CN101562747B

CN101562747B - Method for resolving and rebuilding video coding predictive residue block

Info

Publication number: CN101562747B
Application number: CN 200910062165
Authority: CN
Inventors: 陈加忠; 周敬利; 黎单; 孙自龙
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2009-05-19
Filing date: 2009-05-19
Publication date: 2010-08-25
Anticipated expiration: 2029-05-19
Also published as: CN101562747A

Abstract

The invention provides a method for dividing and rebuilding a video coding predictive residue block, and belongs to the technical field of video coding in signal treatment. The invention aims to reduce the blocking effect due to the prior DCT transformation matrix and solve the problem that the value of the matrix cannot be adjusted. The division method comprises a predictive residue block classification step, a transverse division step of first, third and second multi-channel filter groups, a classification step of transverse division coefficient matrixes and a longitudinal division step of the first, the third and the second multi-channel filter groups; and the rebuilding method comprises a classification step of the division coefficient matrixes of the predictive residue block, a longitudinal rebuilding step of the first, the third and the second multi-channel filter groups; a classification step of the transverse division coefficient matrixes and a transverse rebuilding step of thefirst, the third and the second multi-channel filter groups. The method has the advantages of effectively removing correlation, reducing the blocking effect due to the size mismatching of the DCT tra nsformation matrix and the predictive residue block, and improving subjective and objective qualities of coding.

Description

Decomposition and reconstruction method of video coding prediction residual block

Technical Field

The invention belongs to the technical field of video coding in signal processing, and particularly relates to a decomposition and reconstruction method of a video coding prediction residual block.

Background

H.264/AVC is the latest video coding standard proposed jointly by the video coding experts group of ITU-T (International telecommunication Union, telecommunication standardization sector) and the moving Picture experts group of ISO/IEC (International organization for standardization/International electrotechnical Commission). Compared with the MPEG-4 coding standard, the code rate of H.264/AVC can be saved by about 50 percent under the condition of obtaining the same video quality. h.264/AVC to achieve high compression ratio, some new coding methods are proposed, such as intra prediction, motion compensation, loop filtering, context-based entropy coding, etc.

As shown in fig. 1, in a video encoding process, intra-frame or inter-frame prediction is performed on a video image, an optimal prediction mode is selected by calculating a cost function in the prediction process, a residual between an original image and a predicted image is decomposed to obtain a decomposition coefficient matrix, the decomposition coefficient matrix is quantized and entropy-encoded to obtain an output code stream, and meanwhile, quantization coefficients are inversely quantized, and a prediction residual block is reconstructed and reconstructed to obtain a reconstructed image used for a reference image in the next frame encoding.

The intra prediction method of h.264/AVC is based on the spatial domain, i.e., the current macroblock is predicted from the reconstructed values of the pixels of the neighboring macroblocks. h.264/AVC defines two luma prediction residual block sizes (16 × 16 and 4 × 4) and one chroma size (8 × 8), respectively. Among them, 9 prediction modes are defined for a 4 × 4 prediction residual block, and only 5 prediction modes are defined for a 16 × 16 luminance block and an 8 × 8 chrominance block. A prediction residual block size of 16 x 16 is suitable for smooth regions, while a prediction residual block size of 4 x 4 is suitable for regions with rich texture features.

The inter-frame prediction method of h.264/AVC mainly removes temporal redundancy of video images by a motion compensation prediction coding method using correlation between consecutive image sequences. H.264/AVC defines seven luma inter prediction residual block sizes (16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, 4 × 4), respectively. The motion compensation of different block sizes can more accurately represent the motion characteristics of the macro block and reduce the prediction residual error. Motion estimation with larger block sizes is suitable for video objects with simple texture and insignificant motion in macroblocks, while motion estimation with smaller block sizes is suitable for video objects with significant motion in macroblocks.

h.264/AVC uses a 4 × 4 Discrete Cosine Transform (DCT) to decompose the prediction residual blocks of the luminance and chrominance blocks to further remove spatial redundancy. The 4 x 4 discrete cosine transform used in h.264/AVC has low computational complexity and is effective in reducing ringing noise, but for low rate coded pictures, blocking artifacts will exist within blocks and at block boundaries. Since the size of the discrete cosine transform matrix is smaller than the prediction residual block (16 × 16, 16 × 8, 8 × 16), blocking artifacts inevitably occur on the transform boundary of the block; meanwhile, the value of the DCT transform matrix is fixed, and it cannot be adjusted according to the energy distribution of the prediction residual block signal.

Disclosure of Invention

The invention provides a decomposition method of a video coding prediction residual block, and simultaneously provides a reconstruction method of the video coding prediction residual block, which reduces the block effect caused by the size mismatching of the existing DCT matrix and the prediction residual block, and solves the problem that the value of the DCT matrix can not be adjusted according to the energy distribution of the prediction residual block signal, thereby effectively removing the correlation of the prediction residual block and improving the video coding quality.

The invention discloses a decomposition method of a video coding prediction residual block, which comprises the following steps:

(1) classifying the prediction residual blocks: classifying the input intra-frame or inter-frame prediction residual block, and performing the step (2) when the long edge size of the residual block is 4; when the size of the long edge of the residual block is 8, performing the step (3); when the size of the long side of the residual block is 16, performing the step (4);

(2) a first multi-channel filter bank transverse decomposition step: using a first multi-channel filter bank

F_{1} = [\begin{matrix} U_{0} & U_{1} \\ W_{0} & W_{1} \end{matrix}]

And carrying out transverse decomposition on the residual block X to obtain a transverse decomposition coefficient matrix Y, wherein the decomposition expression is as follows:

wherein,

α＝0.1225π；

after the decomposition is finished, performing the step (5);

(3) a third multi-channel filter bank transverse decomposition step: using a third multi-channel filter bank

F_{3} = [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}]

Y = X [\begin{matrix} {R_{0}}^{T} & {T_{0}}^{T} \\ {R_{1}}^{T} & {T_{1}}^{T} \end{matrix}]

wherein,

R₁＝SR₀PSP，

T₁＝ST₀PSP，

P = \frac{1}{2} [\begin{matrix} 1 & 1 & 1 & 1 \\ 1 & - 1 & 1 & - 1 \\ 1 & 1 & - 1 & - 1 \\ 1 & - 1 & - 1 & 1 \end{matrix}],

S = \frac{1}{2} [\begin{matrix} 1 \\ - 1 \\ 1 \\ - 1 \end{matrix}],

α＝0.4362π，β＝0.38π；

after the decomposition is finished, performing the step (5);

(4) transverse decomposition step of second multi-channel filter bankThe method comprises the following steps: using a second multi-channel filter bank

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}]

Carrying out transverse decomposition on the residual block X to obtain a transverse decomposition coefficient matrix Y, wherein the process is as follows:

(4.1) performing horizontal symmetric expansion and decomposition on each row vector of the residual block X:

for the ith row vector N_i＝[x_i，0x_i，1x_i，2x_i，3x_i，4x_i，5x_i，6x_i，7x_i，8x_i，9x_i，10x_i，11x_i，12x_i，13x_i，14x_i，15]Transversely symmetrically expanding to obtain an expanded row vector Z_i，Z_i＝[x_i，1x_i，0x_i，0x_i，1x_i，2x_i，3x_i，4x_i，5x_i，6x_i，7x_i，8x_i，9x_i，10x_i，11x_i，12x_i，13x_i，14x_i，15x_i，15x_i，14]To Z is paired with_iPerforming transverse decomposition to obtain the ith row vector N_iTransverse decomposition vector M of_i：

M_i＝[y_i，0y_i，1y_i，2y_i，3y_i，4y_i，5y_i，6y_i，7y_i，8y_i，9y_i，10y_i，11y_i，12y_i，13y_i，14y_i，15]，

Wherein:

[y_i，0y_i，1y_i，2y_i，3]＝[x_i，1x_i，0x_i，0x_i，1x_i，2x_i，3x_i，4x_i，5]F₂ ^T，

[y_i，4y_i，5y_i，6y_i，7]＝[x_i，2x_i，3x_i，4x_i，5x_i，6x_i，7x_i，8x_i，9]F₂ ^T，

[y_i，8y_i，9y_i，10y_i，11]＝[x_i，6x_i，7x_i，8x_i，9x_i，10x_i，11x_i，12x_i，13]F₂ ^T，

[y_i，12y_i，13y_i，14y_i，15]＝[x_i，10x_i，11x_i，12x_i，13x_i，14x_i，15x_i，15x_i，14]F₂ ^T，

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}],

alpha is-0.595 pi, i is 0-k-1, k is less than or equal to 16, k is the wide side of the residual block XDetermining the size; (4.2) transverse decomposition vector M of each row vector_iA matrix Y of transverse decomposition coefficients is constructed,

Y = [\begin{matrix} M_{0} \\ M_{1} \\ . \\ . \\ . \\ M_{k - 1} \end{matrix}],

wherein k is determined by the broadside size of the residual block X;

after the decomposition is finished, performing the step (5);

(5) and transverse decomposition coefficient matrix classification: classifying the transverse decomposition coefficient matrix Y, and performing the step (6) when the width side size of Y is 4; when the width side size of Y is 8, performing the step (7); when the width side size of Y is 16, performing the step (8);

(6) a first multi-channel filter bank longitudinal decomposition step: using a first multi-channel filter bank

F_{1} = [\begin{matrix} U_{0} & U_{1} \\ W_{0} & W_{1} \end{matrix}]

And longitudinally decomposing the transverse decomposition coefficient matrix Y to obtain a decomposition coefficient matrix D of the residual block, wherein the decomposition expression is as follows:

wherein, U₀、U₁、W₀、W₁And the value of α is the same as shown in step (2);

finishing the decomposition;

(7) a third multi-channel filter bank longitudinal decomposition step: using a third multi-channel filter bank

F_{3} = [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}]

D = [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}] Y

wherein R is₀、R₁、T₀、T₁The values of (a) are the same as shown in step (3);

finishing the decomposition;

(8) a second multi-channel filter bank longitudinal decomposition step: using a second multi-channel filter bank

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}]

And longitudinally decomposing the transverse decomposition coefficient matrix Y to obtain a decomposition coefficient matrix D of the residual block, wherein the process is as follows:

(8.1) longitudinally symmetrically expanding and decomposing each column vector of the transverse decomposition coefficient matrix Y:

for the ith column vector L_i＝[y_0，iy_1，iy_2，iy_3，iy_4，iy_5，iy_6，iy_7，iy_8，iy_9，iy_10，iy_11，iy_12，iy_13，iy_14，iy_15，i]^TLongitudinally symmetric expansion to obtain an expanded column vector Z_i，Z_i＝[y_1，iy_0，iy_0，iy_1，iy_2，iy_3，iy_4，iy_5，iy_6，iy_7，iy_8，iy_9，iy_10，iy_11，iy_12，iy_13，iy_14，iy_15，iy_15，iy_14，i]^TTo Z is paired with_iPerforming longitudinal decomposition to obtain the ith column vector L_iLongitudinal decomposition vector J of_i：

J_i＝[d_0，id_1，id_2，id_3，id_4，id_5，id_6，id_7，id_8，id_9，id_10，id_11，id_12，id_13，id_14，id_15，i]^T，

Wherein:

[d_0，id_1，id_2，id_3，i]^T＝F₂[y_1，iy_0，iy_0，iy_1，iy_2，iy_3，iy_4，iy_5，i]^T，

[d_4，id_5，id_6，id_7，i]^T＝F₂[y_2，iy_3，iy_4，iy_5，iy_6，iy_7，iy_8，iy_9，i ]^T，

[d_8，id_9，id_10，id_11，i]^T＝F₂[y_6，iy_7，iy_8，iy_9，iy_10，i y_11，iy_12，iy_13，i]^T，

[d_12，id_13，id_14，id_15，i]^T＝F₂[y_10，iy_11，iy_12，iy_13，iy_14，iy_15，iy_15，iy_14，i]^T，

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}],

H₀、H₁、H₂、H₃、G₀、G₁、G₂、G₃the value of (c) is the same as that shown in step (4.1), i is 0-k-1, k is less than or equal to 16, and k is determined by the size of the long side of the residual block X;

(8.2) vertical decomposition vector J of each column vector_iA matrix D of decomposition coefficients constituting the residual block,

D＝[J₀J₁…J_k-1]where k is determined by the size of the long side of the residual block X;

and finishing the decomposition.

The invention discloses a reconstruction method of a video coding prediction residual block, which reconstructs the prediction residual block by utilizing a decomposition coefficient matrix and comprises the following steps:

(1) and classifying a decomposition coefficient matrix of the prediction residual block: classifying the input decomposition coefficient matrix D, and performing the step (2) when the broadside size of the decomposition coefficient matrix is 4; when the size of the wide side of the decomposition coefficient matrix is 8, performing the step (3); when the size of the wide side of the decomposition coefficient matrix is 16, performing the step (4);

(2) a first multi-channel filter bank longitudinal reconstruction step: using a first multi-channel filter bank

F_{1} = [\begin{matrix} U_{0} & U_{1} \\ W_{0} & W_{1} \end{matrix}]

Longitudinally reconstructing the decomposition coefficient matrix D to obtain a transverse decomposition coefficient matrix Y, wherein the reconstruction expression is as follows:

wherein,

α＝0.1225π；

after reconstruction is completed, performing the step (5);

(3) a third multi-channel filter bank longitudinal reconstruction step: using a third multi-channel filter bank

F_{3} = [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}]

Y = [\begin{matrix} {R_{0}}^{T} & {T_{0}}^{T} \\ {R_{1}}^{T} & {T_{1}}^{T} \end{matrix}] D

wherein,

R₁＝SR₀PSP，

T₁＝ST₀PSP，

P = \frac{1}{2} [\begin{matrix} 1 & 1 & 1 & 1 \\ 1 & - 1 & 1 & - 1 \\ 1 & 1 & - 1 & - 1 \\ 1 & - 1 & - 1 & 1 \end{matrix}],

S = \frac{1}{2} [\begin{matrix} 1 \\ - 1 \\ 1 \\ - 1 \end{matrix}],

α＝0.4362π，β＝0.38π；

after reconstruction is completed, performing the step (5);

(4) a second multi-channel filter bank longitudinal reconstruction step: using a second multi-channel filter bank

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}] = [\begin{matrix} h_{00} & h_{01} & h_{02} & h_{03} & h_{04} & h_{05} & h_{06} & h_{07} \\ h_{10} & h_{11} & h_{12} & h_{13} & h_{14} & h_{15} & h_{16} & h_{17} \\ g_{00} & g_{01} & g_{02} & g_{03} & g_{04} & g_{05} & g_{06} & g_{07} \\ g_{10} & g_{11} & g_{12} & g_{13} & g_{14} & g_{15} & g_{16} & g_{17} \end{matrix}]

Carrying out longitudinal and longitudinal reconstruction on the decomposition coefficient matrix D to obtain a transverse and longitudinal decomposition coefficient matrix Y, wherein the process is as follows:

(4.1) reconstructing each column vector of the decomposition coefficient matrix D:

for ith row vector J of decomposition coefficient matrix D_i＝[d_0，id_1，id_2，id_3，id_4，id_5，id_6，i d_7，id_8，i d_9，i d_10，i d_11，i d_12，i d_13，i d_14，i d_15，i]^TReconstructing to obtain the ith column vector L of the transverse decomposition coefficient matrix Y_i＝[y_0，iy_1，iy_2，iy_3，iy_4，iy_5，iy_6，iy_7，iy_8，iy_9，iy_10，iy_11，iy_12，iy_13，iy_14，iy_15，i]^TThe calculation expression is as follows:

y_0，i＝d_0，i×h₀₂+d_0，i×h₀₆+d_1，i×h₁₂-d_1，i×h₁₆+d_2，i×g₀₂+d_2，i×g₀₆+d_3，i×g₁₂-d_3，i×g₁₆

y_1，i＝d_0，i×h₀₃+d_0，i×h₀₇+d_1，i×h₁₃-d_1，i×h₁₇+d_2，i×g₀₃+d_2，i×g₀₇+d_3，i×g₁₃-d_3，i×g₁₇

y_2，i＝d_0，i×h₀₄+d_4，i×h₀₀+d_1，i×h₁₄+d_5，i×h₁₀+d_2，i×g₀₄+d_6，i×g₀₀+d_3，i×g₁₄+d_7，i×g₁₀

y_3，i＝d_0，i×h₀₅+d_4，i×h₀₁+d_1，i×h₁₅+d_5，i×h₁₁+d_2，i×g₀₅+d_6，i×g₀₁+d_3，i×g₁₅+d_7，i×g₁₁

y_4，i＝d_0，i×h₀₆+d_4，i×h₀₂+d_1，i×h₁₆+d_5，i×h₁₂+d_2，i×g₀₆+d_6，i×g₀₂+d_3，i×g₁₆+d_7，i×g₁₂

y_5，i＝d_0，i×h₀₇+d_4，i×h₀₃+d_1，i×h₁₇+d_5，i×h₁₃+d_2，i×g₀₇+d_6，i×g₀₃+d_3，i×g₁₇+d_7，i×g₁₃

y_6，i＝d_4，i×h₀₄+d_8，i×h₀₀+d_5，i×h₁₄+d_9，i×h₁₀+d_6，i×g₀₄+d_10，i×g₀₀+d_7，i×g₁₄+d_11，i×g₁₀

y_7，i＝d_4，i×h₀₅+d_8，i×h₀₁+d_5，i×h₁₅+d_9，i×h₁₁+d_6，i×g₀₅+d_10，i×g₀₁+d_7，i×g₁₅+d_11，i×g₁₁

y_8，i＝d_4，i×h₀₆+d_8，i×h₀₂+d_5，i×h₁₆+d_9，i×h₁₂+d_6，i×g₀₆+d_10，i×g₀₂+d_7，i×g₁₆+d_11，i×g₁₂

y_9，i＝d_4，i×h₀₇+d_8，i×h₀₃+d_5，i×h₁₇+d_9，i×h₁₃+d_6，i×g₀₇+d_10，i×g₀₃+d_7，i×g₁₇+d_11，i×g₁₃

y_10，i＝d_8，i×h₀₄+d_12，i×h₀₀+d_9，i×h₁₄+d_13，i×h₁₀+d_10，i×g₀₄+d_14，i×g₀₀+d_11，i×g₁₄+d_15，i×g₁₀

y_11，i＝d_8，i×h₀₅+d_12，i×h₀₁+d_9，i×h₁₅+d_13，i×h₁₁+d_10，i×g₀₅+d_14，i×g₀₁+d_11，i×g₁₅+d_15，i×g₁₁

y_12，i＝d_8，i×h₀₆+d_12，i×h₀₂+d_9，i×h₁₆+d_13，i×h₁₂+d_10，i×g₀₆+d_14，i×g₀₂+d_11，i×g₁₆+d_15，i×g₁₂

y_13，i＝d_8，i×h₀₇+d_12，i×h₀₃+d_9，i×h₁₇+d_13，i×h₁₃+d_10，i×g₀₇+d_14，i×g₀₃+d_11，i×g₁₇+d_15，i×g₁₃

y_14，i＝d_12，i×h₀₄+d_12，i×h₀₀+d_13，i×h₁₄-d_13，i×h₁₀+d_14，i×g₀₄+d_14，i×g₀₀+d_15，i×g₁₄-d_15，i×g₁₀

y_15，i＝d_12，i×h₀₅+d_12，i×h₀₁+d_13，i×h₁₅-d_13，i×h₁₁+d_14，i×g₀₅+d_14，i×g₀₁+d_15，i×g₁₅-d_15，i×g₁₁

wherein:

alpha is-0.595 pi, i is 0-k-1, k is less than or equal to 16, and k is determined by the long side size of the decomposition coefficient matrix D;

(4.2) longitudinal reconstructed vector L of each column vector_iA matrix Y of transverse decomposition coefficients is constructed,

Y＝[L₀L₁…L_k-1]wherein k is determined by the dimension of the long side of the decomposition coefficient matrix D;

after reconstruction is completed, performing the step (5);

(5) and transverse decomposition coefficient matrix classification: classifying the transverse decomposition coefficient matrix Y, and performing the step (6) when the long side size of Y is 4; when the long side size of Y is 8, performing the step (7); when the long side size of Y is 16, performing the step (8);

(6) a first multi-channel filter bank transverse reconstruction step: using a first multi-channel filter bank

F_{1} = [\begin{matrix} U_{0} & U_{1} \\ W_{0} & W_{1} \end{matrix}]

And transversely reconstructing the transverse decomposition coefficient matrix Y to obtain a residual block X, wherein the reconstruction expression is as follows:

wherein, U₀、U₁、W₀、W₁And the value of α is the same as shown in step (2); finishing reconstruction;

(7) a third multi-channel filter bank transverse reconstruction step: using a third multi-channel filter bank

F_{3} = [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}]

X = Y [\begin{matrix} R_{0} & R_{1} \\ T_{0} & T_{1} \end{matrix}]

finishing reconstruction;

(8) a second multi-channel filter bank transverse reconstruction step: using a second multi-channel filter bank

F_{2} = [\begin{matrix} H_{0} & H_{1} & H_{2} & H_{3} \\ G_{0} & G_{1} & G_{2} & G_{3} \end{matrix}] = [\begin{matrix} h_{00} & h_{01} & h_{02} & h_{03} & h_{04} & h_{05} & h_{06} & h_{07} \\ h_{10} & h_{11} & h_{12} & h_{13} & h_{14} & h_{15} & h_{16} & h_{17} \\ g_{00} & g_{01} & g_{02} & g_{03} & g_{04} & g_{05} & g_{06} & g_{07} \\ g_{10} & g_{11} & g_{12} & g_{13} & g_{14} & g_{15} & g_{16} & g_{17} \end{matrix}]

To transverse decomposition system

And performing transverse reconstruction on the number matrix Y to obtain a residual block X, wherein the process is as follows:

(8.1) reconstructing each row vector of the transverse decomposition coefficient matrix Y:

for the ith row vector M of the transverse decomposition coefficient matrix Y_i＝[y_i，0y_i，1y_i，2y_i，3y_i，4y_i，5y_i，6y_i，7y_i，8y_i，9y_i，10y_i，11y_i，12y_i，13y_i，14y_i，15]Rebuilding to obtain ith row vector N of residual block X_i＝[x_i，0x_i，1x_i，2x_i，3x_i，4x_i，5x_i，6x_i，7x_i，8x_i，9x_i，10x_i，11x_i，12x_i，13x_i，14x_i，15]The calculation expression is as follows:

x_i，0＝y_i，0×h₀₂+y_i，0×h₀₆+y_i，1×h₁₂-y_i，1×h₁₆+y_i，2×g₀₂+y_i，2×g₀₆+y_i，3×g₁₂-y_i，3×g₁₆

x_i，1＝y_i，0×h₀₃+y_i，0×h₀₇+y_i，1×h₁₃-y_i，1×h₁₇+y_i，2×g₀₃+y_i，2×g₀₇+y_i，3×g₁₃-y_i，3×g₁₇

x_i，2＝y_i，0×h₀₄+y_i，4×h₀₀+y_i，1×h₁₄+y_i，5×h₁₀+y_i，2×g₀₄+y_i，6×g₀₀+y_i，3×g₁₄+y_i，7×g₁₀

x_i，3＝y_i，0×h₀₅+y_i，4×h₀₁+y_i，1×h₁₅+y_i，5×h₁₁+y_i，2×g₀₅+y_i，6×g₀₁+y_i，3×g₁₅+y_i，7×g₁₁

x_i，4＝y_i，0×h₀₆+y_i，4×h₀₂+y_i，1×h₁₆+y_i，5×h₁₂+y_i，2×g₀₆+y_i，6×g₀₂+y_i，3×g₁₆+y_i，7×g₁₂

x_i，5＝y_i，0×h₀₇+y_i，4×h₀₃+y_i，1×h₁₇+y_i，5×h₁₃+y_i，2×g₀₇+y_i，6×g₀₃+y_i，3×g₁₇+y_i，7×g₁₃

x_i，6＝y_i，4×h₀₄+y_i，8×h₀₀+y_i，5×h₁₄+y_i，9×h₁₀+y_i，6×g₀₄+y_i，10×g₀₀+y_i，7×g₁₄+y_i，11×g₁₀

x_i，7＝y_i，4×h₀₅+y_i，8×h₀₁+y_i，5×h₁₅+y_i，9×h₁₁+y_i，6×g₀₅+y_i，10×g₀₁+y_i，7×g₁₅+y_i，11×g₁₁

x_i，8＝y_i，4×h₀₆+y_i，8×h₀₂+y_i，5×h₁₆+y_i，9×h₁₂+y_i，6×g₀₆+y_i，10×g₀₂+y_i，7×g₁₆+y_i，11×g₁₂

x_i，9＝y_i，4×h₀₇+y_i，8×h₀₃+y_i，5×h₁₇+y_i，9×h₁₃+y_i，6×g₀₇+y_i，10×g₀₃+y_i，7×g₁₇+y_i，11×g₁₃

x_i，10＝y_i，8×h₀₄+y_i，12×h₀₀+y_i，9×h₁₄+y_i，13×h₁₀+y_i，10×g₀₄+y_i，14×g₀₀+y_i，11×g₁₄+y_i，15×g₁₀

x_i，11＝y_i，8×h₀₅+y_i，12×h₀₁+y_i，9×h₁₅+y_i，13×h₁₁+y_i，10×g₀₅+y_i，14×g₀₁+y_i，11×g₁₅+y_i，15×g₁₁

x_i，12＝y_i，8×h₀₆+y_i，12×h₀₂+y_i，9×h₁₆+y_i，13×h₁₂+y_i，10×g₀₆+y_i，14×g₀₂+y_i，11×g₁₆+y_i，15×g₁₂

x_i，13＝y_i，8×h₀₇+y_i，12×h₀₃+y_i，9×h₁₇+y_i，13×h₁₃+y_i，10×g₀₇+y_i，14×g₀₃+y_i，11×g₁₇+y_i，15×g₁₃

x_i，14＝y_i，12×h₀₄+y_i，12×h₀₀+y_i，13×h₁₄-y_i，13×h₁₀+y_i，14×g₀₄+y_i，14×g₀₀+y_i，15×g₁₄-y_i，15×g₁₀

x_i，15＝y_i，12×h₀₅+y_i，12×h₀₁+y_i，13×h₁₅-y_i，13×h₁₁+y_i，14×g₀₅+y_i，14×g₀₁+y_i，15×g₁₅-y_i，15×g₁₁

wherein H₀、H₁、H₂、H₃、G₀、G₁、G₂、G₃The value of (c) is the same as that shown in step (4.1), i is 0-k-1, k is less than or equal to 16, and k is determined by the broadside size of the transverse decomposition coefficient matrix Y;

(8.2) transverse reconstructed vector N of each row vector_iA residual block X is formed,

X = [\begin{matrix} N_{0} \\ N_{1} \\ . \\ . \\ . \\ N_{k - 1} \end{matrix}],

wherein k is determined by the broadside size of the transverse decomposition coefficient matrix Y; and finishing the reconstruction.

The invention uses three multi-channel filter banks to carry out decomposition (transformation) processing on video data, the three multi-channel filter banks provided by the invention not only have orthogonal characteristic, but also have linear phase, in addition, the distribution of the spectral coefficients in a decomposition coefficient matrix obtained by the decomposition of the multi-channel filter banks is the same as that of 4 multiplied by 4 or 8 multiplied by 8 discrete cosine transform, and the spectral coefficients from the upper left corner to the lower right corner are changed from low to high; meanwhile, the invention can effectively perform decorrelation, reduce the blocking effect caused by the mismatching of the sizes of the DCT transformation matrix and the prediction residual block, and improve the subjective and objective quality of coding.

Experimental results show that the video coding quality of the decomposition processing by using the multi-channel filter bank is superior to the video coding quality of the decomposition processing by using DCT.

Drawings

FIG. 1 is a schematic flow chart of video encoding;

FIG. 2 is a schematic flow diagram of the decomposition process of the present invention;

FIG. 3 is a schematic flow chart of a reconstruction method according to the present invention.

Detailed Description

The decomposition method of the present invention is used for decomposing the prediction disparity block to obtain a decomposition coefficient matrix, and the flow of the decomposition coefficient matrix is shown in fig. 2.

The reconstruction method of the present invention reconstructs the prediction parameter block by using the decomposition coefficient matrix, and the flow is shown in fig. 3.

The encoding process will be implemented using 4 × 4, 8 × 8, 16 × 16 intra prediction modes in h.264 and inter prediction modes for 7 different sized blocks, with the filter banks being stored in table form in both the encoder and decoder. After the prediction residual block decomposition method based on the multi-channel filter bank is adopted, the decomposition coefficients are directly quantized and dequantized according to the quantization step size corresponding to the quantization parameter. The test platform is JM10.1, and a representative CIF (352X 288) with different colors and different texture characteristics of international standard sequences Bus, Mobile, Foreman and Coastguard is selected as a test sequence. The parameter settings of this embodiment are as follows:

1. and (3) coding structure: i frame 1 frame, P frame 29 frame

2. Entropy coding mode: CAVLC;

3. using a rate-distortion optimization model;

4. reference frame number: 1;

5. the search range is as follows: plus or minus 16 pixel points;

this embodiment shows that, compared with Discrete Cosine Transform (DCT), after the decomposition method of the present invention is adopted, the average peak signal-to-noise ratio is improved by more than 0.2dB under the same code rate. In particular, blocking artifacts will be generated near moving objects when using a DCT-based h.264/AVC encoder, whereas no blocking artifacts are generated when using a multi-channel filterbank-based h.264 encoder. Therefore, the prediction residual block decomposition method based on the multi-channel filter bank can effectively perform decorrelation in video coding, reduce the block effect caused by the mismatching of the sizes of the DCT transform matrix and the prediction residual block, and improve the coding subjective and objective quality.

Claims

1. A method of decomposition of a video coding prediction residual block, comprising:

wherein,

α＝0.1225π；

after the decomposition is finished, performing the step (5);

wherein,

R₁＝SR₀PSP，T₁＝ST₀PSP，

α＝0.4362π，β＝0.38π；

after the decomposition is finished, performing the step (5);

(4) and a second multi-channel filter bank transverse decomposition step: using a second multi-channel filter bank

Carrying out transverse decomposition on the residual block X to obtain a transverse decomposition coefficient matrix Y, wherein the process is as follows: (4.1) performing horizontal symmetric expansion and decomposition on each row vector of the residual block X:

Wherein:

[y_i，4 y_i，5 y_i，6 y_i，7]＝[x_i，2 x_i，3 x_i，4 x_i，5 x_i，6 x_i，7 x_i，8 x_i，9]F₂ ^T，

[y_i，8 y_i，9 y_i，10 y_i，11]＝[x_i，6 x_i，7 x_i，8 x_i，9 x_i，10 x_i，11 x_i，12 x_i，13]F₂ ^T，

[y_i，12 y_i，13 y_i，14 y_i，15]＝[x_i，10 x_i，11 x_i，12 x_i，13 x_i，14 x_i，15 x_i，15 x_i，14]F₂ ^T，

alpha is-0.595 pi, i is 0-k-1, k is less than or equal to 16, and k is determined by the width size of the residual block X;

(4.2) transverse decomposition vector M of each row vector_iA matrix Y of transverse decomposition coefficients is constructed,

wherein k is determined by the broadside size of the residual block X;

after the decomposition is finished, performing the step (5);

(5) and transverse decomposition coefficient matrix classification: the matrix of transverse decomposition coefficients Y is classified,

when the width side size of Y is 4, performing the step (6); when the width side size of Y is 8, performing the step (7); when the width side size of Y is 16, performing the step (8);

(6) a first multi-channel filter bank longitudinal decomposition step: using a first multi-channel filter bankAnd longitudinally decomposing the transverse decomposition coefficient matrix Y to obtain a decomposition coefficient matrix D of the residual block, wherein the decomposition expression is as follows:

finishing the decomposition;

(7) a third multi-channel filter bank longitudinal decomposition step: using a third multi-channel filter bankAnd longitudinally decomposing the transverse decomposition coefficient matrix Y to obtain a decomposition coefficient matrix D of the residual block, wherein the decomposition expression is as follows:

finishing the decomposition;

Wherein:

[d_4，id_5，id_6，id_7，i]^T＝F₂[y_2，iy_3，iy_4，iy_5，iy_6，iy_7，iy_8，iy_9，i]^T，

[d_8，id_9，id_10，id_11，i]^T＝F₂[y_6，iy_7，iy_8，iy_9，iy_10，iy_11，iy_12，iy_13，i]^T，

and finishing the decomposition.

2. A method for reconstructing a prediction residual block for video coding using a matrix of decomposition coefficients, comprising:

wherein,

α＝0.1225π；

after reconstruction is completed, performing the step (5);

wherein,

R₁＝SR₀PSP，

T₁＝ST₀PSP，

α＝0.4362π，β＝0.38π；

after reconstruction is completed, performing the step (5);

Longitudinally reconstructing the decomposition coefficient matrix D to obtain a transverse decomposition coefficient matrix Y, wherein the process comprises the following steps:

for ith row vector J of decomposition coefficient matrix D_i＝[d_0，i d_1，i d_2，i d_3，i d_4，i d_5，i d_6，i d_7，id_8，i d_9，i d_10，i d_11，i d_12，i d_13，i d_14，i d_15，i]^TReconstructing to obtain the ith column vector L of the transverse decomposition coefficient matrix Y_i＝[y_0，i y_1，i y_2，i y_3，i y_4，iy_5，i y_{7， i} y_8，i y_9，i y_10，i y_11，i y_12，i y_13，i y_14，i y_15，i]^TThe calculation expression is as follows:

y_1，i＝d_0，i×h₀₃+d_0，i×h₀₇+d_1，i×h₁₃-d_1，i×h₁₇+d_2，i×g₀₃+d_2，i ×g₀₇+d_3，i×g₁₃-d_3，i×g₁₇

y_3，i＝d_0，i×h₀₅+d_4，i×h₀₁+d_1，i ×h₁₅+d_5，i×h₁₁+d_2，i×g₀₅+d_6，i×g₀₁+d_3，i×g₁₅+d_7，i×g₁₁

y_9，i＝d_4，i×h₀₇+d_8，i×h₀₃+d_5，i×h₁₇+d_9，i×h₁₃+d_6，i×g₀₇+dx×g₀₃+d_7，i×g₁₇+d_11，i×g₁₃

wherein:

Y＝[L₀ L₁…L_k-1]wherein k is determined by the dimension of the long side of the decomposition coefficient matrix D;

after reconstruction is completed, performing the step (5);

finishing reconstruction;

(7) a third multi-channel filter bank transverse reconstruction step: using a third multi-channel filter bankAnd transversely reconstructing the transverse decomposition coefficient matrix Y to obtain a residual block X, wherein the reconstruction expression is as follows:

finishing reconstruction;

Performing transverse reconstruction on the transverse decomposition coefficient matrix Y to obtain a residual block X, wherein the process is as follows:

for the ith row vector M of the transverse decomposition coefficient matrix Y_i＝[y_i，0 y_i，1 y_i，2 y_i，3 y_i，4 y_i，5y_i，6 y_i，7 y_i，8 y_i，9 y_i，10 y_i，11 y_i，12 y_i，13 y_i，14 y_i，15]Rebuilding to obtain ith row vector N of residual block X_i＝[x_i，0 x_i，1 x_i，2 x_i，3 x_i，4 x_i，5 x_i，6 x_i，7 x_i，8 x_i，9 x_i，10 x_i，11 x_i，12 x_i，13 x_i，14 x_i，15]The calculation expression is as follows:

wherein k is determined by the broadside size of the transverse decomposition coefficient matrix Y;

and finishing the reconstruction.