CN102708568B - Stereoscopic image objective quality evaluation method on basis of structural distortion - Google Patents

Abstract

The invention discloses a stereoscopic image objective quality evaluation method on the basis of the structural distortion, which comprises the following steps: firstly, respectively carrying out regional division on left and right viewpoint images of an undistorted stereoscopic image and a distorted stereoscopic image to obtain an eye sensitive region and a corresponding nonsensitive region and then respectively obtaining evaluation indexes of the sensitive region and the nonsensitive region from two aspects of the structural amplitude distortion and the structural direction distortion; secondly, acquiring quality evaluation values of the left and right viewpoint images; thirdly, sampling singular value difference and a mean deviation ratio of residual images of which singular values are deprived to evaluate the distortion condition of the depth perception of a stereoscopic image so as to obtain an evaluation value of stereoscopic perceived quality; and finally, combining the quality of the left and right viewpoint images with the stereoscopic perceived quality to obtain a final quality evaluation result of the stereoscopic image. The method disclosed by the invention avoids simulating each composition part of an eye vision system, but sufficiently utilizes structure information of the stereoscopic image, so the consistency of the objective evaluation result and the subjective perception is effectively improved.

Description

A kind of three-dimensional image objective quality evaluation method based on structure distortion

Technical field

The present invention relates to a kind of image quality evaluation technology, especially relate to a kind of three-dimensional image objective quality evaluation method based on structure distortion.

Background technology

Stereo image quality is evaluated in stereoscopic image/video system in occupation of consequence very, not only can pass judgment on the quality of Processing Algorithm in stereoscopic image/video system, and can also optimize and design this algorithm, to improve the efficiency of stereoscopic image/video disposal system.Stereo image quality evaluation method is mainly divided into two classes: subjective quality assessment and evaluating objective quality.Subjective quality assessment method is exactly that several observers are weighted to average comprehensive evaluation to the quality of stereo-picture to be evaluated, its result meets human visual system's characteristic, but it is subject to, and calculating is inconvenient, speed is slow, the restriction of high in cost of production factors, cause embedded system difficult, thereby cannot extensively be promoted in actual applications.And that method for evaluating objective quality has is simple to operate, cost is low, be easy to realize and the feature such as real-time optimization algorithm, become the emphasis of stereo image quality evaluation study.

At present, the three-dimensional image objective quality evaluation model of main flow comprises left and right view-point image quality evaluation and depth perception quality assessment two parts.But, because the mankind are limited to human visual system's understanding, be difficult to simulate exactly each ingredient of human eye, so the consistance between these models and subjective perception is not fine.

Summary of the invention

Technical matters to be solved by this invention is to provide a kind of three-dimensional image objective quality evaluation method based on structure distortion, and it can effectively improve the consistance between three-dimensional image objective quality evaluation result and subjective perception.

The present invention solves the problems of the technologies described above adopted technical scheme: a kind of three-dimensional image objective quality evaluation method based on structure distortion, is characterized in that comprising the following steps:

1. make S _orgundistorted stereo-picture for original, makes S _disfor the stereo-picture of distortion to be evaluated, by original undistorted stereo-picture S _orgleft viewpoint gray level image be designated as L _org, by original undistorted stereo-picture S _orgright viewpoint gray level image be designated as R _org, by the stereo-picture S of distortion to be evaluated _disleft viewpoint gray level image be designated as L _dis, by the stereo-picture S of distortion to be evaluated _disright viewpoint gray level image be designated as R _dis;

2. to L _organd L _dis, R _organd R _dis4 width images are implemented respectively region and are divided, and obtain respectively L _organd L _dis, R _organd R _diseach self-corresponding sensitizing range matrix mapping graph of 4 width images, by L _organd L _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _l, for A _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _l(i, j), by R _organd R _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _r, for A _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _r(i, j), wherein, 0≤i≤(W-8), 0≤j≤(H-8), W represents L herein _org, L _dis, R _organd R _diswide, H represents L _org, L _dis, R _organd R _disheight;

3. by L _organd L _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate L _organd L _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _l, for B _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _l(i, j), $B_L(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},L}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},L}(i,j)\right)}^2+C_1},$ Wherein, B _l(i, j) also represents L _orgoverlapping block and L that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))}^2},$ $U_{\mathrm{org},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))}^2},$ $U_{\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))\×(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))),$ L _org(i+x, j+y) represents L _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), L _dis(i+x, j+y) represents L _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, 0≤i≤(W-8) herein, 0≤j≤(H-8);

By R _organd R _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate R _organd R _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _r, for B _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _r(i, j), $B_R(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},R}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},R}(i,j)\right)}^2+C_1},$ Wherein, B _r(i, j) also represents R _orgoverlapping block and R that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))}^2},$ $U_{\mathrm{org},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))}^2},$ $U_{\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))\×(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))),$ R _org(i+x, j+y) represents R _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), R _dis(i+x, j+y) represents R _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, 0≤i≤(W-8) herein, 0≤j≤(H-8);

4. to L _organd L _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively L _organd L _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by L _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, L}, for I _{h, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i+2,j)+2L_{\mathrm{org}}(i+2,j+1)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i,j+1)-L_{\mathrm{org}}(i,j+2)\end{array},$ By L _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, L}, for I _{v, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i,j+2)+2L_{\mathrm{org}}(i+1,j+2)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i+1,j)-L_{\mathrm{org}}(i+2,j)\end{array},$ By L _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, L}, for I _{h, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i+2,j)+2L_{\mathrm{dis}}(i+2,j+1)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i,j+1)-L_{\mathrm{dis}}(i,j+2)\end{array},$ By L _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, L}, for I _{v, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i,j+2)+2L_{\mathrm{dis}}(i+1,j+2)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i+1,j)-L_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, L _org(i+2, j), L _org(i+2, j+1), L _org(i+2, j+2), L _org(i, j), L _org(i, j+1), L _org(i, j+2), L _org(i+1, j+2), L _org(i+1, j) be the corresponding L that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), L _dis(i+2, j), L _dis(i+2, j+1), L _dis(i+2, j+2), L _dis(i, j), L _dis(i, j+1), L _dis(i, j+2), L _dis(i+1, j+2), L _dis(i+1, j) be the corresponding L that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j);

To R _organd R _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively R _organd R _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by R _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, R}, for I _{h, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i+2,j)+2R_{\mathrm{org}}(i+2,j+1)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i,j+1)-R_{\mathrm{org}}(i,j+2)\end{array},$ By R _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, R}, for I _{v, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i,j+2)+2R_{\mathrm{org}}(i+1,j+2)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i+1,j)-R_{\mathrm{org}}(i+2,j)\end{array},$ By R _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, R}, for I _{h, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i+2,j)+2R_{\mathrm{dis}}(i+2,j+1)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i,j+1)-R_{\mathrm{dis}}(i,j+2)\end{array},$ By R _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, R}, for I _{v, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i,j+2)+2R_{\mathrm{dis}}(i+1,j+2)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i+1,j)-R_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, R _org(i+2, j), R _org(i+2, j+1), R _org(i+2, j+2), R _org(i, j), R _org(i, j+1), R _org(i, j+2), R _org(i+1, j+2), R _org(i+1, j) be the corresponding R that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), R _dis(i+2, j), R _dis(i+2, j+1), R _dis(i+2, j+2), R _dis(i, j), R _dis(i, j+1), R _dis(i, j+2), R _dis(i+1, j+2), R _dis(i+1, j) be the corresponding R that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j);

5. calculate L _organd L _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _l, for E _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as E _l(i, j), $E_L(i,j)=\frac{I_{h,\mathrm{org},L}(i,j)\×I_{h,\mathrm{dis},L}(i,j)+I_{v,\mathrm{org},L}(i,j)\×I_{v,\mathrm{dis},L}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},L}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},L}(i,j)\right)}^2}+C_2},$ Wherein, C ₂represent constant;

Calculate R _organd R _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _r, for E _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as E _r(i, j), $E_R(i,j)=\frac{I_{h,\mathrm{org},R}(i,j)\×I_{h,\mathrm{dis},R}(i,j)+I_{v,\mathrm{org},R}(i,j)\×I_{v,\mathrm{dis},R}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},R}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},R}(i,j)\right)}^2}+C_2};$

6. calculate L _organd L _disstructure distortion evaluation of estimate, be designated as Q ^l, Q ^l=ω ₁* Q _m,L+ ω ₂* Q _{nm, L}, $Q_{m,L}=\frac{1}{N_{L,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×A_L(i,j)),$ $N_{L,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_L(i,j),$ $Q_{\mathrm{nm},L}=\frac{1}{N_{L,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×(1-A_L(i,j))),$ $N_{L,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_L(i,j)),$ Wherein, ω ₁represent L _organd L _disthe weighted value of middle sensitizing range, ω ₂represent L _organd L _disthe weighted value of middle de-militarized zone;

Calculate R _organd R _disstructure distortion evaluation of estimate, be designated as Q ^r, Q ^r=ω ' ₁* Q _m,R+ ω ' ₂* Q _{nm, R}, $Q_{m,R}=\frac{1}{N_{R,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×A_R(i,j)),$ $N_{R,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_R(i,j),$ $Q_{\mathrm{nm},R}=\frac{1}{N_{R,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×(1-A_R(i,j))),$ $N_{R,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_R(i,j)),$ Wherein, ω ' ₁represent R _organd R _disthe weighted value of middle sensitizing range, ω ' ₂represent R _organd R _disthe weighted value of middle de-militarized zone;

7. according to Q ^land Q ^rcalculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgspatial frequency measuring similarity, be designated as Q _f, Q _f=β ₁* Q ^l+ (1-β ₁) * Q ^r, wherein, β ₁represent Q ^lweights;

8. calculate L _organd R _orgabsolute difference image, take matrix representation as calculate L _disand R _disabsolute difference image, take matrix representation as wherein, " || " is the symbol that takes absolute value;

9. will with width image is divided into respectively the size of individual non-overlapping copies is 8 * 8 piece, then right with implement respectively svd, obtain for all in width image the singular value mapping graph that the corresponding singular value matrix by its each piece forms with the singular value mapping graph that the corresponding singular value matrix by its each piece forms, will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _org, for G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _dis, for G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as wherein, W _lRrepresent with wide, H _lRrepresent with height, 0≤p≤7,0≤q≤7;

10. calculate corresponding singular value mapping graph and the singular value deviation evaluation of estimate of corresponding singular value mapping graph, is designated as K, $K=\frac{64}{W_{\mathrm{LR}}\×H_{\mathrm{LR}}}\×\underset{n=0}{\overset{W_{\mathrm{LR}}\×H_{\mathrm{LR}}/64-1}{\mathrm{\Σ}}}\frac{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}(G_{\mathrm{org}}^n(p,p)\×|G_{\mathrm{org}}^n(p,p)-G_{\mathrm{dis}}^n(p,p)\left|\right)}{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}{G}_{\mathrm{org}}^n(p,p)},$ Wherein, represent G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates, represent G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates;

right with implement respectively svd, obtain respectively with each self-corresponding 2 orthogonal matrixes and 1 singular value matrix, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _organd V _org, will the singular value matrix of implementing to obtain after svd is designated as O _org, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _disand V _dis, will the singular value matrix of implementing to obtain after svd is designated as O _dis,

calculate respectively with width image is deprived the residual matrix diagram after singular value, will the residual matrix diagram of depriving after singular value is designated as X _org, X _org=χ _org* Λ * V _org, will the residual matrix diagram of depriving after singular value is designated as X _dis, X _dis=χ _dis* Λ * V _dis, wherein, Λ representation unit matrix, the size of Λ and O _organd O _disin the same size;

calculate X _organd X _dismean bias rate, be designated as wherein, x represents X _organd X _disin the horizontal ordinate of pixel, y represents X _organd X _disin the ordinate of pixel;

calculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgthree-dimensional perception evaluating deg amount, be designated as Q _s, wherein, τ represents constant, for regulate K and at Q _smiddle risen importance;

according to Q _fand Q _s, calculate the stereo-picture S of distortion to be evaluated _disimage quality evaluation score value, be designated as Q, Q=Q _f* (Q _s) ^ρ, wherein, ρ represents weight coefficient value.

Described step is middle L 2. _organd L _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _lacquisition process be:

2.-a1, to L _orgmake level and vertical direction Sobel operator and process, obtain L _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l1}and Z _{v, l1}, then calculate L _orggradient magnitude figure, be designated as Z _l1, wherein, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l1}(x, y) represents Z _{h, l1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l1}(x, y) represents Z _{v, l1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l1wide, H' represents Z _l1height;

2.-a2, to L _dismake level and vertical direction Sobel operator and process, obtain L _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l2}and Z _{v, l2}, then calculate L _disgradient magnitude figure, be designated as Z _l2, wherein, Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l2}(x, y) represents Z _{h, l2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l2}(x, y) represents Z _{v, l2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l2wide, H' represents Z _l2height;

2.-a3, required threshold value T while calculating zoning, $T=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l2}(x,y)),$ Wherein, α is constant, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y);

2.-a4, by Z _l1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l1(i, j), by Z _l2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l2(i, j), judgement Z _l1(i, j) >T or Z _l2whether (i, j) >T sets up, if set up, determines L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _l(i, j)=1, otherwise, determine L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _l(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8);

Described step is middle R 2. _organd R _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _racquisition process be:

2.-b1, to R _orgmake level and vertical direction Sobel operator and process, obtain R _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r1}and Z _{v, r1}, then calculate R _orggradient magnitude figure, be designated as Z _r1, wherein, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r1}(x, y) represents Z _{h, r1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r1}(x, y) represents Z _{v, r1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r1wide, H' represents Z _r1height;

2.-b2, to R _dismake level and vertical direction Sobel operator and process, obtain R _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r2}and Z _{v, r2}, then calculate R _disgradient magnitude figure, be designated as Z _r2, wherein, Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r2}(x, y) represents Z _{h, r2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r2}(x, y) represents Z _{v, r2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r2wide, H' represents Z _r2height;

2.-b3, required threshold value T' while calculating zoning, $T^{\′}=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r2}(x,y)),$ Wherein, α is constant, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y);

2.-b4, by Z _r1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r1(i, j), by Z _r2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r2(i, j), judgement Z _r1(i, j) >T or Z _r2whether (i, j) >T sets up, if set up, determines R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _r(i, j)=1, otherwise, determine R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _r(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8).

Described step is middle β 7. ₁acquisition process be:

7.-1, adopt n undistorted stereo-picture to set up its distortion stereographic map image set under the different distortion levels of different type of distortion, this distortion stereographic map image set comprises the stereo-picture of several distortions, wherein, and n >=1;

7.-2, utilize subjective quality assessment method to obtain the average subjective scoring difference of the stereo-picture of the concentrated every width distortion of distortion stereo-picture, be designated as DMOS, DMOS=100-MOS, wherein, MOS represents subjective scoring average, DMOS ∈ [0,100];

7.-3, according to step 1. to step operating process 6., the left visual point image of the stereo-picture of every width distortion that calculated distortion stereo-picture is concentrated is with respect to the evaluation of estimate Q of the sensitizing range of the left visual point image of the undistorted stereo-picture of correspondence _m,Levaluation of estimate Q with de-militarized zone _{nm, L};

7.-4, adopt Mathematical Fitting method matching distortion stereo-picture to concentrate average subjective scoring difference DMOS and the corresponding Q of the stereo-picture of distortion _m,Land Q _{nm, L}thereby, obtain β ₁value.

Compared with prior art, the invention has the advantages that first respectively the left visual point image of the stereo-picture of undistorted stereo-picture and distortion and right visual point image are carried out to region division, obtain human eye sensitizing range and corresponding de-militarized zone, then from structure amplitude distortion and structure direction distortion two aspects, draw respectively the evaluation index of sensitizing range and de-militarized zone; Next adopts linear weighted function to obtain respectively left view-point image quality evaluation of estimate and right view-point image quality evaluation of estimate, and then obtains left and right view-point image quality evaluation of estimate; Again according to singular value, can characterize preferably stereo-picture structural information characteristic, sampling singular value difference is weighed the distortion situation of the depth perception of stereo-picture with the mean bias rate of depriving the residual image after singular value, thereby obtains the evaluation of estimate of three-dimensional perceived quality; Finally by left and right view-point image quality and three-dimensional perceived quality with nonlinear way combination, obtain the final mass evaluation result of stereo-picture, because the inventive method avoids simulating each ingredient of human visual system, and take full advantage of the structural information of stereo-picture, therefore effectively improved the consistance of objective evaluation result and subjective perception.

Accompanying drawing explanation

Fig. 1 be the inventive method totally realize block diagram;

Fig. 2 a is Akko & Kayo (640 * 480) stereo-picture;

Fig. 2 b is Alt Moabit (1024 * 768) stereo-picture;

Fig. 2 c is Balloons (1024 * 768) stereo-picture;

Fig. 2 d is Door Flowers (1024 * 768) stereo-picture;

Fig. 2 e is Kendo (1024 * 768) stereo-picture;

Fig. 2 f is Leaving Laptop (1024 * 768) stereo-picture;

Fig. 2 g is Lovebird1 (1024 * 768) stereo-picture;

Fig. 2 h is Newspaper (1024 * 768) stereo-picture;

Fig. 2 i is Xmas (640 * 480) stereo-picture;

Fig. 2 j is Puppy (720 * 480) stereo-picture;

Fig. 2 k is Soccer2 (720 * 480) stereo-picture;

Fig. 2 l is Horse (480 * 270) stereo-picture;

Fig. 3 is that the left view-point image quality of the inventive method is evaluated block diagram;

Fig. 4 a is different α and ω ₁under left view-point image quality and the CC performance change figure between subjective perceptual quality;

Fig. 4 b is different α and ω ₁under left view-point image quality and the SROCC performance change figure between subjective perceptual quality;

Fig. 4 c is different α and ω ₁under left view-point image quality and the RMSE performance change figure between subjective perceptual quality;

Fig. 5 a is at ω ₁in=1 situation, the left view-point image quality under different α and the CC performance change figure between subjective perceptual quality;

Fig. 5 b is at ω ₁in=1 situation, the left view-point image quality under different α and the SROCC performance change figure between subjective perceptual quality;

Fig. 5 c is at ω ₁in=1 situation, the left view-point image quality under different α and the RMSE performance change figure between subjective perceptual quality;

Fig. 6 a is different beta ₁under left and right view-point image quality and the CC performance change figure between subjective perceptual quality;

Fig. 6 b is different beta ₁under left and right view-point image quality and the SROCC performance change figure between subjective perceptual quality;

Fig. 6 c is different beta ₁under left and right view-point image quality and the RMSE performance change figure between subjective perceptual quality;

Fig. 7 a is three-dimensional depth perception quality under different τ and the CC performance change figure between subjective perceptual quality;

Fig. 7 b is three-dimensional depth perception quality under different τ and the SROCC performance change figure between subjective perceptual quality;

Fig. 7 c is three-dimensional depth perception quality under different τ and the RMSE performance change figure between subjective perceptual quality;

Fig. 8 a is stereo image quality under different ρ and the CC performance change figure between subjective perceptual quality;

Fig. 8 b is stereo image quality under different ρ and the SROCC performance change figure between subjective perceptual quality;

Fig. 8 c is stereo image quality under different ρ and the RMSE performance change figure between subjective perceptual quality.

Embodiment

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

A kind of three-dimensional image objective quality evaluation method based on structure distortion that the present invention proposes, its angle from structure distortion has been evaluated respectively the three-dimensional perceived quality of left and right view-point image quality and stereo-picture, and the mode of employing nonlinear weight obtains the final mass evaluation of estimate of stereo-picture.What Fig. 1 had provided the inventive method totally realizes block diagram, and it comprises the following steps:

1. make S _orgundistorted stereo-picture for original, makes S _disfor the stereo-picture of distortion to be evaluated, by original undistorted stereo-picture S _orgleft viewpoint gray level image be designated as L _org, by original undistorted stereo-picture S _orgright viewpoint gray level image be designated as R _org, by the stereo-picture S of distortion to be evaluated _disleft viewpoint gray level image be designated as L _dis, by the stereo-picture S of distortion to be evaluated _disright viewpoint gray level image be designated as R _dis.

2. to L _organd L _dis, R _organd R _dis4 width images are implemented respectively region and are divided, and obtain respectively L _organd L _dis, R _organd R _diseach self-corresponding sensitizing range matrix mapping graph of 4 width images, by L _organd L _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _l, for A _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _l(i, j), by R _organd R _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _r, for A _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _r(i, j), wherein, 0≤i≤(W-8), 0≤j≤(H-8), W represents L herein _org, L _dis, R _organd R _diswide, H represents L _org, L _dis, R _organd R _disheight.

In this specific embodiment, step is middle L 2. _organd L _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _lacquisition process be:

2.-a1, to L _orgmake level and vertical direction Sobel operator and process, obtain L _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l1}and Z _{v, l1}, then calculate L _orggradient magnitude figure, be designated as Z _l1, wherein, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l1}(x, y) represents Z _{h, l1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l1}(x, y) represents Z _{v, l1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l1wide, H' represents Z _l1height.

2.-a2, to L _dismake level and vertical direction Sobel operator and process, obtain L _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l2}and Z _{v, l2}, then calculate L _disgradient magnitude figure, be designated as Z _l2, wherein, Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l2}(x, y) represents Z _{h, l2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l2}(x, y) represents Z _{v, l2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l2wide, H' represents Z _l2height.

2.-a3, required threshold value T while calculating zoning, $T=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l2}(x,y)),$ Wherein, W' represents Z _l1and Z _l2wide, H' represents Z _l1and Z _l2height, α is constant, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y).

2.-a4, by Z _l1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l1(i, j), by Z _l2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l2(i, j), judgement Z _l1(i, j) >T or Z _l2whether (i, j) >T sets up, if set up, determines L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _l(i, j)=1, otherwise, determine L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _l(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8).

In this specific embodiment, step is middle R 2. _organd R _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _racquisition process be:

2.-b1, to R _orgmake level and vertical direction Sobel operator and process, obtain R _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r1}and Z _{v, r1}, then calculate R _orggradient magnitude figure, be designated as Z _r1, wherein, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r1}(x, y) represents Z _{h, r1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r1}(x, y) represents Z _{v, r1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r1wide, H' represents Z _r1height.

2.-b2, to R _dismake level and vertical direction Sobel operator and process, obtain R _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r2}and Z _{v, r2}, then calculate R _disgradient magnitude figure, be designated as Z _r2, wherein, Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r2}(x, y) represents Z _{h, r2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r2}(x, y) represents Z _{v, r2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r2wide, H' represents Z _r2height.

2.-b3, required threshold value T' while calculating zoning, $T^{\′}=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r2}(x,y)),$ Wherein, W' represents Z _r1and Z _r2wide, H' represents Z _r1and Z _r2height, α is constant, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y).

2.-b4, by Z _r1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r1(i, j), by Z _r2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r2(i, j), judgement Z _r1(i, j) >T or Z _r2whether (i, j) >T sets up, if set up, determines R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _r(i, j)=1, otherwise, determine R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _r(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8).

In the present embodiment, utilize as Fig. 2 a, Fig. 2 b, Fig. 2 c, Fig. 2 d, Fig. 2 e, Fig. 2 f, Fig. 2 g, Fig. 2 h, Fig. 2 i, Fig. 2 j, 12 undistorted stereo-pictures shown in Fig. 2 k and Fig. 2 l are set up its distortion stereographic map image set under the different distortion levels of different type of distortion, type of distortion comprises JPEG compression, JP2K compression, white Gaussian noise, Gaussian Blur and H264 coding distortion, and the left visual point image of stereo-picture and right visual point image are simultaneously with degree distortion, this distortion stereographic map image set comprises the stereo-picture of 312 width distortions altogether, the stereo-picture of the distortion that wherein JPEG compresses is totally 60 width, the stereo-picture of the distortion of JPEG2000 compression is totally 60 width, the stereo-picture of white Gaussian noise distortion is totally 60 width, the stereo-picture of Gaussian Blur distortion is totally 60 width, the stereo-picture of H264 coding distortion is totally 72 width.Above-mentioned 312 width stereo-pictures are carried out to Region Segmentation as above.

In the present embodiment, α value has determined the order of accuarcy that sensitizing range is divided, if value is excessive, sensitizing range can be mistaken as de-militarized zone, if value is too small, de-militarized zone can be mistaken as sensitizing range, so the contribution of the deterministic process of its value and left view-point image quality or right view-point image quality stereoscopic image quality decides.

3. by L _organd L _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate L _organd L _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _l, for B _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _l(i, j), $B_L(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},L}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},L}(i,j)\right)}^2+C_1},$ Wherein, B _l(i, j) also represents L _orgoverlapping block and L that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))}^2},$ $U_{\mathrm{org},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))}^2},$ $U_{\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))\×(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))),$ L _org(i+x, j+y) represents L _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), L _dis(i+x, j+y) represents L _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, C ₁be for fear of $B_L(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},L}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},L}(i,j)\right)}^2+C_1}$ Denominator there is zero situation, desirable C in actual application ₁=0.01,0≤i≤(W-8) herein, 0≤j≤(H-8).

At this, consider the correlativity between the pixel of image, left or right that overlapping block that size is 8 * 8 is the most adjacent with it has 7 column weights folded, equally, this 8 * 8 overlapping block the most adjacent with it upper piece or lower have 7 row overlapping.

By R _organd R _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate R _organd R _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _r, for B _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _r(i, j), $B_R(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},R}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},R}(i,j)\right)}^2+C_1},$ Wherein, B _r(i, j) also represents R _orgoverlapping block and R that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))}^2},$ $U_{\mathrm{org},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))}^2},$ $U_{\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))\×(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))),$ R _org(i+x, j+y) represents R _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), R _dis(i+x, j+y) represents R _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, C ₁be for fear of $B_R(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},R}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},R}(i,j)\right)}^2+C_1}$ Denominator there is zero situation, desirable C in actual application ₁=0.01,0≤i≤(W-8) herein, 0≤j≤(H-8).

4. to L _organd L _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively L _organd L _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by L _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, L}, for I _{h, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i+2,j)+2L_{\mathrm{org}}(i+2,j+1)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i,j+1)-L_{\mathrm{org}}(i,j+2)\end{array},$ By L _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, L}, for I _{v, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i,j+2)+2L_{\mathrm{org}}(i+1,j+2)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i+1,j)-L_{\mathrm{org}}(i+2,j)\end{array},$ By L _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, L}, for I _{h, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i+2,j)+2L_{\mathrm{dis}}(i+2,j+1)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i,j+1)-L_{\mathrm{dis}}(i,j+2)\end{array},$ By L _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, L}, for I _{v, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i,j+2)+2L_{\mathrm{dis}}(i+1,j+2)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i+1,j)-L_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, L _org(i+2, j), L _org(i+2, j+1), L _org(i+2, j+2), L _org(i, j), L _org(i, j+1), L _org(i, j+2), L _org(i+1, j+2), L _org(i+1, j) be the corresponding L that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), L _dis(i+2, j), L _dis(i+2, j+1), L _dis(i+2, j+2), L _dis(i, j), L _dis(i, j+1), L _dis(i, j+2), L _dis(i+1, j+2), L _dis(i+1, j) be the corresponding L that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j).

To R _organd R _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively R _organd R _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by R _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, R}, for I _{h, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i+2,j)+2R_{\mathrm{org}}(i+2,j+1)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i,j+1)-R_{\mathrm{org}}(i,j+2)\end{array},$ By R _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, R}, for I _{v, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i,j+2)+2R_{\mathrm{org}}(i+1,j+2)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i+1,j)-R_{\mathrm{org}}(i+2,j)\end{array},$ By R _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, R}, for I _{h, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i+2,j)+2R_{\mathrm{dis}}(i+2,j+1)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i,j+1)-R_{\mathrm{dis}}(i,j+2)\end{array},$ By R _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, R}, for I _{v, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i,j+2)+2R_{\mathrm{dis}}(i+1,j+2)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i+1,j)-R_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, R _org(i+2, j), R _org(i+2, j+1), R _org(i+2, j+2), R _org(i, j), R _org(i, j+1), R _org(i, j+2), R _org(i+1, j+2), R _org(i+1, j) be the corresponding R that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), R _dis(i+2, j), R _dis(i+2, j+1), R _dis(i+2, j+2), R _dis(i, j), R _dis(i, j+1), R _dis(i, j+2), R _dis(i+1, j+2), R _dis(i+1, j) be the corresponding R that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j).

5. calculate L _organd L _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _l, for E _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as E _l(i, j), $E_L(i,j)=\frac{I_{h,\mathrm{org},L}(i,j)\×I_{h,\mathrm{dis},L}(i,j)+I_{v,\mathrm{org},L}(i,j)\×I_{v,\mathrm{dis},L}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},L}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},L}(i,j)\right)}^2}+C_2},$ Wherein, C ₂represent constant, C ₂be for fear of $E_L(i,j)=\frac{I_{h,\mathrm{org},L}(i,j)\×I_{h,\mathrm{dis},L}(i,j)+I_{v,\mathrm{org},L}(i,j)\×I_{v,\mathrm{dis},L}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},L}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},L}(i,j)\right)}^2}+C_2}$ Denominator appear as zero situation, desirable C in actual application ₂=0.02.

Calculate R _organd R _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _r, for E _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as E _r(i, j), $E_R(i,j)=\frac{I_{h,\mathrm{org},R}(i,j)\×I_{h,\mathrm{dis},R}(i,j)+I_{v,\mathrm{org},R}(i,j)\×I_{v,\mathrm{dis},R}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},R}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},R}(i,j)\right)}^2}+C_2},$ C ₂be for fear of $E_R(i,j)=\frac{I_{h,\mathrm{org},R}(i,j)\×I_{h,\mathrm{dis},R}(i,j)+I_{v,\mathrm{org},R}(i,j)\×I_{v,\mathrm{dis},R}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},R}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},R}(i,j)\right)}^2}+C_2}$ Denominator appear as zero situation, desirable C in actual application ₂=0.02.

6. calculate L _organd L _disstructure distortion evaluation of estimate, be designated as Q ^l, Q ^l=ω ₁* Q _m,L+ ω ₂* Q _{nm, L}, $Q_{m,L}=\frac{1}{N_{L,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×A_L(i,j)),$ $N_{L,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_L(i,j),$ $Q_{\mathrm{nm},L}=\frac{1}{N_{L,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×(1-A_L(i,j))),$ $N_{L,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_L(i,j)),$ Wherein, ω ₁represent L _organd L _disthe weighted value of middle sensitizing range, ω ₂represent L _organd L _disthe weighted value of middle de-militarized zone.

Calculate R _organd R _disstructure distortion evaluation of estimate, be designated as Q ^r, Q ^r=ω ' ₁* Q _m,R+ ω ' ₂* Q _{nm, R}, $Q_{m,R}=\frac{1}{N_{R,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×A_R(i,j)),$ $N_{R,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_R(i,j),$ $Q_{\mathrm{nm},R}=\frac{1}{N_{R,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×(1-A_R(i,j))),$ $N_{R,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_R(i,j)),$ Wherein, ω ' ₁represent R _organd R _disthe weighted value of middle sensitizing range, ω ' ₂represent R _organd R _disthe weighted value of middle de-militarized zone.

In the present embodiment, Fig. 3 has provided the block diagram of realizing of left view-point image quality evaluation.The undistorted stereo-pictures of as shown in Fig. 2 a to Fig. 2 l 12 of utilization are set up the distortion stereographic map image set that the stereo-picture by 312 width distortions forms, to the stereo-picture of this 312 width distortion, adopt known subjective quality assessment method to carry out subjective quality assessment, the stereo-picture that obtains 312 width distortions average subjective scoring difference (DMOS separately, Difference Mean Opinion Scores), i.e. the subjective quality score value of the stereo-picture of every width distortion.DMOS is the difference of subjective scoring average (MOS) and full marks (100), i.e. DMOS=100-MOS, therefore, the quality of the stereo-picture of the larger expression distortion of DMOS value is poorer, the quality of the stereo-picture of the less expression distortion of DMOS value is better, and the span of DMOS is [0,100].On the other hand, 6. 1. the stereo-picture of above-mentioned 312 width distortions extremely calculated to the corresponding Q of stereo-picture of every width distortion by the inventive method step _m,Land Q _{nm, L}; Then adopt Q ^l=ω ₁* Q _m,L+ (1-ω ₁) * Q _{nm, L}make four parameter L ogistic function nonlinear fittings, obtain α and ω ₁value.Here, utilize 3 conventional objective parameters of evaluate image quality evaluating method as evaluation index, be Pearson correlation coefficient (the Correlation Coefficient under non-linear regression condition, CC), Spearman related coefficient (Spearman Rank-Order Correlation Coefficient, SROCC) and square error coefficient (Rooted Mean Squared Error, RMSE), the precision of this objective models of stereo-picture evaluation function of CC reflection distortion, the monotonicity situation between SROCC reflection objective models and subjective perception.RMSE reflects the accuracy of its prediction.CC and SROCC value higher explanation three-dimensional image objective evaluation method and DMOS correlativity are better, and RMSE value lower explanation three-dimensional image objective evaluation method and DMOS correlativity are better.Fig. 4 a has provided different α and ω ₁under the left view-point image quality of 312 width stereo-pictures and the CC performance change between subjective perceptual quality, Fig. 4 b has provided different α and ω ₁under the left view-point image quality of 312 width stereo-pictures and the SROCC performance change between subjective perceptual quality, Fig. 4 c has provided different α and ω ₁under the left view-point image quality of 312 width stereo-pictures and the RMSE performance change between subjective perceptual quality, analysis chart 4a, Fig. 4 b and Fig. 4 c, known CC and SROCC are can be along with the change of ω 1 value large and become large, and RMSE is along with ω ₁the change of value is large and diminish, and illustrate that left view-point image quality is mainly that quality by sensitizing range determines, and the change of α value is little to the performance impact between left view-point image quality and subjective perception.Fig. 5 a has provided at ω ₁=1, ω ₂in=0 situation, the left view-point image quality of 312 width stereo-pictures under different α and the CC performance change between subjective perceptual quality; Fig. 5 b has provided at ω ₁=1, ω ₂in=0 situation, the left view-point image quality of 312 width stereo-pictures under different α and the SROCC performance change between subjective perceptual quality; Fig. 5 c has provided at ω ₁=1, ω ₂in=0 situation, the left view-point image quality of 312 width stereo-pictures under different α and the RMSE performance change between subjective perceptual quality; Analysis chart 5a, Fig. 5 b and Fig. 5 c, known CC, SROCC and RMSE value all fluctuate on hundredths, but all have a peak value.Therefore, in the present embodiment, get ω ₁=1, α=2.1.

7. according to Q ^land Q ^rcalculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgspatial frequency measuring similarity, be designated as Q _f, Q _f=β ₁* Q ^l+ (1-β ₁) * Q ^r, wherein, β ₁represent Q ^lweights.

In this specific embodiment, step is middle β 7. ₁acquisition process be:

7.-1, adopt n undistorted stereo-picture to set up its distortion stereographic map image set under the different distortion levels of different type of distortion, this distortion stereographic map image set comprises the stereo-picture of several distortions, wherein, and n >=1.

7.-2, utilize subjective quality assessment method to obtain the average subjective scoring difference of the stereo-picture of the concentrated every width distortion of distortion stereo-picture, be designated as DMOS, DMOS=100-MOS, wherein, MOS represents subjective scoring average, DMOS ∈ [0,100].

7.-3, according to step 1. to step operating process 6., the left visual point image of the stereo-picture of every width distortion that calculated distortion stereo-picture is concentrated is with respect to the evaluation of estimate Q of the sensitizing range of the left visual point image of the undistorted stereo-picture of correspondence _m,Levaluation of estimate Q with de-militarized zone _{nm, L}.

7.-4, adopt Mathematical Fitting method matching distortion stereo-picture to concentrate average subjective scoring difference DMOS and the corresponding Q of the stereo-picture of distortion _m,Land Q _{nm, L}thereby, obtain β ₁value.

In the present embodiment, β ₁determined Q ^lthe contribution of stereoscopic image quality, for blocking effect, half of stereo image quality the chances are the quality of left visual point image and the quality sum of right visual point image, for fuzzy distortion, stereo image quality depends primarily on quality that viewpoint preferably.Due to the distortion that left visual point image and the right visual point image of this stereo-picture test library is subject to same degree simultaneously, the quality of left visual point image and the mass change of right visual point image are little, therefore β ₁the subjective performance impact that changes stereoscopic image is little.First 6. 1. the stereo-picture of above-mentioned 312 width distortions extremely calculated to the corresponding Q of stereo-picture of every width distortion by the inventive method step ^land Q ^r, then adopt four parameter fittings to obtain β ₁value.Fig. 6 a has provided different beta ₁under the quality of left and right visual point image and the CC performance change between subjective perceptual quality, Fig. 6 a has provided different beta ₁under the quality of left and right visual point image and the SROCC performance change between subjective perceptual quality, Fig. 6 a has provided different beta ₁under the quality of left and right visual point image and the RMSE performance change between subjective perceptual quality, analysis chart 6a, Fig. 6 b and Fig. 6 c, known along with β ₁the variation of value, CC, SROCC and RMSE value change little, fluctuate, but all have peak value on hundredths.Here, get β ₁=0.5.

8. calculate L _organd R _orgabsolute difference image matrix represent, calculate L _disand R _disabsolute difference image matrix represent, wherein, " || " is the symbol that takes absolute value.

9. will with width image is divided into respectively the size of individual non-overlapping copies is 8 * 8 piece, then right with implement respectively svd, obtain for all in width image the singular value mapping graph that the corresponding singular value matrix by its each piece forms with the singular value mapping graph that the corresponding singular value matrix by its each piece forms, will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _org, for G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _dis, for G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as wherein, W _lRrepresent with wide, H _lRrepresent with height, 0≤p≤7,0≤q≤7.

At this, in order to reduce computation complexity, the piece of 8 * 8 the most adjacent with it left or right or upper piece or lower do not repeat row or repeated rows, i.e. piece and piece non-overlapping copies.

10. calculate corresponding singular value mapping graph and the singular value deviation evaluation of estimate of corresponding singular value mapping graph, is designated as K, $K=\frac{64}{W_{\mathrm{LR}}\×H_{\mathrm{LR}}}\×\underset{n=0}{\overset{W_{\mathrm{LR}}\×H_{\mathrm{LR}}/64-1}{\mathrm{\Σ}}}\frac{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}(G_{\mathrm{org}}^n(p,p)\×|G_{\mathrm{org}}^n(p,p)-G_{\mathrm{dis}}^n(p,p)\left|\right)}{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}{G}_{\mathrm{org}}^n(p,p)},$ Wherein, represent G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates, represent G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates.

right with implement respectively svd, obtain respectively with each self-corresponding 2 orthogonal matrixes and 1 singular value matrix, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _organd V _org, will the singular value matrix of implementing to obtain after svd is designated as O _org, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _disand V _dis, will the singular value matrix of implementing to obtain after svd is designated as O _dis,

calculate respectively with width image is deprived the residual matrix diagram after singular value, will the residual matrix diagram of depriving after singular value is designated as X _org, X _org=χ _org* Λ * V _org, will the residual matrix diagram of depriving after singular value is designated as X _dis, X _dis=χ _dis* Λ * V _dis, wherein, Λ representation unit matrix, the size of Λ and O _organd O _disin the same size.

calculate X _organd X _dismean bias rate, be designated as wherein, x represents X _organd X _disin the horizontal ordinate of pixel, y represents X _organd X _disin the ordinate of pixel.

calculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgthree-dimensional perception evaluating deg amount, be designated as Q _s, wherein, τ represents constant, for regulate K and at Q _smiddle risen importance.

In the present embodiment, first ask for the absolute difference image of stereo-picture and 10 undistorted stereo-pictures of above-mentioned 312 width distortions, then according to the step of the inventive method 8. extremely calculate every width distortion the corresponding K of stereo-picture and at this, τ value size has determined the importance that singular value deviation and residual risk are risen in depth perception is evaluated.Fig. 7 a has provided the three-dimensional perceived quality of stereo-picture of 312 width distortions under different τ and the CC performance change between subjective perception, Fig. 7 b has provided the three-dimensional perceived quality of stereo-picture of 312 width distortions under different τ and the SROCC performance change between subjective perception, Fig. 7 c has provided the three-dimensional perceived quality of stereo-picture of 312 width distortions under different τ and the RMSE performance change between subjective perception, Fig. 7 a, in Fig. 7 b and Fig. 7 c, τ changes in [164] scope, analysis chart 7a, Fig. 7 b and Fig. 7 c, known CC, all there is an extreme value in the variation of SROCC and RMSE and τ, and position is roughly the same, here get τ=-8.

according to Q _fand Q _s, calculate the stereo-picture S of distortion to be evaluated _disimage quality evaluation score value, be designated as Q, Q=Q _f* (Q _s) ^ρ, wherein, ρ represents weight coefficient value.

In the present embodiment, to the stereo-picture of above-mentioned 312 width distortions by the inventive method step 1. extremely calculate the corresponding Q of stereo-picture of every width distortion _fand Q _s, then adopt Q=Q _f* (Q _s) ^ρmake four parameter L ogistic function nonlinear fittings, obtain ρ, ρ value has determined quality and the contribution of three-dimensional perceived quality in stereo image quality of left and right visual point image.Q _fand Q _svalue is all along with stereo-picture distortion level is deepened and diminishes, therefore the span of ρ value is for being greater than 0.Fig. 8 a has provided the quality of 312 width stereo-pictures under different ρ values and the CC performance change between subjective perceptual quality, Fig. 8 b has provided the quality of 312 width stereo-pictures under different ρ values and the SROCC performance change between subjective perceptual quality, Fig. 8 c has provided the quality of 312 width stereo-pictures under different ρ values and the RMSE performance change between subjective perceptual quality, analysis chart 8a, Fig. 8 b, Fig. 8 c, known ρ value obtains and too large or too littlely all can affect the consistance between stereo image quality objective evaluation model and subjective perception, under ρ value situation of change, CC, all there is extreme point in SROCC and RMSE value, and approximate location is identical, here get ρ=0.3.

The image quality evaluation function Q=Q of the stereo-picture of the distortion that analysis the present embodiment obtains _f* (Q _s) ^0.3final appraisal results and the correlativity between subjective scoring DMOS.The image quality evaluation function Q=Q of the stereo-picture of the distortion first obtaining by the present embodiment _f* (Q _s) ^0.3the output valve Q of the final stereo image quality evaluation result calculating, is then output valve Q four parameter L ogistic function nonlinear fittings, finally obtains the performance index value between three-dimensional objective evaluation model and subjective perception.Here, utilize 4 of evaluate image quality evaluating method conventional objective parameters as evaluation index, CC, SROCC, be often worth ratio (Outlier Ratio, OR), RMSE.The dispersion degree of the objective rating model of OR reflection stereo image quality, the distortion stereo-picture number proportion that in all distortion stereo-pictures, the evaluation of estimate after four parameter fittings and the difference between DMOS are greater than a certain threshold value.Table 1 has provided assess performance CC, SROCC, OR and RMSE coefficient, as seen from the data in Table 1, and the image quality evaluation function Q=Q of the stereo-picture of the distortion obtaining by the present embodiment _f* (Q _s) ^0.3correlativity between the output valve Q of the final appraisal results that calculate and subjective scoring DMOS is very high, CC value and SROCC value all surpass 0.92, RMSE value, lower than 6.5, shows that the result of objective evaluation result and human eye subjective perception is more consistent, and the validity of the inventive method has been described.

The image quality evaluation score value of the stereo-picture of the distortion that this enforcement of table 1 obtains and the correlativity between subjective scoring

?	Gblur	JP2K	JPEG	WN	H264	ALL
							Number	60	60	60	60	72	312
CC	0.9658	0.9479	0.9533	0.9554	0.9767	0.9235
							SROCC	0.9655	0.9489	0.9524	0.9274	0.9545	0.9430
OR	0	0	0	0	0	0
							RMSE	5.4719	3.8180	4.3010	4.6151	3.0135	6.5890

Claims (3)

1. the three-dimensional image objective quality evaluation method based on structure distortion, is characterized in that comprising the following steps:

1. make S _orgundistorted stereo-picture for original, makes S _disfor the stereo-picture of distortion to be evaluated, by original undistorted stereo-picture S _orgleft viewpoint gray level image be designated as L _org, by original undistorted stereo-picture S _orgright viewpoint gray level image be designated as R _org, by the stereo-picture S of distortion to be evaluated _disleft viewpoint gray level image be designated as L _dis, by the stereo-picture S of distortion to be evaluated _disright viewpoint gray level image be designated as R _dis;

2. to L _organd L _dis, R _organd R _dis4 width images are implemented respectively region and are divided, and obtain respectively L _organd L _dis, R _organd R _diseach self-corresponding sensitizing range matrix mapping graph of 4 width images, by L _organd L _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _l, for A _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _l(i, j), by R _organd R _disthe matrix of coefficients of implementing respectively each self-corresponding sensitizing range matrix mapping graph of obtaining after region is divided is all designated as A _r, for A _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as A _r(i, j), wherein, 0≤i≤(W-8), 0≤j≤(H-8), W represents L herein _org, L _dis, R _organd R _diswide, H represents L _org, L _dis, R _organd R _disheight;

3. by L _organd L _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate L _organd L _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _l, for B _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _l(i, j), $B_L(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},L}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},L}(i,j)\right)}^2+C_1},$ Wherein, B _l(i, j) also represents L _orgoverlapping block and L that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))}^2},$ $U_{\mathrm{org},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},L}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))}^2},$ $U_{\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{L}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},L}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((L_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},L}(i,j))\×(L_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},L}(i,j))),$ L _org(i+x, j+y) represents L _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), L _dis(i+x, j+y) represents L _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, 0≤i≤(W-8) herein, 0≤j≤(H-8);

By R _organd R _dis2 width images are divided into respectively the overlapping block that (W-7) * (H-7) individual size is 8 * 8, then calculate R _organd R _disthe structure amplitude distortion mapping graph of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as B by the matrix of coefficients of this structure amplitude distortion mapping graph _r, for B _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as B _r(i, j), $B_R(i,j)=\frac{2\×{\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)+C_1}{{\left({\mathrm{\σ}}_{\mathrm{org},R}(i,j)\right)}^2+{\left({\mathrm{\σ}}_{\mathrm{dis},R}(i,j)\right)}^2+C_1},$ Wherein, B _r(i, j) also represents R _orgoverlapping block and R that the size that middle upper left corner coordinate position is (i, j) is 8 * 8 _disthe structure amplitude distortion value of the overlapping block that the size that middle upper left corner coordinate position is (i, j) is 8 * 8, ${\mathrm{\σ}}_{\mathrm{org},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))}^2},$ $U_{\mathrm{org},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{org}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{dis},R}(i,j)=\sqrt{\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))}^2},$ $U_{\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}{R}_{\mathrm{dis}}(i+x,j+y),$ ${\mathrm{\σ}}_{\mathrm{org},\mathrm{dis},R}(i,j)=\frac{1}{64}\underset{x=0}{\overset{7}{\mathrm{\Σ}}}\underset{y=0}{\overset{7}{\mathrm{\Σ}}}((R_{\mathrm{org}}(i+x,j+y)-U_{\mathrm{org},R}(i,j))\×(R_{\mathrm{dis}}(i+x,j+y)-U_{\mathrm{dis},R}(i,j))),$ R _org(i+x, j+y) represents R _orgmiddle coordinate position is the pixel value of the pixel of (i+x, j+y), R _dis(i+x, j+y) represents R _dismiddle coordinate position is the pixel value of the pixel of (i+x, j+y), C ₁represent constant, 0≤i≤(W-8) herein, 0≤j≤(H-8);

4. to L _organd L _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively L _organd L _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by L _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, L}, for I _{h, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i+2,j)+2L_{\mathrm{org}}(i+2,j+1)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i,j+1)-L_{\mathrm{org}}(i,j+2)\end{array},$ By L _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, L}, for I _{v, org, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},L}(i,j)=L_{\mathrm{org}}(i,j+2)+2L_{\mathrm{org}}(i+1,j+2)+L_{\mathrm{org}}(i+2,j+2)\\ -L_{\mathrm{org}}(i,j)-2L_{\mathrm{org}}(i+1,j)-L_{\mathrm{org}}(i+2,j)\end{array},$ By L _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, L}, for I _{h, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, L}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i+2,j)+2L_{\mathrm{dis}}(i+2,j+1)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i,j+1)-L_{\mathrm{dis}}(i,j+2)\end{array},$ By L _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, L}, for I _{v, dis, L}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, L}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},L}(i,j)=L_{\mathrm{dis}}(i,j+2)+2L_{\mathrm{dis}}(i+1,j+2)+L_{\mathrm{dis}}(i+2,j+2)\\ -L_{\mathrm{dis}}(i,j)-2L_{\mathrm{dis}}(i+1,j)-L_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, L _org(i+2, j), L _org(i+2, j+1), L _org(i+2, j+2), L _org(i, j), L _org(i, j+1), L _org(i, j+2), L _org(i+1, j+2), L _org(i+1, j) be the corresponding L that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), L _dis(i+2, j), L _dis(i+2, j+1), L _dis(i+2, j+2), L _dis(i, j), L _dis(i, j+1), L _dis(i, j+2), L _dis(i+1, j+2), L _dis(i+1, j) be the corresponding L that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j);

To R _organd R _dis2 width images are carrying out horizontal and the processing of vertical direction Sobel operator respectively, obtains respectively R _organd R _diseach self-corresponding horizontal direction gradient matrix mapping graph of 2 width images and vertical gradient matrix mapping graph, by R _orgthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, org, R}, for I _{h, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, org, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i+2,j)+2R_{\mathrm{org}}(i+2,j+1)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i,j+1)-R_{\mathrm{org}}(i,j+2)\end{array},$ By R _orgthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, org, R}, for I _{v, org, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, org, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{org},R}(i,j)=R_{\mathrm{org}}(i,j+2)+2R_{\mathrm{org}}(i+1,j+2)+R_{\mathrm{org}}(i+2,j+2)\\ -R_{\mathrm{org}}(i,j)-2R_{\mathrm{org}}(i+1,j)-R_{\mathrm{org}}(i+2,j)\end{array},$ By R _disthe matrix of coefficients of the corresponding horizontal direction gradient matrix mapping graph that carrying out horizontal direction Sobel operator obtains after processing is designated as I _{h, dis, R}, for I _{h, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{h, dis, R}(i, j), $\begin{array}{c}{I}_{h,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i+2,j)+2R_{\mathrm{dis}}(i+2,j+1)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i,j+1)-R_{\mathrm{dis}}(i,j+2)\end{array},$ By R _disthe matrix of coefficients of the corresponding vertical gradient matrix mapping graph that enforcement vertical direction Sobel operator obtains after processing is designated as I _{v, dis, R}, for I _{v, dis, R}middle coordinate position is the coefficient value that (i, j) locates, and is designated as I _{v, dis, R}(i, j), $\begin{array}{c}{I}_{v,\mathrm{dis},R}(i,j)=R_{\mathrm{dis}}(i,j+2)+2R_{\mathrm{dis}}(i+1,j+2)+R_{\mathrm{dis}}(i+2,j+2)\\ -R_{\mathrm{dis}}(i,j)-2R_{\mathrm{dis}}(i+1,j)-R_{\mathrm{dis}}(i+2,j)\end{array},$ Wherein, R _org(i+2, j), R _org(i+2, j+1), R _org(i+2, j+2), R _org(i, j), R _org(i, j+1), R _org(i, j+2), R _org(i+1, j+2), R _org(i+1, j) be the corresponding R that represents respectively _orgmiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j), R _dis(i+2, j), R _dis(i+2, j+1), R _dis(i+2, j+2), R _dis(i, j), R _dis(i, j+1), R _dis(i, j+2), R _dis(i+1, j+2), R _dis(i+1, j) be the corresponding R that represents respectively _dismiddle coordinate position is the pixel value of the pixel of (i+2, j), (i+2, j+1), (i+2, j+2), (i, j), (i, j+1), (i, j+2), (i+1, j+2), (i+1, j);

5. calculate L _organd L _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _l, for E _lmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as E _l(i, j), $E_L(i,j)=\frac{I_{h,\mathrm{org},L}(i,j)\×I_{h,\mathrm{dis},L}(i,j)+I_{v,\mathrm{org},L}(i,j)\×I_{v,\mathrm{dis},L}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},L}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},L}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},L}(i,j)\right)}^2}+C_2},$ Wherein, C ₂represent constant;

Calculate R _organd R _disthe structure direction distortion map figure of two overlapping blocks that in 2 width images, all coordinate positions are identical, is designated as E by the matrix of coefficients of this structure direction distortion map figure _r, for E _rmiddle coordinate position is the coefficient value that (i, j) locates, and is designated as ER (i, j), $E_R(i,j)=\frac{I_{h,\mathrm{org},R}(i,j)\×I_{h,\mathrm{dis},R}(i,j)+I_{v,\mathrm{org},R}(i,j)\×I_{v,\mathrm{dis},R}(i,j)+C_2}{\sqrt{{\left(I_{h,\mathrm{org},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{org},R}(i,j)\right)}^2}\×\sqrt{{\left(I_{h,\mathrm{dis},R}(i,j)\right)}^2+{\left(I_{v,\mathrm{dis},R}(i,j)\right)}^2}+C_2};$

6. calculate L _organd L _disstructure distortion evaluation of estimate, be designated as Q ^l, Q ^l=ω ₁* Q _m,L+ ω ₂* Q _{nm, L}, $Q_{m,L}=\frac{1}{N_{L,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×A_L(i,j)),$ $N_{L,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_L(i,j),$ $Q_{\mathrm{nm},L}=\frac{1}{N_{L,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_L(i,j)+E_L(i,j))\×(1-A_L(i,j))),$ $N_{L,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_L(i,j)),$ Wherein, ω ₁represent L _organd L _disthe weighted value of middle sensitizing range, ω ₂represent L _organd L _disthe weighted value of middle de-militarized zone;

Calculate R _organd R _disstructure distortion evaluation of estimate, be designated as Q ^r, Q ^r=ω ' ₁* Q _m,R+ ω ' ₂* Q _{nm, R}, $Q_{m,R}=\frac{1}{N_{R,m}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×A_R(i,j)),$ $N_{R,m}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}{A}_R(i,j),$ $Q_{\mathrm{nm},R}=\frac{1}{N_{R,\mathrm{nm}}}\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(0.5\×(B_R(i,j)+E_R(i,j))\×(1-A_R(i,j))),$ $N_{R,\mathrm{nm}}=\underset{i=0}{\overset{W-8}{\mathrm{\Σ}}}\underset{j=0}{\overset{H-8}{\mathrm{\Σ}}}(1-A_R(i,j)),$ Wherein, ω ' ₁represent R _organd R _disthe weighted value of middle sensitizing range, ω ' ₂represent R _organd R _disthe weighted value of middle de-militarized zone;

7. according to Q ^land Q ^rcalculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgspatial frequency measuring similarity, be designated as Q _f, Q _f=β ₁* Q ^l+ (1-β ₁) * Q ^r, wherein, β ₁represent Q ^lweights;

8. calculate L _organd R _orgabsolute difference image, take matrix representation as calculate L _disand R _disabsolute difference image, take matrix representation as wherein, " || " is the symbol that takes absolute value;

9. will with width image is divided into respectively the size of individual non-overlapping copies is 8 * 8 piece, then right with implement respectively svd, obtain for all in width image the singular value mapping graph that the corresponding singular value matrix by its each piece forms with the singular value mapping graph that the corresponding singular value matrix by its each piece forms, will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _org, for G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as will the matrix of coefficients of the singular value mapping graph obtaining after enforcement svd is designated as G _dis, for G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, q) locates, be designated as wherein, W _lRrepresent with wide, H _lRrepresent with height, 0≤p≤7,0≤q≤7;

10. calculate corresponding singular value mapping graph and the singular value deviation evaluation of estimate of corresponding singular value mapping graph, is designated as K, $K=\frac{64}{W_{\mathrm{LR}}\×H_{\mathrm{LR}}}\×\underset{n=0}{\overset{W_{\mathrm{LR}}\×H_{\mathrm{LR}}/64-1}{\mathrm{\Σ}}}\frac{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}(G_{\mathrm{org}}^n(p,p)\×|G_{\mathrm{org}}^n(p,p)-G_{\mathrm{dis}}^n(p,p)\left|\right)}{\underset{p=0}{\overset{7}{\mathrm{\Σ}}}{G}_{\mathrm{org}}^n(p,p)},$ Wherein, represent G _orgin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates, represent G _disin in the singular value matrix of n piece coordinate position be the singular value that (p, p) locates;

right with implement respectively svd, obtain respectively with each self-corresponding 2 orthogonal matrixes and 1 singular value matrix, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _organd V _org, will the singular value matrix of implementing to obtain after svd is designated as O _org, will 2 orthogonal matrixes implementing to obtain after svd are designated as respectively χ _disand V _dis, will the singular value matrix of implementing to obtain after svd is designated as O _dis,

calculate respectively with width image is deprived the residual matrix diagram after singular value, will the residual matrix diagram of depriving after singular value is designated as X _org, X _org=χ _org* Λ * V _org, will the residual matrix diagram of depriving after singular value is designated as X _dis, X _dis=χ _dis* Λ * V _dis, wherein, Λ representation unit matrix, the size of Λ and O _organd O _disin the same size;

calculate X _organd X _dismean bias rate, be designated as wherein, x represents X _organd X _disin the horizontal ordinate of pixel, y represents X _organd X _disin the ordinate of pixel;

calculate the stereo-picture S of distortion to be evaluated _diswith respect to original undistorted stereo-picture S _orgthree-dimensional perception evaluating deg amount, be designated as Q _s, wherein, τ represents constant, for regulate K and at Q _smiddle risen importance;

according to Q _fand Q _s, calculate the stereo-picture S of distortion to be evaluated _disimage quality evaluation score value, be designated as Q, Q=Q _f* (Q _s) ^ρ, wherein, ρ represents weight coefficient value.

2. a kind of three-dimensional image objective quality evaluation method based on structure distortion according to claim 1, is characterized in that 2. middle L of described step _organd L _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _lacquisition process be:

2.-a1, to L _orgmake level and vertical direction Sobel operator and process, obtain L _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l1}and Z _{v, l1}, then calculate L _orggradient magnitude figure, be designated as Z _l1, wherein, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l1}(x, y) represents Z _{h, l1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l1}(x, y) represents Z _{v, l1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l1wide, H' represents Z _l1height;

2.-a2, to L _dismake level and vertical direction Sobel operator and process, obtain L _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, l2}and Z _{v, l2}, then calculate L _disgradient magnitude figure, be designated as Z _l2, wherein, Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, l2}(x, y) represents Z _{h, l2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, l2}(x, y) represents Z _{v, l2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _l2wide, H' represents Z _l2height;

2.-a3, required threshold value T while calculating zoning, $T=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{l2}(x,y)),$ Wherein, α is constant, Z _l1(x, y) represents Z _l1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _l2(x, y) represents Z _l2middle coordinate position is the gradient magnitude of the pixel of (x, y);

2.-a4, by Z _l1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l1(i, j), by Z _l2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _l2(i, j), judgement Z _l1(i, j) >T or Z _l2whether (i, j) >T sets up, if set up, determines L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _l(i, j)=1, otherwise, determine L _organd L _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _l(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8);

Described step is middle R 2. _organd R _disthe coefficient matrices A of each self-corresponding sensitizing range matrix mapping graph _racquisition process be:

2.-b1, to R _orgmake level and vertical direction Sobel operator and process, obtain R _orghorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r1}and Z _{v, r1}, then calculate R _orggradient magnitude figure, be designated as Z _r1, wherein, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r1}(x, y) represents Z _{h, r1}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r1}(x, y) represents Z _{v, r1}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r1wide, H' represents Z _r1height;

2.-b2, to R _dismake level and vertical direction Sobel operator and process, obtain R _dishorizontal direction gradient image and vertical gradient image, be designated as respectively Z _{h, r2}and Z _{v, r2}, then calculate R _disgradient magnitude figure, be designated as Z _r2, wherein, Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _{h, r2}(x, y) represents Z _{h, r2}middle coordinate position is the horizontal direction Grad of the pixel of (x, y), Z _{v, r2}(x, y) represents Z _{v, r2}middle coordinate position is the vertical gradient value of the pixel of (x, y), 1≤x≤W', and 1≤y≤H', W' represents Z herein _r2wide, H' represents Z _r2height;

2.-b3, required threshold value T' while calculating zoning, $T^{\′}=\mathrm{\α}\×\frac{1}{W^{\′}\×H^{\′}}\×(\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r1}(x,y)+\underset{x=0}{\overset{W^{\′}}{\mathrm{\Σ}}}\underset{y=0}{\overset{H^{\′}}{\mathrm{\Σ}}}{Z}_{r2}(x,y)),$ Wherein, α is constant, Z _r1(x, y) represents Z _r1middle coordinate position is the gradient magnitude of the pixel of (x, y), Z _r2(x, y) represents Z _r2middle coordinate position is the gradient magnitude of the pixel of (x, y);

2.-b4, by Z _r1middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r1(i, j), by Z _r2middle coordinate position is that the gradient magnitude of the pixel of (i, j) is designated as Z _r2(i, j), judgement Z _r1(i, j) >T or Z _r2whether (i, j) >T sets up, if set up, determines R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to sensitizing range, and makes A _r(i, j)=1, otherwise, determine R _organd R _dismiddle coordinate position is that the pixel of (i, j) belongs to de-militarized zone, and makes A _r(i, j)=0, wherein, 0≤i≤(W-8), 0≤j≤(H-8).

3. a kind of three-dimensional image objective quality evaluation method based on structure distortion according to claim 1 and 2, is characterized in that 7. middle β of described step ₁acquisition process be:

7.-1, adopt n undistorted stereo-picture to set up its distortion stereographic map image set under the different distortion levels of different type of distortion, this distortion stereographic map image set comprises the stereo-picture of several distortions, wherein, and n >=1;

7.-2, utilize subjective quality assessment method to obtain the average subjective scoring difference of the stereo-picture of the concentrated every width distortion of distortion stereo-picture, be designated as DMOS, DMOS=100-MOS, wherein, MOS represents subjective scoring average, DMOS ∈ [0,100];

7.-3, according to step 1. to step operating process 6., the left visual point image of the stereo-picture of every width distortion that calculated distortion stereo-picture is concentrated is with respect to the evaluation of estimate Q of the sensitizing range of the left visual point image of the undistorted stereo-picture of correspondence _m,Levaluation of estimate Q with de-militarized zone _{nm, L};

7.-4, adopt Mathematical Fitting method matching distortion stereo-picture to concentrate average subjective scoring difference DMOS and the corresponding Q of the stereo-picture of distortion _m,Land Q _{nm, L}thereby, obtain β ₁value.