CN105872543B - A kind of fast schema selection method in MVD viewpoints graduated encoding - Google Patents

Abstract

The invention discloses a fast mode selection method in MVD viewpoint scalable coding. Basic layer coding, texture video T2 adopts the full traversal coding mode selection method, texture video T1 simplifies candidate coding modes according to the texture complexity of the current MB, spatially adjacent MBs, and inter-view corresponding MBs; texture video T2 coding further considers The motion characteristics of the current MB. Enhancement layer encoding, depth video D2 uses the depth boundary distribution of the current MB and the encoding mode of the MB corresponding to the texture video T2, depth video D0/D1 uses the CBP value of the current MB in SKIP or Inter16x16 mode and the encoding mode corresponding to the depth video D2 to simplify Encode candidate patterns. When encoding the texture video of the base layer and the depth video of the enhancement layer, the present invention uses information such as the correlation between viewpoints, the complexity of the macroblock itself, and the motion situation to terminate the selection process of the encoding mode in advance, reduce the time complexity of encoding, and solve the problem of It solves the problem that the encoding mode selection time overhead is too large in the MVD encoding process.

Description

A Fast Mode Selection Method in MVD Viewpoint Scalable Coding

technical field

The invention relates to a fast mode selection method in viewpoint scalable coding, in particular to a fast mode selection method in MVD viewpoint scalable coding, and belongs to the technical field of video coding.

Background technique

The three-dimensional stereoscopic television (3DTV) system uses the principle of recognizing the distance of an object by observing the angle difference of the object with two eyes, and transmits the encoded 3D video to the left and right eyes of the person, so that the audience can experience a strong stereoscopic effect when watching TV programs. sense. The 3D video in multi-view video plus depth (Multi-view Video plus Depth, MVD) format combines the characteristics of multi-view video (Multi View Video, MVV) and texture video plus depth (Video plus Depth, V+D). Occupy an important position in the current 3DTV system.

The depth video is composed of successive depth frames, and the depth pixel value of each point in the depth frame represents the distance between the point in the shooting scene and the camera, and its value range is [0-255]. When the depth map is expressed in the YUV color space, there is only the brightness component Y, but no chroma components U and V, so the depth map can be regarded as a 256-color grayscale map. In the H.264/AVC video coding standard, the coding technology based on the macro block (Macro Block, MB) is adopted. The MB size is 16×16, and it can be divided into 4 sub-macro blocks. Each sub-macro block The size is 8x8. For the coded block pattern (Coded Block Pattern, CBP) value of MB, it indicates the residual coding method of MB luma and chrominance, using 6 binary numbers to represent: b ₅ b ₄ b ₃ b ₂ b ₁ b ₀ , where b ₃ b ₂ b ₁ b ₀ respectively represent 4 luma blocks, b ₅ b ₄ represent two chrominance blocks, if the residual coefficient of the sub-macroblock has not been encoded and transmitted, then this position is 0, otherwise it is set to 1, and each binary See Figure 1 for the distribution of bits. Since depth video has no chroma component, b ₅ b ₄ =00. Therefore, the CBP value range of the depth video MB is 0≤CBP≤15.

In MVD video coding, the time overhead is mainly concentrated in the MB coding mode selection process, accounting for more than 60% of the total coding time, which severely restricts the application of real-time video transmission. In the coding mode selection process, there are two types of MB coding modes: intra coding mode and inter coding mode. The intra-frame coding mode can be specifically divided into Intra16x16, Intra8x8, and Intra4x4; the inter-frame coding mode can be specifically divided into SKIP, Inter16x16, Inter16x8, Inter8x16, and Inter8x8. Among them, Inter8x8 includes four division methods: Inter8x8, Inter8x4, Inter4x8 and Inter4x4.

In the joint multi-view video coding reference software (Joint Multi-view Video Model, JMVM), the MB coding mode is selected according to the rate-distortion optimization criterion (Rate Distortion Optimization, RDO). After the Lagrangian parameter λ _mode and the quantization parameter Q are determined, the macroblock coding method that minimizes the rate-distortion cost function J _m is selected as the optimal coding method for the current macroblock S. The rate-distortion cost function is as follows:

J _m (I _m |Q,λ _mode )＝λ _mode ×R(I _m |Q)+D(I _m |Q)m∈Mode (1)

Among them, _Im is the coding mode of the macroblock, R is the bit rate of the coded bit stream, Mode is the coding mode of the candidate macroblock, and D is the distortion degree, which is determined by the original pixels and reconstructed pixels of the MB. In JMVM, it has There are two calculation methods, namely:

Wherein, (x, y) represents the coordinates of the pixel in the macroblock, p(x, y) is the original pixel value of the macroblock, and p'(x, y) is the reconstructed pixel value of the macroblock.

The final encoding method I of MB is obtained by the following relational expression:

In JMVM, MB mode selection follows the above-mentioned rate-distortion optimization criterion, calculates the rate-distortion cost value of each macroblock coding mode, and then selects the MB coding mode. This method can ensure that the MB gets the optimal coding mode, but the selection method of full traversal greatly increases the time complexity of MB coding. Therefore, it is necessary to reduce the time complexity of MVD in scalable coding. The core problem is to quickly select the coding mode of MB under the condition that the rate-distortion is acceptable.

Contents of the invention

Object of the present invention The object of the present invention is to provide a method for fast mode selection in MVD viewpoint scalable coding, aiming at the defects existing in the prior art.

To achieve the above object, the present invention adopts the following technical solutions:

A fast mode selection method in MVD viewpoint scalable coding, encoding viewpoint V0, V1 and V2; each viewpoint is composed of texture video and corresponding depth video; texture video of the viewpoint V0, V1 and V2 T0, T1, and T2 are used as the base layer, and the depth videos D0, D1, and D2 of the viewpoints V0, V1, and V2 are used as the enhancement layer, and the encoding sequence is the texture video T0 of the viewpoint V0-the texture video T2 of the viewpoint V2-the texture of the viewpoint V1 Video T1-depth video D2 of viewpoint V2-depth video D1 of viewpoint V1-depth video D0 of viewpoint V0; the texture video and depth video of each viewpoint are divided into MBs of equal size; the following specific steps are included:

Step 1. Base layer coding: use the full traversal coding mode selection process to select the best coding mode for each MB of the texture video T0 of the viewpoint V0; use the texture video T2 of the viewpoint V2 to use the texture of the current MB The complexity and its correlation with the coding mode of the corresponding MB in the texture video T0 of the viewpoint V0 simplify the coding candidate mode set, and select the coding mode of the current MB; the texture video T1 of the viewpoint V1 uses the motion speed of the current MB and its Simplify the encoding candidate mode set for the correlation between the texture video T0 of the viewpoint V0 and the coding mode of the corresponding MB in the T2 of the viewpoint V2, and select the coding mode of the current MB;

Step 2. Enhancement layer coding: the depth video D2 of the viewpoint V2, using the depth boundary distribution in the current MB and its correlation with the coding mode of the texture video T2 of the viewpoint V2, simplify the coding candidate mode, and select the coding of the current MB mode: for the depth video D1 of the viewpoint V1, use the size of the CBP value of the current MB in SKIP or Inter16x16 mode and its correlation with the corresponding MB coding mode in the depth video D2 of the viewpoint V2 to simplify the encoding candidate mode, and select the current MB encoding mode; the encoding mode selection method of the depth video D0 of the viewpoint V0 is the same as the encoding mode selection method of the depth video D1 of the viewpoint V1.

The method for simplifying the encoding candidate mode set of the current MB for the texture video T2 of the viewpoint V2 described in step 1 includes the following steps:

Step 1-A1: Set the initial values of the lower limit threshold Th ₁ and the upper limit threshold Th ₂ respectively; determine whether the encoding mode of the corresponding MB in the left MB, the upper MB, and the texture video T0 of the viewpoint V0 of the current MB is SKIP or Inter16x16 encoding mode; If it is go to step 1-A2; otherwise go to step 1-A3;

Step 1-A2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-A8;

Step 1-A3: Determine whether the texture complexity MAD of the current MB satisfies MAD<Th ₁ ; if yes, go to step 1-A4; otherwise, go to step 1-A5;

The calculation method of the texture complexity of the current MB is:

Among them, p(i, j) represents the pixel value at the coordinate (i, j) in the current MB, and avg represents the pixel average value of the current MB.

Step 1-A4: The candidate mode of the current MB is set to SKIP, Inter16x16 and Intra16x16 coding mode, and Th ₁ is updated as:

Go to step 1-A8;

Step 1-A5: Determine whether the texture complexity MAD of the current MB satisfies MAD>Th ₂ ; if yes, go to step 1-A6; otherwise, go to step 1-A7;

Step 1-A6: The candidate mode of the current MB is set to Inter8x8, Inter8x4, Inter4x8, Inter4x4, Intra8x8 and Intra4x4, and update Th ₂ as:

Go to step 1-A8;

Step 1-A7: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and all intra-frame coding modes;

Step 1-A8: End.

The method for simplifying the coding candidate mode of the current MB for the texture video T1 of the viewpoint V1 described in step 1 includes the following steps:

Step 1-B1: Determine whether the encoding mode of the left MB, the upper MB of the current MB, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2 is SKIP or Inter16x16. Step 1-B2; otherwise go to step 1-B3;

Step 1-B2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-B8;

Step 1-B3: According to the average value of the motion vectors of the left MB, the upper MB of the current MB, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2, the current MB is divided into motion There are three types of situations: slow, moderate and vigorous exercise:

In the formula, D1 and D2 are the _first and _second division thresholds respectively; D is the motion vector distance of the current MB, and its calculation method is:

D＝|x|+|y| (5)

Wherein, i=1...6, x _i and y _i are the left MB of the current MB of the reference macroblock, the upper MB, the corresponding MB in the forward reference frame of the texture video T1 of the viewpoint V1, the viewpoint V1 The horizontal and vertical motion vector values corresponding to the MB in the backward reference frame of the texture video T1, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2;

If the current MB is of slow movement type, turn to step 1-B4; if the current MB is of moderate movement type, turn to step 1-B5; if the current MB is of vigorous movement type, turn to step 1-B7;

Step 1-B4: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-B7;

Step 1-B5: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and Inter8x8 encoding modes; turn to step 1-B7;

Step 1-B6: the candidate mode of the current MB is set to all encoding modes; turn to step 1-B7;

Step 1-B7: End.

The method for simplifying the encoding candidate mode of the current MB for the depth video D2 of the viewpoint V2 described in step 2 includes the following steps:

Step 2-A1: Determine whether the encoding mode of the MB corresponding to the texture video T2 of the viewpoint V2 is SKIP or Inter16x16; if so, turn to step 1-A2; otherwise, turn to step 1-A3;

Step 2-A2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 2-A6;

Step 2-A3: Calculate the number Num of sub-macroblocks containing depth boundaries in the current MB; if the number Num of sub-macroblocks is

0, turn to step 2-A2; if the number Num of sub-macroblocks is 1, turn to step 2-A4; otherwise, turn to step 2-A5;

Step 2-A4: The candidate mode of the current MB is set to SKIP, Inter16x16, Inter16x16, Inter16x8 and Inter8x16 encoding modes; turn to step 1-B8; turn to step 2-A6;

Step 2-A5: The candidate mode of the current MB is set to Inter16x8, Inter8x16, Inter8x8, Inter8x4, Inter4x8 and Inter4x4 coding mode; turn to step 2-A6;

Step 2-A6: End.

The method for simplifying the coding candidate mode of the current MB for the depth video D1 of the viewpoint V1 described in step 2 includes the following steps:

Step 2-B1: Determine whether the encoding mode of the MB corresponding to the current MB in the depth video D2 of the viewpoint V2 is SKIP or Inter16x16, and the CBP of the MB corresponding to the current MB in the depth video D2 of the viewpoint V2 is CBP=0; if it is, go to step 1 -B2; otherwise turn to step 1-B3;

Step 2-B2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 2-B7;

Step 2-B3: In the SKIP and Inter16x16 modes, the current MB first selects a pre-selected mode according to the rate-distortion optimization criterion, if CBP=0, turn to step 2-B4; if CBP=7, 11, 13, 14, 15, turn to Step 2-B5; otherwise turn to step 2-B6;

Step 2-B4: The inter prediction mode of the current MB selects the prediction mode selected in step 2-B3; turn to step 2-B7;

Step 2-B5: The candidate mode of the current MB is set to Inter8x8, Inter8x4, Inter4x8 and Inter4x4 encoding modes; turn to step 2-B7;

Step 2-B6: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and Inter8x8 encoding modes; turn to step 2-B7;

Step 2-B7: End.

In the step 1, the method of determining the current MB of the texture video T2 of the viewpoint V2 is in the corresponding MB of the texture video T1 of the viewpoint V1 and determining the texture of the texture video T1 of the viewpoint V1 and the texture of the viewpoint V2 of the current MB of the texture video T0 of the V0 The method of corresponding MB of video T2 is the same, determined by the global disparity vector GDV, and its calculation method is as follows:

Among them, POC _curr , POC _ahead and POC _behind represent the POC value of the image sequence number on the time axis, that is, the display order of the image frames, GDV _ahead and GDV _behind represent the GDV values of the two key frames closest to the current encoding frame, respectively, It is calculated as follows:

Among them, img_l and img_r respectively represent the corresponding key frames in the left and right viewpoints, R is the number of pixels in a fixed area, x and y represent the horizontal and vertical position offsets of the left and right viewpoints respectively;

In the step 2, the method of determining the current MB of the depth video D2 of the viewpoint V2 in the corresponding MB of the depth video D1 of the viewpoint V1 and determining the depth of the current MB of the depth video D0 of V0 in the depth video D1 of the viewpoint V1 and the depth of the viewpoint V2 The method of the corresponding MB of the video D2 is the same as the method of determining the corresponding MB of the texture video T1 of the viewpoint V1 in the current MB of the texture video T2 of the viewpoint V2 in step 1.

Compared with the prior art, the present invention has the following obvious outstanding substantive features and significant technological progress:

When encoding the texture video of the base layer and the depth video of the enhancement layer, the present invention uses information such as the correlation between viewpoints, the complexity of the macroblock itself, and the motion characteristics to terminate the selection process of the encoding mode in advance, reduce the time complexity of encoding, and solve the problem of It solves the problem that the encoding mode selection time overhead is too large in the MVD encoding process.

Description of drawings:

Fig. 1 is each binary bit distribution figure in the macroblock CBP value;

Fig. 2 is the inter-view coding prediction structure of MVD video scalable coding;

Fig. 3 is a flowchart of the present invention;

Fig. 4 is the encoding mode selection process of the texture video T1 in the base layer viewpoint V1 in the present invention;

Fig. 5 is the encoding mode selection process of the texture video T2 in the base layer viewpoint V2 in the present invention;

Fig. 6 is the encoding mode selection process of the depth video D2 in the enhancement layer viewpoint V2 in the present invention;

FIG. 7 shows the encoding mode selection process of the depth video D1 in the enhancement layer view V1 in the present invention.

Detailed ways:

Preferred embodiments of the present invention are described in detail as follows in conjunction with accompanying drawings:

Example 1:

Referring to Figures 1 to 7, a fast mode selection method in MVD viewpoint scalable coding encodes viewpoints V0, V1 and V2; each viewpoint is composed of texture video and corresponding depth video; the viewpoint V0 , the texture videos T0, T1 and T2 of V1 and V2 are used as the base layer, the depth videos D0, D1 and D2 of the viewpoint V0, V1 and V2 are used as the enhancement layer, and the encoding sequence is the texture video T0 of the viewpoint V0-the texture of the viewpoint V2 Video T2-texture video T1-viewpoint V1-depth video D2-viewpoint V1-depth video D1-viewpoint V0 depth video D0; including the following specific steps:

Step 1. Basic layer coding: the texture video T0 of the viewpoint V0 is divided into more than one MB, and the best coding mode for each MB is selected using the full traversal coding mode selection process; the texture video T2 of the viewpoint V2 is used The texture complexity of the current MB and its correlation with the encoding mode of the corresponding MB in the texture video T0 of the viewpoint V0 simplify the encoding candidate mode set, and select the encoding mode of the current MB; the texture video T1 of the viewpoint V1 uses the motion of the current MB Simplify the encoding candidate mode set and select the encoding mode of the current MB based on the speed and slowness and its correlation with the encoding mode of the corresponding macroblock in the texture video T0 of the viewpoint V0 and the corresponding macroblock in the T2 of the viewpoint V2;

Step 2. Enhancement layer coding: the depth video D2 of the viewpoint V2, using the depth boundary distribution in the current MB and its correlation with the coding mode of the texture video T2 of the viewpoint V2, simplify the coding candidate mode, and select the coding of the current MB mode: for the depth video D1 of the viewpoint V1, use the size of the CBP value of the current MB in SKIP or Inter16x16 mode and its correlation with the corresponding MB coding mode in the depth video D2 of the viewpoint V2 to simplify the encoding candidate mode, and select the current MB encoding mode; the encoding mode selection method of the depth video D0 of the viewpoint V0 is the same as the encoding mode selection method of the depth video D1 of the viewpoint V1.

The method for simplifying the encoding candidate mode set of the current MB for the texture video T2 of the viewpoint V2 described in step 1 includes the following steps:

Step 1-A1: Set the initial values of the lower threshold Th ₁ and the upper threshold Th ₂ to 5 and 15 respectively; determine whether the coding mode of the corresponding MB in the left MB, upper MB of the current MB, and texture video T0 of the viewpoint V0 is SKIP or Inter16x16 encoding mode; if it is turn to step 1-A2; otherwise turn to step 1-A3;

Step 1-A2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-A8;

Step 1-A3: Determine whether the texture complexity MAD of the current MB satisfies MAD<Th ₁ ; if yes, turn to step 1-A4; otherwise turn to step 1-A5;

The calculation method of the texture complexity of the current MB is:

Among them, p(i, j) represents the pixel value at the coordinate (i, j) in the current MB, and avg represents the pixel average value of the current MB.

Step 1-A4: The candidate mode of the current MB is set to SKIP, Inter16x16 and Intra16x16 coding mode, and Th ₁ is updated as:

Go to step 1-A8;

Step 1-A5: Judging whether the texture complexity MAD of the current MB satisfies MAD>Th ₂ ; if so, go to step 1-A6; otherwise, go to step 1-A7;

Step 1-A6: The candidate mode of the current MB is set to Inter8x8, Inter8x4, Inter4x8, Inter4x4, Intra8x8 and Intra4x4, and update Th ₂ as:

Go to step 1-A8;

Step 1-A7: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and all intra-frame coding modes;

Step 1-A8: End.

The method for simplifying the coding candidate mode of the current MB for the texture video T1 of the viewpoint V1 described in step 1 includes the following steps:

Step 1-B1: Determine whether the encoding mode of the left MB, the upper MB of the current MB, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2 is SKIP or Inter16x16. Step 1-B2; otherwise go to step 1-B3;

Step 1-B2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-B8;

Step 1-B3: According to the average value of the motion vectors of the left MB, the upper MB of the current MB, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2, the current MB is divided into motion There are three types of situations: slow, moderate and vigorous exercise:

In the formula, D1 and D2 are the _first and _second division thresholds respectively; D is the distance of the motion vector of the current MB, and its calculation method is:

D＝|x|+|y| (5)

Wherein, i=1...6, x _i and y _i are the left MB of the current MB of the reference macroblock, the upper MB, the corresponding MB in the forward reference frame of the texture video T1 of the viewpoint V1, the viewpoint V1 The horizontal and vertical motion vector values corresponding to the MB in the backward reference frame of the texture video T1, the corresponding MB in the texture video T0 of the viewpoint V0, and the corresponding MB in the texture video T2 of the viewpoint V2;

If the current MB is of slow movement type, turn to step 1-B4; if the current MB is of moderate movement type, turn to step 1-B5; if the current MB is of vigorous movement type, turn to step 1-B7;

Step 1-B4: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 1-B7;

Step 1-B5: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and Inter8x8 encoding modes; turn to step 1-B7;

Step 1-B6: the candidate mode of the current MB is set to all encoding modes; turn to step 1-B7;

Step 1-B7: End.

The method for simplifying the encoding candidate mode of the current MB for the depth video D2 of the viewpoint V2 described in step 2 includes the following steps:

Step 2-A1: Determine whether the encoding mode of the MB corresponding to the texture video T2 of the viewpoint V2 is SKIP or Inter16x16; if so, turn to step 1-A2; otherwise, turn to step 1-A3;

Step 2-A2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 2-A6;

Step 2-A3: Calculate the number Num of sub-macroblocks containing depth boundaries in the current MB; if the number Num of sub-macroblocks is 0, turn to step 2-A2; if the number Num of sub-macroblocks is 1, turn to Step 2-A4; Otherwise, go to Step 2-A5;

Step 2-A4: The candidate mode of the current MB is set to SKIP, Inter16x16, Inter16x16, Inter16x8 and Inter8x16 encoding modes; turn to step 1-B8; turn to step 2-A6;

Step 2-A5: The candidate mode of the current MB is set to Inter16x8, Inter8x16, Inter8x8, Inter8x4, Inter4x8 and Inter4x4 coding mode; turn to step 2-A6;

Step 2-A6: End.

The method for simplifying the coding candidate mode of the current MB for the depth video D1 of the viewpoint V1 described in step 2 includes the following steps:

Step 2-B1: Determine whether the encoding mode of the corresponding MB of the current MB in the depth video D2 of the viewpoint V2 is SKIP or Inter16x16, and the CBP of the corresponding MB of the current MB in the depth video D2 of the viewpoint V2=0; if yes, go to step 1-B2; otherwise go to step 1-B3;

Step 2-B2: The candidate mode of the current MB is set to SKIP and Inter16x16 encoding mode; turn to step 2-B7;

Step 2-B3: In the SKIP and Inter16x16 modes, the current MB first selects a pre-selected mode according to the rate-distortion optimization criterion, if CBP=0, turn to step 2-B4; if CBP=7, 11, 13, 14, 15, turn to Step 2-B5; otherwise turn to step 2-B6;

Step 2-B4: The inter prediction mode of the current MB selects the prediction mode selected in step 2-B3; turn to step 2-B7;

Step 2-B5: The candidate mode of the current MB is set to Inter8x8, Inter8x4, Inter4x8 and Inter4x4 encoding modes; turn to step 2-B7;

Step 2-B6: The candidate mode of the current MB is set to Inter16x16, Inter16x8, Inter8x16 and Inter8x8 encoding modes; turn to step 2-B7;

Step 2-B7: End.

In the step 1, the method of determining the current MB of the texture video T2 of the viewpoint V2 is in the corresponding MB of the texture video T1 of the viewpoint V1 and determining the texture of the texture video T1 of the viewpoint V1 and the texture of the viewpoint V2 of the current MB of the texture video T0 of the V0 The method of corresponding MB of video T2 is the same, determined by the global disparity vector GDV, and its calculation method is as follows:

Among them, POC _curr , POC _ahead and POC _behind represent the POC value of the image sequence number on the time axis, that is, the display order of the image frames, GDV _ahead and GDV _behind represent the GDV values of the two key frames closest to the current encoding frame, respectively, It is calculated as follows;

Among them, img_l and img_r respectively represent the corresponding key frames in the left and right viewpoints, R is the number of pixels in a fixed area, x and y represent the horizontal and vertical position offsets of the left and right viewpoints respectively;

In the step 2, the method of determining the current MB of the depth video D2 of the viewpoint V2 in the corresponding MB of the depth video D1 of the viewpoint V1 and determining the depth of the current MB of the depth video D0 of V0 in the depth video D1 of the viewpoint V1 and the depth of the viewpoint V2 The method of the corresponding MB of the video D2 is the same as the method of determining the corresponding MB of the texture video T1 of the viewpoint V1 in the current MB of the texture video T2 of the viewpoint V2 in step 1.

The specific embodiments described herein are only examples to illustrate the spirit of the present invention. Those skilled in the technical field to which the present invention belongs may modify, supplement or replace in a similar manner to the described specific implementation cases, but they will not deviate from the spirit of the present invention or go beyond the scope defined in the appended claims.

Claims (6)

1. the fast schema selection method in a kind of MVD viewpoints graduated encoding, to vision point 0, V1 and V2 are encoded；It is special Sign is：Each viewpoint is made of texture video and corresponding deep video；The texture video of the vision point 0, V1 and V2 T0, T1 and T2 are as Primary layer, the vision point 0, and as enhancement layer, coded sequence is deep video D0, D1 and D2 of V1 and V2 The deep video D2- of the texture video T1- vision points 2 of the texture video T2- vision points 1 of the texture video T0- vision points 2 of vision point 0 The deep video D0 of the deep video D1- vision points 0 of vision point 1；It is characterized in that：Including step in detail below：

Step 1. Primary layer encodes：The texture video T0 of the vision point 0 is divided into 1 or more MB, uses the coding mould traversed entirely Formula selection process selecting goes out the forced coding pattern of each MB；The texture video T2 of the vision point 2 is multiple using the texture of current MB Miscellaneous degree and its coding mode correlation that MB is corresponded to the texture video T0 of vision point 0 simplify coding candidate pattern set, selection The coding mode of current MB；The texture video T1 of the vision point 1, using current MB movement speed situation and its with vision point 0 Texture video T0 and vision point 2 T2 in correspond to the coding mode correlation of MB and simplify coding candidate pattern set, selection is current The coding mode of MB；

Step 2. enhancement layer coding：The deep video D2 of the vision point 2, using in current MB depth boundary distribution situation and Coding mode correlation between its texture video T2 with vision point 2 simplifies coding candidate pattern, selects the coding mould of current MB Formula；The deep video D1 of the vision point 1, using CBP values of the current MB under SKIP or Inter16x16 patterns size and its The correlation that MB coding modes are corresponded to the deep video D2 of vision point 2 simplifies coding candidate pattern, therefrom selects current MB's Coding mode；The volume of the coding mode selection method of the deep video D0 of the vision point 0 and the deep video D1 of the vision point 1 Pattern selection method is identical.

2. the fast schema selection method in MVD viewpoints graduated encoding according to claim 1；It is characterized in that：Institute State vision point 2 texture video T2 simplify current MB coding candidate pattern set method, include the following steps：

Step 1-A1：Lower threshold Th is respectively set₁With upper limit threshold Th₂Initial value；Judge left MB, the upper MB of current MB and regards Whether the coding mode that MB is corresponded in the texture video T0 of point V0 is SKIP or Inter16x16 coding modes；It is walked if it is turning to Rapid 1-A2；Otherwise step 1-A3 is turned to；

Step 1-A2：The candidate pattern of current MB is set as SKIP and Inter16x16 coding modes；Turn to step 1-A8；

Step 1-A3：Judge whether the Texture complication MAD of current MB meets MAD < Th₁, if so, turning to step 1-A4；It is no Then turn to step 1-A5；

Currently the computational methods of the Texture complication of MB are：

Wherein, p (i, j) indicates that the pixel value at (i, j) coordinate, avg indicate the pixel average of current MB in current MB；

Step 1-A4：The candidate pattern of current MB is set as SKIP, Inter16x16 and Intra16x16 coding mode, and updates Th₁For：

Turn to step 1-A8；

Step 1-A5：Judge whether the Texture complication MAD of current MB meets MAD > Th₂；If so, turning to step 1-A6；It is no Then turn to step 1-A7；

Step 1-A6：The candidate pattern of current MB be set as Inter8x8, Inter8x4, Inter4x8, Inter4x4, Intra8x8 and Intra4x4, and update Th₂For：

Turn to step 1-A8；

Step 1-A7：The candidate pattern of current MB is set as compiling in Inter16x16, Inter16x8, Inter8x16 and whole frame Pattern；

Step 1-A8：Terminate.

3. the fast schema selection method in MVD viewpoints graduated encoding according to claim 1；It is characterized in that：Step The method that the texture video T1 of vision point 1 described in rapid 1 simplifies the coding candidate pattern of current MB, includes the following steps：

Step 1-B1：Judge to correspond to MB, the vision point 2 in the texture video T0 of the left MB of current MB, upper MB, the vision point 0 Texture video T2 in correspond to whether the coding mode of MB is SKIP or Inter16x16, if it is turn to step 1-B2；Otherwise Turn to step 1-B3；

Step 1-B2：The candidate pattern of current MB is set as SKIP and Inter16x16 coding modes；Turn to step 1-B7；

Step 1-B3：According to the left MB of current MB, upper MB, the vision point 0 texture video T0 in correspond to MB, the vision point 2 Texture video T2 in correspond to MB motion vector average value by current MB be divided into movement slowly, moderate activity and movement it is acute Strong three kinds of situation types：

D in formula₁And D₂Respectively first and second divide threshold value；D is the distance of the motion vector of current MB, and computational methods are：

D=&#124;x&#124;+&#124;y&#124; (5)

Wherein, i=1...6, x_iAnd y_iThe left MB of respectively current MB, upper MB, the vision point 1 texture video T1 in forward direction join Examine the texture video for corresponding to MB, the vision point 0 in the texture video T1 for corresponding to MB, the vision point 1 in frame in backward reference frame The horizontal and vertical movement vector value that MB is corresponded in the texture video T2 of MB and the vision point 2 is corresponded in T0；

If current MB is to move slow type, step 1-B4 is turned to；If current MB is moderate activity type, step 1- is turned to B5；If current MB is motion intense type, step 1-B7 is turned to；

Step 1-B4：The candidate pattern of current MB is set as SKIP and Inter16x16 coding modes；Turn to step 1-B7；

Step 1-B5：The candidate pattern of current MB is set as Inter16x16, and Inter16x8, Inter8x16 and Inter8x8 are compiled Pattern；Turn to step 1-B7；

Step 1-B6：The candidate pattern of current MB is set as all coding modes；Turn to step 1-B7；

Step 1-B7：Terminate.

4. the fast schema selection method in MVD viewpoints graduated encoding according to claim 1；It is characterized in that：Step The method that the deep video D2 of vision point 2 described in rapid 2 simplifies the coding candidate pattern of current MB, includes the following steps：

Step 2-A1：Judge current MB corresponded in the texture video T2 of vision point 2 MB coding mode whether be SKIP or Inter16x16；If it is steering step 2-A2；Otherwise step 2-A3 is turned to；

Step 2-A2：The candidate pattern of current MB is set as SKIP and Inter16x16 coding modes；Turn to step 2-A6；

Step 2-A3：Calculate the number Num of the sub-macroblock containing depth boundary in current MB；If the number Num of sub-macroblock is 0, turn to step 2-A2；If the number Num of sub-macroblock is 1, step 2-A4 is turned to；Otherwise, step 2-A5 is turned to；

Step 2-A4：The candidate pattern of current MB be set as SKIP, Inter16x16, Inter16x16, Inter16x8 and Inter8x16 coding modes；Turn to step 2-A6；

Step 2-A5：The candidate pattern of current MB be set as Inter16x8, Inter8x16, Inter8x8, Inter8x4, Inter4x8 and with Inter4x4 coding modes；Turn to step 2-A6；

Step 2-A6：Terminate.

5. the fast schema selection method in MVD viewpoints graduated encoding according to claim 1；It is characterized in that：On State vision point 1 described in step 2 deep video D1 simplify current MB coding candidate pattern method, include the following steps：

Step 2-B1：Judge current MB corresponded in the deep video D2 of vision point 2 MB coding mode whether be SKIP or Inter16x16, and current MB corresponds to the CBP=0 of MB in the deep video D2 of vision point 2；If it is steering step 2-B2； Otherwise step 2-B3 is turned to；

Step 2-B2：The candidate pattern of current MB is set as SKIP and Inter16x16 coding modes；Turn to step 2-B7；

Step 2-B3：Current MB selects one in advance according to rate-distortion optimization criterion first in SKIP and Inter16x16 patterns Lectotype turns to step 2-B4 if CBP=0；If CBP=7,11,13,14,15, turn to step 2-B5；Otherwise step is turned to Rapid 2-B6；

Step 2-B4：The prediction mode selected in the inter-frame forecast mode selection step 2-B3 of current MB；Turn to step 2-B7；

Step 2-B5：The candidate pattern of current MB is set as Inter8x8, Inter8x4, Inter4x8 and Inter4x4 coding mould Formula；Turn to step 2-B7；

Step 2-B6：The candidate pattern of current MB is set as Inter16x16, Inter16x8, Inter8x16 and Inter8x8 Coding mode；Turn to step 2-B7；

Step 2-B7：Terminate.

6. the fast schema selection method in MVD viewpoints graduated encoding according to claim 1；It is characterized in that：Institute State in step 1, determine the current MB of the texture video T2 of vision point 2 in the method for the correspondence MB of the texture video T1 of vision point 1 and Determine the current MB of the texture video T0 of V0 the texture video T1 of vision point 1 and the texture video T2 of vision point 2 correspondence MB Method is identical, determines that calculation is as follows by global disparity vector GDV：

Wherein, POC_curr、POC_aheadAnd POC_behindIndicate image sequence POC values on a timeline, the i.e. display of picture frame Sequentially, GDV_aheadAnd GDV_behindThe GDV values of two key frames nearest from current encoded frame are indicated respectively, and calculation is such as Under：

Wherein, img_l, img_r indicate that corresponding key frame in left and right viewpoint, R are the pixel of one piece of fixed area respectively Number, x and y indicate left and right viewpoint horizontal and vertical position offset respectively；

In the step 2, determine the current MB of the deep video D2 of vision point 2 the deep video D1 of vision point 1 correspondence MB Pairs of the current MB of the deep video D0 of method and determining V0 in the deep video D1 of vision point 1 and the deep video D2 of vision point 2 Answer the method for MB with the current MB for the texture video T2 that vision point 2 is determined in step 1 vision point 1 texture video T1 it is corresponding The method of MB is identical.