

The video coding intra prediction method according to claim 1, wherein the “sub-block-based chroma intra prediction is performed according to a degree of association between a pixel of a current coding sub-block and a neighboring coded pixel, The chrominance prediction is performed by using different pixels for different coding sub-blocks. Specifically, the chrominance information of the current coded macroblock is evenly divided into four sub-blocks according to the correlation degree between the pixels of the current coded sub-block and the adjacent coded pixels. Prediction is performed using different pixels for different coded sub-blocks, including the following prediction modes: a sub-block based progressive DC mode, a second fade mode, a sub-block based upper mode, and a sub-block based left mode.

The video coding intra prediction method according to claim 15, wherein the sub-block-based progressive DC mode is specifically: the upper left sub-block pixel of the current coded macro block is used on the upper side and the left side. The average of the row and column pixels of the coded macroblock, which is closest to the current macroblock, is described; and the other three subblocks respectively utilize the mean of the row and column pixels closest to the current macroblock and the encoded of the currently coded macroblock. The mean of the sub-blocks is represented.

The video coding intra prediction method according to claim 15, wherein the second gradation mode is specifically: the pixels in the current coded macroblock use the coded macroblocks on the upper side and the left side, the most A linear function close to the pixel mean of the current macroblock.

The video coding intra prediction method according to claim 15, wherein the sub-block-based upper side mode is specifically: the pixel in the current coded macro block uses the nearest neighbor of the coded macro block. Sub-block information to represent.

The video coding intra prediction method according to claim 15, wherein the sub-block-based left mode is specifically: the pixels in the current coded macro block are encoded on the left side.