CN103442226B - The multiple views color video fast encoding method of distortion just can be perceived based on binocular - Google Patents

The multiple views color video fast encoding method of distortion just can be perceived based on binocular Download PDF

Info

Publication number
CN103442226B
CN103442226B CN201310325370.8A CN201310325370A CN103442226B CN 103442226 B CN103442226 B CN 103442226B CN 201310325370 A CN201310325370 A CN 201310325370A CN 103442226 B CN103442226 B CN 103442226B
Authority
CN
China
Prior art keywords
macro
block
binocular
macro block
value
Prior art date
Application number
CN201310325370.8A
Other languages
Chinese (zh)
Other versions
CN103442226A (en
Inventor
蒋刚毅
朱亚培
郁梅
邵枫
彭宗举
陈芬
王晓东
李福翠
Original Assignee
宁波大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 宁波大学 filed Critical 宁波大学
Priority to CN201310325370.8A priority Critical patent/CN103442226B/en
Publication of CN103442226A publication Critical patent/CN103442226A/en
Application granted granted Critical
Publication of CN103442226B publication Critical patent/CN103442226B/en

Links

Abstract

The invention discloses a kind of multiple views color video fast encoding method that just can perceive distortion based on binocular, first with the parallax information of left viewpoint video and right viewpoint video, it determines that the binocular of each macro block in the non-borderline region in the right visual point image of every frame in right viewpoint video just can perceive distortion value, secondly the size that just can perceive distortion value according to binocular terminates Macroblock Mode Selection in advance, this fast encoding method is not on the basis of causing distortion performance to decline, the code efficiency of multiple views color video can be effectively improved, the scramble time saved is up to 66.48% to 71.90%, the average saving scramble time 68.46%.

Description

The multiple views color video fast encoding method of distortion just can be perceived based on binocular

Technical field

The present invention relates to the processing method of a kind of multiple views colour-video signal, especially relate to one and just may be used based on binocular Perceive the multiple views color video fast encoding method of distortion.

Background technology

Three-dimensional television and free view-point TV are widely used multiple views color video and carry out scene description.Multiple views colour regards Frequency includes the colour of multiple viewpoint, by carrying out virtual viewpoint rendering at display end after encoding, transmit, decoding.Color at multiple views In color plus depth video, multiple views color video encoding is studied the most widely, and the encoding platform being the most more suitable for has Associating multi-view point video model and associating multiple view video coding.But owing to the research of the characteristic about human eye vision is sent out Zhan Zhong, therefore utilizes the perception characteristic of human visual system to need further to be studied in the compression of multiple views color video.

At present, in many characteristics of research human visual system, just can perceive distortion is the spy that vast researcher is inclined to One of levy.Just can perceive distortion table traveller on a long journey observe when seeing piece image change to image pixel can the visible threshold of perception, Depend on the brightness and contrast of image.Liu et al. utilizes border and the texture that just can perceive distortion model differentiation image Region.Recently, some for the research of human eye perception 3-D view and the visible threshold of video just prevailing, as the degree of depth just can be felt Examine distortion and binocular just can perceive distortion.What the degree of depth just can perceive distortion sign is the threshold that in deep video, minimum energy is perceived Value, therefore can be compressed the pixel being less than this threshold value in deep video further.It is beneficially base that binocular just can perceive distortion The model obtained with contrast masking sensitivity experiment, its image representing a viewpoint wherein or video are sheltered in brightness in eyes When distortion just can perceive distortion less than binocular, binocular will not aware the distortion of this image or video.

The data volume huge in order to compress multiple views colour plus depth video further, applicable encoding platform uses and entirely searches Rope model selection, determines the minimum rate distortion costs of macro block, so that it is determined that optimum prediction mode.Higher for full search pattern Computation complexity, research worker proposes some fast mode decision algorithms.Shen et al. proposes a kind of low-complexity mode Selection algorithm, including four kinds of mode selection techniques effectively, judges SKIP pattern in advance, and self adaptation terminates in advance, quick mode Size selects and selectable inner frame coding method, and this method has saved the scramble time of multiple views color video effectively, Almost identical with full search model selection coding result can be kept simultaneously.Zeng et al. utilizes quantization step and rate distortion generation Relation between valency, as threshold value, utilizes the motion vector of the motion vector computation current block of adjacent block.Above-mentioned method can Encoder complexity is effectively saved on the premise of not reducing coding quality, but at the cataloged procedure of multiple views color video In, the visual characteristic of human eye can not fully utilize in these methods, about perceived direction multiple views color video still There is a lot of research spaces.

Summary of the invention

The technical problem to be solved is to provide a kind of multiple views colour that just can perceive distortion based on binocular and regards Frequently fast encoding method, it can efficiently reduce multiple views color video on the basis of maintaining reconstruction viewpoint video performance Scramble time.

The present invention solves the technical scheme that above-mentioned technical problem used: a kind of just can to perceive distortion based on binocular many Viewpoint color video fast encoding method, it is characterised in that comprise the following steps:

1. the left viewpoint video of multiple views color video is designated as { CL(k) }, the right viewpoint of multiple views color video is regarded Frequency is designated as { CR(k) }, wherein, CLK () represents { CL(k) } in kth frame left view dot image, CRK () represents { CR(k) } in kth The right visual point image of frame, 1≤k≤K, K represent the frame number of the image comprised in left viewpoint video and right viewpoint video;

2. by { CR(k) } in the right visual point image of every frame be divided into borderline region and non-borderline region, wherein, borderline region It is made up of the first row macro block in right visual point image, last column macro block, first row macro block and last string macro block;Then calculate {CR(k) } in the right visual point image of every frame in non-borderline region in the binocular of each macro block just can perceive distortion value, it is assumed that {CR(k) } in kth frame right visual point image CRK in (), coordinate position is that (i, macro block j) belongs to { CR(k) } in the kth frame right side regard Dot image CRK the non-borderline region in (), then by { CR(k) } in kth frame right visual point image CRIn (k) coordinate position be (i, j) The binocular of macro block just can perceive distortion value and be designated asWherein, 2≤i≤W/16-1,2≤j≤H/16-1, W represent {CL(k) } in every frame left view dot image and { CR(k) } in the width of the right visual point image of every frame, H represents { CL(k) } in every Frame left view dot image and { CR(k) } in the height of the right visual point image of every frame;

3., on multiple view video coding Knowledge Verification Model JMVC, use HBP forecast coding structure to { CL(k) } in every frame Each macro block in left view dot image and { CR(k) } in the right visual point image of every frame in each macro block encode, coding During select the process of optimum macro-block coding pattern to be for each macro block:

3.-1, macro block the most to be encoded is defined as current macro;

3. it is-2, { C when current macroL(k) } in macro block time, encoder use H.264 mode selection processes search SKIP、Inter16×16、Inter16×8、Inter8×16、Inter8×8、Inter8×8Frext、Intra16×16、 Intra8 × 8 and Intra4 × 4 macro-block coding pattern, the macro block that elimination factor distortion cost is minimum from these macro-block coding patterns Coding mode encodes as the optimum macro-block coding pattern of current macro;

When current macro is { CR(k) } in macro block time, it is judged that current macro belongs to borderline region and still falls within non-border Region, if current macro belongs to borderline region, then encoder uses mode selection processes search SKIP, Inter16 H.264 × 16, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 He Intra4 × 4 macro-block coding pattern, the macro-block coding pattern that elimination factor distortion cost is minimum from these macro-block coding patterns is made Optimum macro-block coding pattern for current macro encodes;If current macro belongs to non-borderline region, judge current the most again The binocular of macro block just can perceive the decision threshold whether distortion value is more than or equal to set, if it is, H.264 encoder uses Mode selection processes search SKIP, Inter16 × 16, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4 macro-block coding pattern, select from these macro-block coding patterns The macro-block coding pattern of rate distortion costs minimum encodes as the optimum macro-block coding pattern of current macro;Otherwise, coding Device uses mode selection processes H.264 only to search for SKIP and Inter16 × 16 macro-block coding pattern, from both macroblock codings The macro-block coding pattern that in pattern, elimination factor distortion cost is minimum encodes as the optimum macro-block coding pattern of current macro;

3.-3, using macro block to be encoded for the next one as current macro, it is then back to step and 3.-2 continues executing with, until {CL(k) } in every frame left view dot image in each macro block and { CR(k) } in the right visual point image of every frame in each macro block equal Complete coding.

Described step 2. in { CR(k) } in kth frame right visual point image CRIn (k) coordinate position be (i, macro block j) Binocular just can perceive distortion value Wherein, d represents { CR(k) } in kth frame right visual point image CRK in (), coordinate position is (i, the parallax value of macro block j), bgL(i + d, j) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, all pixels in macro block j) The meansigma methods of the brightness value of point, bgL(i+d, j) ∈ [0,255], bgL(i × 16+d × 16+m, j × 16+n) represents { CL(k) } in kth frame left view dot image CLIn (k) coordinate position be (i × 16+d × 16+m, j × 16+n) the brightness value of pixel, 0≤m≤15,0≤n≤15, ehL(i+d j) represents { CL(k) } in Kth frame left view dot image CLK in (), coordinate position is (i+d, putting down of the edge intensity value computing of all pixels in macro block j) Average, bgL (i × 16+d × 16+m-3+h, j × 16+n-3+v) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is The brightness value of the pixel of (i × 16+d × 16+m-3+h, j × 16+n-3+v), GH(h v) represents that the horizontal Sobel of 5 × 5 calculates Sub-GHMiddle coordinate position be (h, v) element at place,GV(h v) represents the vertical Sobel operator G of 5 × 5VIn Coordinate position be (h, v) element at place,1≤h≤5,1≤v≤5, Alimt(bgL(i+d,j),ehL (i+d, j))=Alimit(bgL(i+d,j))+K(bgL(i+d,j))×ehL(i+d, j),K(bgL (i+d, j))=-10-6×(0.7×(bgL(i+d,j))2+32×bgL(i+d, j))+0.07, λ represent the right viewpoint noise shadow of control The parameter rung, nL(i+d j) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, macro block j) Noise amplitude.

The described parameter lambda value controlling right viewpoint influence of noise is 1.25.

Described { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, the noise width of macro block j) Value nL(i+d, j) value is 0.3.

The decision threshold value that 3. described step sets in-2 is as 5.

Compared with prior art, it is an advantage of the current invention that: first with left viewpoint video and the parallax of right viewpoint video Information determines that the binocular of each macro block in the non-borderline region in the right visual point image of every frame in right viewpoint video just can be perceived Distortion value, the size that secondly just can perceive distortion value according to binocular terminates Macroblock Mode Selection in advance, and this fast encoding method exists On the basis of not causing distortion performance to decline, it is possible to be effectively improved the code efficiency of multiple views color video, the volume of saving The code time, up to 66.48% to 71.90%, averagely saves the scramble time 68.46%.

Accompanying drawing explanation

Fig. 1 is the FB(flow block) of the inventive method;

Fig. 2 a is that the 1st frame of the 4th viewpoint correspondence color image sequence in " Book Arrival " test video sequence is color Color image;

Fig. 2 b is that the 1st frame of the 5th viewpoint correspondence color image sequence in " Book Arrival " test video sequence is color Color image;

Fig. 2 c is that the binocular of each macro block in the non-borderline region in the coloured image shown in Fig. 2 b just can perceive mistake The image that true value is formed after amplifying 20 times;

Fig. 3 a be multiple views color image sequence " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " right visual point image in non-borderline region in, just can perceive for having identical binocular All macro blocks of distortion value, select SKIP pattern to account for as the macroblock number of optimum macro-block coding pattern and have identical binocular and just may be used Perceive the percentage ratio of all macroblock number of distortion value;

Fig. 3 b be multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right visual point image in non-borderline region in, just may be used for having identical binocular Perceive all macro blocks of distortion value, select Inter16 × 16 pattern to account for as the macroblock number of optimum macro-block coding pattern and have Identical binocular just can perceive the percentage ratio of all macroblock number of distortion value;

Fig. 3 c be multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right visual point image in non-borderline region in, just may be used for having identical binocular Perceive all macro blocks of distortion value, select Inter16 × 8 pattern to account for as the macroblock number of optimum macro-block coding pattern and there is phase The percentage ratio of all macroblock number of distortion value just can be perceived with binocular;

Fig. 3 d be multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right visual point image in non-borderline region in, just may be used for having identical binocular Perceive all macro blocks of distortion value, select Inter8 × 16 pattern to account for as the macroblock number of optimum macro-block coding pattern and there is phase The percentage ratio of all macroblock number of distortion value just can be perceived with binocular;

Fig. 3 e be multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right visual point image in non-borderline region in, just may be used for having identical binocular Perceive all macro blocks of distortion value, select Inter8 × 8 pattern to account for as the macroblock number of optimum macro-block coding pattern and there is phase The percentage ratio of all macroblock number of distortion value just can be perceived with binocular;

Fig. 3 f be multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right visual point image in non-borderline region in, just may be used for having identical binocular Perceive all macro blocks of distortion value, select Others (to include Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 He Intra4 × 4) pattern accounts for as the macroblock number of optimum macro-block coding pattern has identical binocular and just can perceive the institute of distortion value There is the percentage ratio of macroblock number;

Fig. 4 a is obtained after being encoded by multiple views criteria check platform and the inventive method respectively for " Exit " cycle tests The encoding rate distortion performance comparison diagram of multiple views color image sequence;

Fig. 4 b is to obtain after " Vassar " cycle tests is encoded by multiple views criteria check platform and the inventive method respectively The encoding rate distortion performance comparison diagram of the multiple views color image sequence arrived;

Fig. 4 c is that " Champagne tower " cycle tests is respectively by multiple views criteria check platform and the inventive method The encoding rate distortion performance comparison diagram of the multiple views color image sequence obtained after coding;

Fig. 5 a is the 2nd viewpoint the 10th two field picture that " Exit " cycle tests uses that multiple views criteria check platform is rebuild;

Fig. 5 b is the 2nd viewpoint the 10th two field picture that " Exit " cycle tests uses that the fast encoding method of the present invention is rebuild;

Fig. 5 c is the 2nd viewpoint the 10th two field picture that " Vassar " cycle tests uses that multiple views criteria check platform is rebuild;

Fig. 5 d is the 2nd viewpoint the 10th frame figure that " Vassar " cycle tests uses that the fast encoding method of the present invention is rebuild Picture;

Fig. 6 is the relation schematic diagram between the coordinate position of the pixel in the coordinate position of macro block and macro block.

Detailed description of the invention

Below in conjunction with accompanying drawing embodiment, the present invention is described in further detail.

Owing to not having fully utilized human eye visual perception characteristic to compress in multiple views color video encoding, multiple views colour regards Substantial amounts of redundancy is there is, therefore to it is color to obtain reducing multiple views on the premise of not affecting human eye visual perception further in Pin The scramble time of color video, the present invention proposes a kind of multiple views color video fast coding that just can perceive distortion based on binocular Method, the macro block in right viewpoint video was asked for corresponding binocular before to multiple views color video encoding by it just can perceive mistake True value, to reduce the scramble time of coding multiple views color video.

A kind of multiple views color video fast encoding method that just can perceive distortion based on binocular that the present invention proposes, its stream Journey block diagram is as it is shown in figure 1, it comprises the following steps:

1. the left viewpoint video of multiple views color video is designated as { CL(k) }, the right viewpoint of multiple views color video is regarded Frequency is designated as { CR(k) }, wherein, CLK () represents { CL(k) } in kth frame left view dot image, CRK () represents { CR(k) } in kth The right visual point image of frame, 1≤k≤K, K represent the frame number of the image comprised in left viewpoint video and right viewpoint video.

2. exposed region is blocked owing to left view dot image and right visual point image existing, it is impossible to ask for left view dot image with right The parallax value of visual point image, and block the borderline region that exposed region is normally in left view dot image and right visual point image, Therefore { C is being calculatedR(k) } in the right visual point image of every frame in the binocular of macro block need to remove right viewpoint when just can perceive distortion value The borderline region (top line and bottom line macro block and Far Left and rightmost string macro block) of image, i.e. detailed process For: by { CR(k) } in the right visual point image of every frame be divided into borderline region and non-borderline region, wherein, borderline region is regarded by the right side The first row macro block in dot image, last column macro block, first row macro block and last string macro block are constituted, non-borderline region i.e. by All macro blocks that in right visual point image, abscissa is in the range of 2≤i≤W/16-1 and vertical coordinate is in the range of 2≤j≤H/16-1 Constitute;Then { C is calculatedR(k) } in the right visual point image of every frame in non-borderline region in the binocular of each macro block just can feel Examine distortion value, it is assumed that { CR(k) } in kth frame right visual point image CRK in (), coordinate position is that (i, macro block j) belongs to { CR(k)} In kth frame right visual point image CRK the non-borderline region in (), then by { CR(k) } in kth frame right visual point image CRK () sits Mark is set to that (i, the binocular of macro block j) just can be perceived distortion value and be designated as Wherein, 2≤i≤W/16-1,2≤j≤H/16-1, W represent { CL(k) } in every frame left view dot image and { CR(k) } in every frame right side regard The width of dot image, H represents { CL(k) } in every frame left view dot image and { CR(k) } in the height of the right visual point image of every frame, d Represent { CR(k) } in kth frame right visual point image CRK in (), coordinate position is that (i, the parallax value of macro block j) utilize existing Disparity estimation based on the difference of two squares He (the Sum of Square Differences, SSD) asks for { CR(k) } in The each macro block in non-borderline region in every right visual point image of frame and { CL(k) } in synchronization left view dot image in right Answer the parallax value of macro block, bgL(i+d j) represents { CL(k) } in kth frame left view dot image CLIn (k) coordinate position be (i+d, The meansigma methods of the brightness value of all pixels in macro block j), bgL(i+d, j) ∈ [0,255],bgL(i × 16+d × 16+m, j × 16+n) represents {CL(k) } in kth frame left view dot image CLK in (), coordinate position is the brightness of the pixel of (i × 16+d × 16+m, j × 16+n) Value, Fig. 6 gives the relation schematic diagram between the coordinate position of the pixel in the coordinate position of macro block and macro block, 0≤m≤15,0≤ N≤15, ehL(i+d j) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, all pixels in macro block j) The meansigma methods of the edge intensity value computing of point, bgL (i × 16+d × 16+m-3+h, j × 16+n-3+v) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is The brightness value of the pixel of (i × 16+d × 16+m-3+h, j × 16+n-3+v), GH(h v) represents that the horizontal Sobel of 5 × 5 calculates Sub-GHMiddle coordinate position be (h, v) element at place,GV(h v) represents the vertical Sobel operator G of 5 × 5V Middle coordinate position be (h, v) element at place,1≤h≤5,1≤v≤5, Alimt(bgL(i+d,j), ehL(i+d, j))=Alimit(bgL(i+d,j))+K(bgL(i+d,j))×ehL(i+d, j),K(bgL (i+d, j))=-10-6×(0.7×(bgL(i+d,j))2+32×bgL(i+d, j))+0.07, λ represent the right viewpoint noise shadow of control The parameter rung, λ value is 1.25, n in the present embodimentL(i+d j) represents { CL(k) } in kth frame left view dot image CL(k) Middle coordinate position be (i+d, the noise amplitude of macro block j), in the present embodiment, nL(i+d, j) value is 0.3.

Fig. 2 a gives the 1st of the 4th viewpoint correspondence color image sequence in " Book Arrival " test video sequence Color image frame (left view dot image);The 5th viewpoint correspondence that Fig. 2 b gives in " Book Arrival " test video sequence is color 1st color image frame (right visual point image) of color video sequence;Owing to H.264 cataloged procedure is carried out based on macroblock size, Therefore each macro block in the non-borderline region in the coloured image shown in Fig. 2 b is asked for binocular and just can be perceived mistake by the present invention True value, and just can perceive distortion value relatively due to each macro block in the non-borderline region in the coloured image shown in Fig. 2 b Little, human eye can not direct viewing, therefore for convenience of each by the non-borderline region in the coloured image shown in Fig. 2 b of viewing The distortion value of just can perceiving of individual macro block amplifies 20 times, and Fig. 2 c gives in the non-borderline region in the coloured image shown in Fig. 2 b The binocular of each macro block just can perceive the image formed after distortion value amplifies 20 times.

3., on multiple view video coding Knowledge Verification Model JMVC, use HBP forecast coding structure to { CL(k) } in every frame Each macro block in left view dot image and { CR(k) } in the right visual point image of every frame in each macro block encode, coding During select the process of optimum macro-block coding pattern to be for each macro block:

3.-1, macro block the most to be encoded is defined as current macro;

3. it is-2, { C when current macroL(k) } in macro block time, encoder use existing model selection H.264 Journey search SKIP, Inter16 × 16, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4 macro-block coding pattern, from these macro-block coding patterns, elimination factor distortion cost is minimum Macro-block coding pattern encode as the optimum macro-block coding pattern of current macro;

When current macro is { CR(k) } in macro block time, it is judged that current macro belongs to borderline region and still falls within non-border Region, if current macro belongs to borderline region, then encoder use existing H.264 mode selection processes search SKIP, Inter16×16、Inter16×8、Inter8×16、Inter8×8、Inter8×8Frext、Intra16×16、Intra8 × 8 and Intra4 × 4 macro-block coding pattern, the macroblock coding mould that elimination factor distortion cost is minimum from these macro-block coding patterns Formula encodes as the optimum macro-block coding pattern of current macro;If current macro belongs to non-borderline region, judge the most again The binocular of current macro just can perceive the decision threshold whether distortion value is more than or equal to set, if it is, encoder uses Existing mode selection processes search SKIP, Inter16 × 16 H.264, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4 macro-block coding pattern, from these macroblock codings The macro-block coding pattern that in pattern, elimination factor distortion cost is minimum encodes as the optimum macro-block coding pattern of current macro; Otherwise, encoder uses existing mode selection processes H.264 only to search for SKIP and Inter16 × 16 macro-block coding pattern, The macro-block coding pattern that elimination factor distortion cost is minimum from both macro-block coding patterns is as the optimum macro block of current macro Coding mode encodes;

3.-3, using macro block to be encoded for the next one as current macro, it is then back to step and 3.-2 continues executing with, until {CL(k) } in every frame left view dot image in each macro block and { CR(k) } in the right visual point image of every frame in each macro block equal Complete coding.

In the present embodiment, the decision threshold value that 3. step sets in-2 as 5, specifically taking of the decision threshold of this setting Value is obtained by experiment, and experimentation is as follows:

3.-2a, use multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence left viewpoint video in 46 frames, multiple views color video " Alt Moabit ", 46 frames in " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", the right viewpoint video of " Xmas9 " sequence, Calculate " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence successively The binocular of each macro block in the non-borderline region in the right visual point image of each frame in the right viewpoint video of row just can perceive mistake True value, just can perceive the minima of distortion value by the binocular of each macro block in the non-borderline region in right for each frame visual point image Be designated as BJND_min and BJND_max respectively with maximum, the most successively statistics " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right viewpoint video in 46 two field pictures in binocular Just can perceive the number of the macro block of each numerical value in the range of distortion value is equal to from BJND_min to BJND_max, with aggregate form It is designated as { NBJND_min,…,NBJND_max, wherein, NBJND_minRepresent binocular in 46 two field pictures in the right viewpoint video of a sequence Just can perceive the number of the distortion value macro block equal to BJND_min, NBJND_maxRepresent 46 frames in the right viewpoint video of a sequence In image, binocular just can perceive the number of the distortion value macro block equal to BJND_max.

3.-2b, use multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence left viewpoint video in 46 frames, multiple views color video " Alt Moabit ", 46 frames in " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", the right viewpoint video of " Xmas9 " sequence, On multiple view video coding Knowledge Verification Model JMVC, use HBP forecast coding structure to multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", the left viewpoint video of " Xmas9 " sequence With multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", The right viewpoint video of " Xmas9 " sequence encodes, in an encoding process obtain " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right viewpoint video in 46 two field pictures in institute There is the optimum macro-block coding pattern of macro block.

3.-2c, for multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " sequence right viewpoint video in 46 two field pictures in non-borderline region in all grand Block, wherein, when the binocular of macro block just can perceive some numerical value in the range of distortion value is equal to from BJND_min to BJND_max During BJND_x, corresponding set { NBJND_min,…,NBJND_maxA numerical value in }, is designated as NBJND_x, i.e. all just can perceive mistake The true value macro block number equal to some numerical value BJND_x in the range of from BJND_min to BJND_max is NBJND_x, wherein, This NBJND_xIn individual macro block, SKIP pattern is selected to be designated as N as the macroblock number of optimum macro-block coding patternSKIP, select Inter16 × 16 pattern is designated as N as the macroblock number of optimum macro-block coding patternInter16×16, select Inter16 × 8 pattern Macroblock number as optimum macro-block coding pattern is designated as NInter16×8, select Inter8 × 16 pattern as optimum macroblock coding The macroblock number of pattern is designated as NInter8×16, select Inter8 × 8 pattern to be designated as the macroblock number of optimum macro-block coding pattern NInter8×8, select Others (including Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4) pattern to make Macroblock number for optimum macro-block coding pattern is designated as NOthers

3.-2d, calculating NSKIP/NBJND_xValue, be designated as PSKIP, i.e. just can perceive distortion value equal to from BJND_ all The macro block number of some numerical value BJND_x in the range of min to BJND_max is NBJND_xIn, select SKIP pattern as optimum Macroblock number N of macro-block coding patternSKIPAccount for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 a give multiple views color image sequence " Alt Moabit ", " Balloons ", " Ballroom ", In non-borderline region in " Kendo ", " Race1 ", " Xmas3 ", " Xmas9 " respective 46 right visual point images of frame, for having Identical binocular just can perceive all macro blocks of distortion value, selects SKIP pattern to account for as the macroblock number of optimum code pattern and has Identical binocular just can perceive the percentage ratio of all macroblock number of distortion value, i.e. Fig. 3 a is multiple views color image sequence " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", " Race1 ", " Xmas3 ", the right visual point image of " Xmas9 " sequence Middle PSKIPSize variation schematic diagram, as shown in Figure 3 a, PSKIPValue be generally higher than 50%, shared percentage ratio is bigger, i.e. right Distortion value just can be perceived at BJND_min to BJND_ in non-borderline region in the right multi-view video image of multiple views color video All macro blocks in the range of max, in the existing mode selection processes H.264 encoded, select the percentage ratio of SKIP pattern relatively Greatly, so, just can perceive distortion value in the non-borderline region in multi-view video image right for multiple views color video at BJND_ In the range of min to BJND_max all macro blocks, the search of SKIP pattern will be carried out.

Calculate NInter16×16/NBJND_xValue, be designated as PInter16×16, i.e. all just can perceive distortion value equal to from The macro block number of some numerical value BJND_x in the range of BJND_min to BJND_max is NBJND_xIn, select Inter16 × 16 Pattern is as macroblock number N of optimum macro-block coding patternInter16×16Account for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 b give multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", In non-borderline region in " Race1 ", " Xmas3 ", the respective 46 right visual point images of frame of " Xmas9 " sequence, identical for having Binocular just can perceive all macro blocks of distortion value, selects Inter16 × 16 pattern to account for as the macroblock number of optimum code pattern There is the percentage ratio that identical binocular just can perceive all macroblock number of distortion value, as shown in Figure 3 b, PInter16×16Value general More than 20%, shared percentage ratio is bigger, i.e. in the non-borderline region in multi-view video image right for multiple views color video Just can perceive the distortion value all macro blocks in the range of BJND_min to BJND_max, select in the existing pattern H.264 encoded During selecting, the percentage ratio selecting SKIP pattern is relatively big, so, non-in multi-view video image right for multiple views color video Just can perceive the distortion value all macro blocks in the range of BJND_min to BJND_max in borderline region, Inter16 will be carried out The search of × 16 patterns.

Calculate NInter16×8/NBJND_xValue, be designated as PInter16×8, i.e. just can perceive distortion value equal to from BJND_ all The macro block number of some numerical value BJND_x in the range of min to BJND_max is NBJND_xIn, select Inter16 × 8 pattern to make Macroblock number N for optimum macro-block coding patternInter16×8Account for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 c give multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", In non-borderline region in " Race1 ", " Xmas3 ", the respective 46 right visual point images of frame of " Xmas9 " sequence, identical for having Binocular just can perceive all macro blocks of distortion value, selects Inter16 × 8 pattern to account for tool as the macroblock number of optimum code pattern There is identical binocular just can perceive the percentage ratio of all macroblock number of distortion value, as shown in Figure 3 c, PInter16×8Value typically shared Ratio is about about 5%, just can perceive the increase of distortion value along with binocular, and the patient distortion of human eye institute is the least, and macro block need to enter Step refining, shows as in figure 3 c, PInter16×8Increase the most therewith, be in particular in as 0≤BJND≤5, PInter16×8Keep At less numerical value, about 2%, this shows, when the binocular of macro block just can perceive distortion value less than certain threshold value, macro block mode Selection can no longer carry out Inter16 × 8 macro-block coding pattern search procedure, and when 6<BJND<n, during n>6, PInter16×8In increasing Big trend, about increases to 5%, and this shows, when the binocular of macro block just can perceive distortion value more than certain threshold value, macro block needs Inter16 × 8 to be carried out macro-block coding pattern search procedure, wherein, BJND represents that binocular just can perceive distortion value.

Calculate NInter8×16/NBJND_xValue, be designated as PInter8×16, i.e. just can perceive distortion value equal to from BJND_ all The macro block number of some numerical value BJND_x in the range of min to BJND_max is NBJND_xIn, select Inter16 × 8 pattern to make Macroblock number N for optimum macro-block coding patternInter8×16Account for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 d give multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", In non-borderline region in " Race1 ", " Xmas3 ", the respective 46 right visual point images of frame of " Xmas9 " sequence, identical for having Binocular just can perceive all macro blocks of distortion value, selects Inter8 × 16 pattern to account for tool as the macroblock number of optimum code pattern There is identical binocular just can perceive the percentage ratio of all macroblock number of distortion value, as shown in Figure 3 d, PInter8×16Value typically shared Ratio is about about 6%, just can perceive the increase of distortion value along with binocular, and the patient distortion of human eye institute is the least, and macro block need to enter Step refining, shows as, P in Fig. 3 dInter8×16Increase the most therewith;It is in particular in as 0≤BJND≤5, PInter8×16Keep At less numerical value, about 3%, this shows, when the binocular of macro block just can perceive distortion value less than certain threshold value, macro block mode Selection can no longer carry out Inter8 × 16 macro-block coding pattern search procedure, and when 6<BJND<n, during n>6, PInter8×16In increasing Big trend, about increases to 6%, and this shows, when the binocular of macro block just can perceive distortion value more than certain threshold value, macro block needs Inter8 × 16 to be carried out macro-block coding pattern search procedure.

Calculate NInter8×8/NBJND_xValue, be designated as PInter8×8, i.e. just can perceive distortion value equal to from BJND_ all The macro block number of some numerical value BJND_x in the range of min to BJND_max is NBJND_xIn, select Inter16 × 8 pattern to make Macroblock number N for optimum macro-block coding patternInter8×8Account for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 e give multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", In non-borderline region in " Race1 ", " Xmas3 ", the respective 46 right visual point images of frame of " Xmas9 " sequence, identical for having Binocular just can perceive all macro blocks of distortion value, selects Inter8 × 8 pattern to account for tool as the macroblock number of optimum code pattern There is identical binocular just can perceive the percentage ratio of all macroblock number of distortion value, as shown in Figure 3 e, PInter8×8Value typically shared Ratio is about about 4%, just can perceive the increase of distortion value along with binocular, and the patient distortion of human eye institute is the least, and macro block need to enter Step refining, shows as, P in Fig. 3 eInter8×8Increase the most therewith;It is in particular in as 0≤BJND≤5, PInter8×8It is maintained at Less numerical value, about 1.5%, this shows, when the binocular of macro block just can perceive distortion value less than certain threshold value, macro block mode Selection can no longer carry out Inter8 × 8 macro-block coding pattern search procedure, and when 6<BJND<n, during n>6, PInter8×8In increasing Big trend, about increases to 4%, and this shows, when the binocular of macro block just can perceive distortion value more than certain threshold value, macro block needs Inter8 × 8 to be carried out macro-block coding pattern search procedure.

Calculate NOthers/NBJND_xValue, be designated as POthers, i.e. all just can perceive distortion value equal to from BJND_min to The macro block number of some numerical value BJND_x in the range of BJND_max is NBJND_xIn, select Inter16 × 8 patterns as Macroblock number N of excellent macro-block coding patternOthersAccount for this NBJND_xThe percentage ratio of individual macro block.

Fig. 3 f give multiple views color video " Alt Moabit ", " Balloons ", " Ballroom ", " Kendo ", In non-borderline region in " Race1 ", " Xmas3 ", the respective 46 right visual point images of frame of " Xmas9 " sequence, identical for having Binocular just can perceive all macro blocks of distortion value, selects Others (to include Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4) pattern accounts for as the macroblock number of optimum code pattern has identical binocular and just can perceive the institute of distortion value There are the percentage ratio of macroblock number, as illustrated in figure 3f, POthersThe general proportion of value be about about 6%, along with binocular just can be felt Examining the increase of distortion value, the patient distortion of human eye institute is the least, and macro block need to refine further, shows as in Fig. 3 f, POthersAlso Increase therewith;It is in particular in as 0≤BJND≤5, POthersIt is maintained at less numerical value, about 5%, this shows, when macro block Binocular just can perceive distortion value less than certain threshold value time, Macroblock Mode Selection can no longer carry out Others (include Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4) macro-block coding pattern search procedure, and when 6<BJND<n, n>6 Time, POthersIn the trend increased, about increasing to is 8%, and this shows, when the binocular of macro block just can perceive distortion value more than certain During individual threshold value, macro block needs to carry out Others and (includes Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4) macro-block coding pattern search procedure.

To sum up analyzing and can obtaining decision threshold value is 5.

In order to verify effectiveness and the feasibility of the fast encoding method of the present invention, first have chosen MERL laboratory " Exit ", " Vassar " the 0th viewpoint and the 2nd viewpoint color video cycle tests and the " Champagne of Nogaya university of Japan Tower " the 39th viewpoint and the color video cycle tests of the 41st viewpoint, using these color videos as original color video sequence Row, the fast encoding method that the recycling present invention proposes to " Exit ", " Vassar " the 2nd viewpoint of these sequences and The color video of the 41st viewpoint of " Champagne tower " carries out the color video rebuild after fast coding obtains fast coding Sequence.

At this by the peak value noise of the color image sequence from the scramble time of color image sequence, encoder bit rate and reconstruction The performance of the inventive method is weighed than (Peak Signal to Noise Ratio, PSNR).

Table 1 is listed the original color video sequence of above-mentioned each sequence and is encoded by JMVC verification platform and the inventive method After the color image sequence time that (Basis QP=22) encodes under the same conditions and encoder bit rate and reconstruction quality pair Ratio.Can draw from the data listed by table 1, in terms of the color image sequence scramble time, the inventive method and multiple views standard Verification platform is compared and can be saved the scramble time 66.48%~71.90%, and the average saving scramble time 68.46%, original color regards Frequency sequence is by the color image sequence rebuild after the fast coding of the present invention and the coloured silk rebuild by JMVC verification plateau coding The PSNR of color video sequence compares and averagely reduces 0.037dB, and maximum is reduced to 0.04dB, and original color video sequence passes through this The bright code check after fast coding is average compared with the code check by JMVC verification plateau coding increases by 0.46%, maximum increase 1.27%, minimum increase-0.20%, almost it is negligible.

Fig. 4 a, Fig. 4 b and Fig. 4 c sets forth " Exit ", " Vassar " and " Champagne tower " cycle tests and divide The encoding rate distortion performance comparison diagram of the color image sequence obtained after not encoded by JMVC verification platform and the inventive method, The PSNR of the color image sequence that distortion performance contrast is obtained after including being encoded by JMVC verification platform and the inventive method With code check contrast used.From Fig. 4 a it can be seen that on abscissa direction, when the right visual point image of coding " Exit " sequence Code check identical time, the PSNR of color image sequence rebuild after being encoded by JMVC verification platform and the inventive method is protected substantially Hold consistent;On vertical coordinate direction, the color image sequence rebuild after being encoded by JMVC verification platform and the inventive method PSNR identical time, coding " Exit " sequence the code check used by right visual point image essentially identical.It is seen from fig. 4b that at horizontal stroke On coordinate direction, when encoding the code check of right visual point image of " Vassar " sequence and being identical, by JMVC verification platform and Ben Fa The PSNR of the color image sequence rebuild after bright method coding keeps consistent substantially;On vertical coordinate direction, when by JMVC school Test the PSNR of the color image sequence rebuild after platform and the inventive method encode identical time, the right side of coding " Vassar " sequence regards Code check used by dot image is essentially identical.From Fig. 4 c it can be seen that on abscissa direction, as coding " Champagne Tower " sequence right visual point image code check one timing, by JMVC verification platform and the inventive method encode after rebuild coloured silk The PSNR of color video sequence keeps consistent substantially;On vertical coordinate direction, when being compiled by JMVC verification platform and the inventive method When the PSNR of the color image sequence rebuild after Ma is identical, used by the right visual point image of coding " Champagne tower " sequence Code check essentially identical.To sum up analyzing and can obtain, the fast encoding method of the present invention can keep verifying platform one with JMVC substantially The distortion performance caused.

Table 1 original color video sequence by JMVC verification platform and the scramble time of the inventive method and encoder bit rate with And rebuild the quality versus of color image sequence

Subjective, the quality of the color image sequence that JMVC verification platform and the inventive method are rebuild is contrasted, Fig. 5 a gives the 2nd viewpoint the 10th two field picture that " Exit " cycle tests uses multiple views criteria check platform to rebuild, and Fig. 5 b gives Go out the 2nd viewpoint the 10th two field picture that " Exit " cycle tests uses the fast encoding method of the present invention to rebuild, the most several Do not see the difference of Fig. 5 a and Fig. 5 b;Fig. 5 c gives " Vassar " cycle tests and uses multiple views criteria check platform to rebuild The 2nd viewpoint the 10th two field picture, Fig. 5 d give " Vassar " cycle tests use the present invention fast encoding method rebuild 2nd viewpoint the 10th two field picture, does not the most almost see the difference of Fig. 5 c and Fig. 5 d.Therefore, the fast coding side of the present invention The right visual point image of the multi-view video sequences that method and JMVC verification platform are rebuild has essentially no difference intuitively.

To sum up, for the right visual point image of multi-view video sequences, the inventive method is at conservation rate distortion performance and JMVC On the premise of verification platform is consistent, it is possible to be greatly saved the scramble time of the right visual point image of multi-view video sequences, thus save Save the binary encoding time of multi-view video sequences.

Claims (5)

1. the multiple views color video fast encoding method that just can perceive distortion based on binocular, it is characterised in that include following Step:
1. the left viewpoint video of multiple views color video is designated as { CL(k) }, the right viewpoint video of multiple views color video is designated as {CR(k) }, wherein, CLK () represents { CL(k) } in kth frame left view dot image, CRK () represents { CR(k) } in the kth frame right side regard Dot image, 1≤k≤K, K represent the frame number of the image comprised in left viewpoint video and right viewpoint video;
2. by { CR(k) } in the right visual point image of every frame be divided into borderline region and non-borderline region, wherein, borderline region is by the right side The first row macro block, last column macro block, first row macro block and last string macro block in visual point image are constituted;Then { C is calculatedR (k) } in the right visual point image of every frame in non-borderline region in the binocular of each macro block just can perceive distortion value, it is assumed that { CR (k) } in kth frame right visual point image CRK in (), coordinate position is that (i, macro block j) belongs to { CR(k) } in the right viewpoint of kth frame Image CRK the non-borderline region in (), then by { CR(k) } in kth frame right visual point image CRIn (k) coordinate position be (i, j) The binocular of macro block just can be perceived distortion value and be designated asWherein, 2≤i≤W/16-1,2≤j≤H/16-1, W represent {CL(k) } in every frame left view dot image and { CR(k) } in the width of the right visual point image of every frame, H represents { CL(k) } in every Frame left view dot image and { CR(k) } in the height of the right visual point image of every frame;
3., on multiple view video coding Knowledge Verification Model JMVC, use HBP forecast coding structure to { CL(k) } in every frame left view Each macro block in dot image and { CR(k) } in the right visual point image of every frame in each macro block encode, at cataloged procedure In select the process of optimum macro-block coding pattern to be for each macro block:
3.-1, macro block the most to be encoded is defined as current macro;
3. it is-2, { C when current macroL(k) } in macro block time, encoder use H.264 mode selection processes search SKIP, Inter16×16、Inter16×8、Inter8×16、Inter8×8、Inter8×8Frext、Intra16×16、Intra8 × 8 and Intra4 × 4 macro-block coding pattern, the macroblock coding mould that elimination factor distortion cost is minimum from these macro-block coding patterns Formula encodes as the optimum macro-block coding pattern of current macro;
When current macro is { CR(k) } in macro block time, it is judged that current macro belongs to borderline region and still falls within non-borderline region, If current macro belongs to borderline region, then encoder use mode selection processes search SKIP, Inter16 × 16 H.264, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4 macro-block coding patterns, the macro-block coding pattern that elimination factor distortion cost is minimum from these macro-block coding patterns is as currently The optimum macro-block coding pattern of macro block encodes;If current macro belongs to non-borderline region, judge current macro the most again Binocular just can perceive the decision threshold whether distortion value is more than or equal to set, if it is, encoder uses pattern H.264 Selection course search SKIP, Inter16 × 16, Inter16 × 8, Inter8 × 16, Inter8 × 8, Inter8 × 8Frext, Intra16 × 16, Intra8 × 8 and Intra4 × 4 macro-block coding pattern, elimination factor distortion generation from these macro-block coding patterns The macro-block coding pattern of valency minimum encodes as the optimum macro-block coding pattern of current macro;Otherwise, encoder uses H.264 mode selection processes only searches for SKIP and Inter16 × 16 macro-block coding pattern, from both macro-block coding patterns The macro-block coding pattern of elimination factor distortion cost minimum encodes as the optimum macro-block coding pattern of current macro;
3.-3, using macro block to be encoded for the next one as current macro, it is then back to step and 3.-2 continues executing with, until { CL(k)} In every frame left view dot image in each macro block and { CR(k) } in the right visual point image of every frame in each macro block all complete compile Code.
The multiple views color video fast encoding method that just can perceive distortion based on binocular the most according to claim 1, its It is characterised by { C during described step is 2.R(k) } in kth frame right visual point image CRIn (k) coordinate position be (i, macro block j) Binocular just can perceive distortion value Wherein, d represents { CR(k) } in kth frame right visual point image CRK in (), coordinate position is (i, the parallax value of macro block j), bgL(i + d, j) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, all pixels in macro block j) The meansigma methods of the brightness value of point, bgL(i+d, j) ∈ [0,255], bgL(i × 16+d × 16+m, j × 16+n) represents { CL(k) } in kth frame left view dot image CLIn (k) coordinate position be (i × 16+d × 16+m, j × 16+n) the brightness value of pixel, 0≤m≤15,0≤n≤15, ehL(i+d j) represents { CL(k) } in Kth frame left view dot image CLK in (), coordinate position is (i+d, putting down of the edge intensity value computing of all pixels in macro block j) Average, bgL (i × 16+d × 16+m-3+h, j × 16+n-3+v) represents { CL(k) } in kth frame left view dot image CLK in (), coordinate position is The brightness value of the pixel of (i × 16+d × 16+m-3+h, j × 16+n-3+v), GH(h v) represents that the horizontal Sobel of 5 × 5 calculates Sub-GHMiddle coordinate position be (h, v) element at place,GV(h v) represents the vertical Sobel operator G of 5 × 5VIn Coordinate position be (h, v) element at place,1≤h≤5,1≤v≤5, Alimt(bgL(i+d,j),ehL (i+d, j))=Alimit(bgL(i+d,j))+K(bgL(i+d,j))×ehL(i+d, j), K(bgL(i+d, j))=-10-6×(0.7×(bgL(i+d,j))2+32×bgL(i+d, j))+0.07, λ represent the right viewpoint of control The parameter of influence of noise, nL(i+d j) represents { CL(k) } in kth frame left view dot image CLIn (k) coordinate position be (i+d, j) The noise amplitude of macro block.
The multiple views color video fast encoding method that just can perceive distortion based on binocular the most according to claim 2, its The parameter lambda value controlling right viewpoint influence of noise being characterised by described is 1.25.
The multiple views color video fast encoding method that just can perceive distortion based on binocular the most according to claim 3, its It is characterised by described { CL(k) } in kth frame left view dot image CLK in (), coordinate position is (i+d, the noise width of macro block j) Value nL(i+d, j) value is 0.3.
The multiple views color video that just can perceive distortion based on binocular the most according to any one of claim 1 to 4 is quick Coded method, it is characterised in that the decision threshold value that 3. described step sets in-2 is as 5.
CN201310325370.8A 2013-07-30 2013-07-30 The multiple views color video fast encoding method of distortion just can be perceived based on binocular CN103442226B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310325370.8A CN103442226B (en) 2013-07-30 2013-07-30 The multiple views color video fast encoding method of distortion just can be perceived based on binocular

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310325370.8A CN103442226B (en) 2013-07-30 2013-07-30 The multiple views color video fast encoding method of distortion just can be perceived based on binocular

Publications (2)

Publication Number Publication Date
CN103442226A CN103442226A (en) 2013-12-11
CN103442226B true CN103442226B (en) 2016-08-17

Family

ID=49695886

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310325370.8A CN103442226B (en) 2013-07-30 2013-07-30 The multiple views color video fast encoding method of distortion just can be perceived based on binocular

Country Status (1)

Country Link
CN (1) CN103442226B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105611272B (en) * 2015-12-28 2017-05-03 宁波大学 Eye exactly perceptible stereo image distortion analyzing method based on texture complexity

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008136607A1 (en) * 2007-05-02 2008-11-13 Samsung Electronics Co., Ltd. Method and apparatus for encoding and decoding multi-view video data
CN101404766A (en) * 2008-11-05 2009-04-08 宁波大学 Multi-view point video signal encoding method
CN102724525A (en) * 2012-06-01 2012-10-10 宁波大学 Depth video coding method on basis of foveal JND (just noticeable distortion) model

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008136607A1 (en) * 2007-05-02 2008-11-13 Samsung Electronics Co., Ltd. Method and apparatus for encoding and decoding multi-view video data
CN101404766A (en) * 2008-11-05 2009-04-08 宁波大学 Multi-view point video signal encoding method
CN102724525A (en) * 2012-06-01 2012-10-10 宁波大学 Depth video coding method on basis of foveal JND (just noticeable distortion) model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Binocular Just-Noticeable-Difference Model for Stereoscopic Image;Zhao Y,Chen Z Z,Zhu C;《Signal Processing Letters IEEE》;20101028;第18卷(第1期);19-22 *
JND模型及其在视频编码中的应用;连凤宗;《中国优秀硕士学位论文全文数据库(信息科技辑)》;20120515(第05期);I136-364 *

Also Published As

Publication number Publication date
CN103442226A (en) 2013-12-11

Similar Documents

Publication Publication Date Title
Kim et al. Deep learning of human visual sensitivity in image quality assessment framework
CN103281538B (en) Based on the inner frame coding method of rolling Hash and block rank infra-frame prediction
Kiran Adhikarla et al. Towards a quality metric for dense light fields
CN104935941B (en) The method being decoded to intra prediction mode
CN103152600B (en) Three-dimensional video quality evaluation method
US8610758B2 (en) Depth map generation for a video conversion system
Boev et al. Towards compound stereo-video quality metric: a specific encoder-based framework
CN103873861B (en) Coding mode selection method for HEVC (high efficiency video coding)
CN104539962B (en) It is a kind of merge visually-perceptible feature can scalable video coding method
CN100496121C (en) Image signal processing method of the interactive multi-view video system
CN101888566B (en) Estimation method of distortion performance of stereo video encoding rate
CN1233168C (en) Encoding motion image equipment and method
CN100461867C (en) Inage predicting encoding method in frame
CN104125469B (en) A kind of fast encoding method for HEVC
CN104768005A (en) Methods and apparatuses for encoding/decoding high resolution images
JP5988577B2 (en) Image coding apparatus, image coding method, and program
CN101710993B (en) Block-based self-adaptive super-resolution video processing method and system
CN102769746B (en) Method for processing multi-viewpoint depth video
CN103957415B (en) CU dividing methods and device based on screen content video
CN103297781A (en) High efficiency video coding (HEVC) intraframe coding method, device and system based on texture direction
CN103124347B (en) Vision perception characteristic is utilized to instruct the method for multiple view video coding quantizing process
CN103546758B (en) A kind of fast deep graphic sequence inter mode decision fractal coding
CN1551637A (en) Frame interpolating method and apparatus thereof at frame rate conversion
CN1176562A (en) Quantization process for video encoding
CN101986716B (en) Quick depth video coding method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant