CN104408716A - Three-dimensional image quality objective evaluation method based on visual fidelity - Google Patents

Abstract

The invention discloses a three-dimensional image quality objective evaluation method based on visual fidelity. The method includes: in a training stage, selecting multiple original distortionless three-dimensional images to form a training image set, determining whether pixel points in the distortionless three-dimensional images belong to a shielding area or a matching area through area detection, and structuring a monocular vision dictionary table and a binocular vision dictionary table to the training image set through an unsupervised learning mode; in a testing stage, for testing three-dimensional images and the original distortionless three-dimensional images, estimating sparse coefficient array of each subblock, belonging to the shielding area and the matching area, in the testing three-dimensional images and the corresponding distortionless three-dimensional images according to the monocular vision dictionary table and the binocular vision dictionary table, calculating monocular image quality objective evaluation prediction value and binocular image quality objective evaluation prediction value through the sparse coefficient array, and finally combining to acquire an image quality evaluation predication value. The three-dimensional image quality objective evaluation method has the advantage that the acquired image quality objective evaluation predication value is highly uniform with a subjective evaluation value.

Description

The objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity

Technical field

The present invention relates to a kind of image quality evaluating method, especially relate to the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity.

Background technology

Along with developing rapidly of image coding technique and stereo display technique, stereo-picture technology receives to be paid close attention to and application more and more widely, has become a current study hotspot.Stereo-picture technology utilizes the binocular parallax principle of human eye, and binocular receives left visual point image from Same Scene and right visual point image independently of one another, is merged and forms binocular parallax, thus enjoy the stereo-picture with depth perception and realism by brain.Compared with single channel image, stereo-picture needs the picture quality simultaneously ensureing two passages, therefore carries out quality assessment to it and has very important significance.But current stereoscopic image quality lacks effective method for objectively evaluating and evaluates.Therefore, set up effective stereo image quality objective evaluation model tool to be of great significance.

Owing to affecting the many factors of stereo image quality, as left viewpoint and right viewpoint quality distortion situation, stereoscopic sensation are known the inside story condition, observer's visual fatigue etc., therefore how effectively carrying out stereo image quality evaluation is the difficulties needing solution badly.Existing method carrys out prediction and evaluation model by machine learning at present, but its computation complexity is higher, and training pattern needs the subjective assessment value predicting each evaluation map picture, and be not suitable for actual application scenario, has some limitations.Signal decomposes by rarefaction representation on known collection of functions, makes every effort to approach original signal with few basis function of trying one's best on transform domain.A key issue of rarefaction representation is exactly how effectively to construct the essential characteristic that dictionary carrys out token image.The dictionary construction algorithm proposed at present comprises: 1) have the dictionary construction method of learning process: trained obtaining dictionary information, as support vector machine etc. by machine learning; 2) without the dictionary construction method of learning process: directly utilize the feature of image to construct dictionary, as multiple dimensioned Gabor dictionary, multiple dimensioned Gauss's dictionary etc.Therefore, the dictionary how carried out without learning process constructs, and how to carry out quality estimation according to dictionary, and from dictionary, how to extract the visual signature of reflection eyefidelity, be all the technical matters needing emphasis to solve in stereo image quality evaluation study.

Summary of the invention

Technical matters to be solved by this invention is to provide the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity, and its computation complexity is low, and effectively can improve the correlativity between objective evaluation result and subjective perception.

The present invention solves the problems of the technologies described above adopted technical scheme: the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity, it is characterized in that comprising training stage and test phase two processes, and the described training stage comprises the following steps:

-1 1., choose N original undistorted stereo-picture composing training image set, be designated as { S _{i, org}| 1≤i≤N}, wherein, N>1, S _{i, org}represent { S _{i, org}| i-th in 1≤i≤N} original undistorted stereo-picture;

-2 1., utilize Region detection algorithms, judge { S _{i, org}| each pixel in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image belongs to occlusion area and still belongs to matching area;

-3 1., adopt Gabor filter, obtain { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image;

Then according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture;

Equally, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture;

Further, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, obtains { S _{i, org}| the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image;

1.-4, to { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture carry out non-overlapped point sub-block process;

Then K-SVD method is adopted, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _ncin a jth visual dictionary, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Equally, adopt K-SVD method, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _bfin a jth visual dictionary, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Described test phase comprises the following steps:

2. for any one secondary test stereo-picture S _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org;

According to step 1.-2 to step 1.-3 process, with identical operation obtain S _testand S _orgthe amplitude of each pixel under different center frequency and the different directions factor in respective left visual point image and right visual point image, and obtain S _testand S _orgthe amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in respective left visual point image and right visual point image;

By S _testleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, by S _orgleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; By S _testleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, by S _orgleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

According to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition occlusion area is correlated with; And { the S obtained according to the training stage _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition matching area is correlated with;

Two the proper vectors sparse coefficient matrix separately corresponding according to position in two proper vector set that occlusion area is correlated with, obtains S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block; And two proper vectors sparse coefficient matrix separately that in two proper vector set of being correlated with according to matching area, position is corresponding, obtain S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block;

According to S _testleft visual point image and right visual point image in belong to local objective evaluation metric and the S of each sub-block of occlusion area _testleft visual point image and right visual point image in belong to the local objective evaluation metric of each sub-block of matching area, obtain S _testpicture quality objective evaluation predicted value.

The detailed process of the Region detection algorithms of described step 1. in-2 is:

A1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1;

A2, the anaglyph adopting Block Matching Algorithm calculating to work as front left visual point image and work as between front right visual point image, be designated as { d _{i, org}(x, y) }, wherein, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, d _{i, org}(x, y) represents { d _{i, org}(x, y) } in coordinate position be the pixel value of the pixel of (x, y);

A3, basis { d _{i, org}(x, y) } in the pixel value of each pixel, judge still to belong to matching area when each pixel in front left visual point image belongs to occlusion area; For being the pixel of (x, y) when coordinate position in front left visual point image, if d _{i, org}(x, y)=255, then judge when coordinate position in front left visual point image belongs to occlusion area, if d as the pixel of (x, y) _{i, org}(x, y) ≠ 255, then judge when coordinate position in front left visual point image belongs to matching area as the pixel of (x, y);

A4, when in front right visual point image, the pixel corresponding with when each pixel belonging to matching area in front left visual point image is judged to belong to matching area; Then be judged to belong to occlusion area by when all pixels in front right visual point image except belonging to matching area;

A5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step a2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

Described step 1.-3 detailed process be:

B1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1;

B2, employing Gabor filter carry out filtering process to when front left visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front left visual point image

Equally, Gabor filter is adopted to carry out filtering process to when front right visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $G_{i,\mathrm{org}}^R(x,y;\mathrm{\ω},\mathrm{\θ});$

Above-mentioned, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, ω represents the centre frequency of adopted Gabor filter, ω ∈ Ω _ω, Ω _ωrepresent the set of the centre frequency of the Gabor filter adopted, θ represents the direction factor of adopted Gabor filter, θ ∈ Ω _θ, Ω _θrepresent the set of the direction factor of the Gabor filter adopted;

B3, basis are when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front left visual point image $E_{i,\mathrm{org}}^L(x,y;\mathrm{\ω},\mathrm{\θ})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\ω},\mathrm{\θ})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix;

Equally, according to the frequency response of each pixel under different center frequency and the different directions factor of working as in front right visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $E_{i,\mathrm{org}}^R(x,y;\mathrm{\ω},\mathrm{\θ})={\left|\right|G_{i,\mathrm{org}}^R(x,y;\mathrm{\ω},\mathrm{\θ})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix;

B4, basis are when front left visual point image with when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, calculate when front left visual point image with when the amplitude of each pixel in front right visual point image under different center frequency, the skew of different directions Summing Factor out of phase, by work as front left visual point image and when front right visual point image separately middle coordinate position be (x, y) pixel centre frequency be ω, the direction factor amplitude that to be θ and phase offset be under Δ ψ is designated as $E_{i,\mathrm{org}}^{\mathrm{LR}}(x,y;\mathrm{\ω},\mathrm{\θ},\mathrm{\Δ\ψ})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\ω},\mathrm{\θ})+e^{\mathrm{j\Δ\ψ}}\×G_{i,\mathrm{org}}^R(x,y;\mathrm{\ω},\mathrm{\θ})\left|\right|}_2^2,$ Wherein, Δ ψ ∈ Ω _{Δ ψ}, Ω _{Δ ψ}represent the set of the phase offset of the Gabor filter adopted, symbol " || || ₂" for asking for the 2-norm sign of matrix, e represents nature radix, j is imaginary unit;

B5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step b2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

Described step 1.-4 detailed process be:

C1, by { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture are divided into respectively individual size is the sub-block of the non-overlapping copies of 8 × 8, wherein, and W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N};

C2, determine { S _{i, org}| belong to all sub-blocks of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area;

And by { S _{i, org}| all sub-blocks in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image except the sub-block belonging to occlusion area are defined as belonging to matching area;

C3, acquisition { S _{i, org}| belong to the proper vector of each sub-block of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

And obtain { S _{i, org}| belong to the proper vector of each sub-block of matching area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

C4, by { S _{i, org}| all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { y _t| 1≤t≤M ₁, wherein, y _tfor { y _t| 1≤t≤M ₁in t proper vector, y _tdimension be 64 × N _ω× N _θ, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, M ₁represent { S _{i, org}| belong to total number of the sub-block of occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image,

And by { S _{i, org}| the proper vector belonging to all sub-blocks of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { z _t| 1≤t≤M ₂, wherein, z _tfor { z _t| 1≤t≤M ₂in t proper vector, z _tdimension be 64 × N _ω× N _θ× N _{Δ ψ}, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter, M ₂represent { S _{i, org}| belong to total number of the sub-block of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image, $M_2<\frac{W\&times;H\&times;N}{64};$

C5, employing K-SVD method are to { y _t| 1≤t≤M ₁carry out dictionary training and operation, obtain { y _t| 1≤t≤M ₁visual dictionary table, and by { y _t| 1≤t≤M ₁visual dictionary table as { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, d _ncsolved by K-SVD method obtain, constraint condition be: || x _t|| ₀≤ τ, wherein, D _ncdimension be (64 × N _ω× N _θ) × K, K represents total number of the dictionary of setting, K>=1, represent D _ncin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, y _ncdimension be (64 × N _ω× N _θ) × M ₁, y ₁represent { y _t| 1≤t≤M ₁in the 1st proper vector, y _trepresent { y _t| 1≤t≤M ₁in t proper vector, represent { y _t| 1≤t≤M ₁in M ₁individual proper vector, X _ncrepresent sparse matrix, x _ncdimension be K × M ₁, x ₁represent X _ncin the 1st row, x _trepresent X _ncin t row, represent X _ncin M ₁row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient;

And adopt K-SVD method to { z _t| 1≤t≤M ₂carry out dictionary training and operation, obtain { z _t| 1≤t≤M ₂visual dictionary table, and by { z _t| 1≤t≤M ₂visual dictionary table as { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, d _bfsolved by K-SVD method obtain, constraint condition be: || f _t|| ₀≤ τ, wherein, D _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × K, K represents total number of the dictionary of setting, K>=1, represent D _bfin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, z _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × M ₂, z ₁for { z _t| 1≤t≤M ₂in the 1st proper vector, z _tfor { z _t| 1≤t≤M ₂in t proper vector, for { z _t| 1≤t≤M ₂in M ₂individual proper vector, F _bfrepresent sparse matrix, f _bfdimension be K × M ₂, f ₁represent F _bfin the 1st row, f _trepresent F _bfin t row, represent F _bfin M ₂row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient.

τ=0.1 is got in described step c5.

Described step detailed process is 2.:

2.-1, by S _testleft visual point image be designated as L _test, by S _testright visual point image be designated as R _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org, by S _orgleft visual point image be designated as L _org, by S _orgright visual point image be designated as R _org;

-2 2., according to step 1.-2 process, judge L with identical operation _testand R _testin each pixel belong to occlusion area and still belong to matching area, and judge L _organd R _orgin each pixel belong to occlusion area and still belong to matching area;

-3 2., according to step 1.-3 process, obtain L with identical operation _testin amplitude, the R of each pixel under different center frequency and the different directions factor _testin amplitude, the L of each pixel under different center frequency and the different directions factor _testand R _testin the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase, and obtain L _orgin amplitude, the R of each pixel under different center frequency and the different directions factor _orgin amplitude, the L of each pixel under different center frequency and the different directions factor _organd R _orgin each pixel different center frequency, different directions Summing Factor out of phase skew under amplitude;

2.-4, to L _testand R _testand L _organd R _orgcarry out non-overlapped point sub-block process respectively;

Then L is determined _testand R _testand L _organd R _orgin belong to all sub-blocks of occlusion area, for L _testand R _testand L _organd R _orgin any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area; And by L _testand R _testand L _organd R _orgin all sub-blocks except the sub-block belonging to occlusion area be defined as belonging to matching area;

Then L is obtained _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of occlusion area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; And obtain L _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of matching area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Afterwards by L _testand R _testin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', test}| 1≤t'≤M ₁', and by L _testand R _testin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', test}| 1≤t'≤M ₂', wherein, y _{t', test}for { y _{t', test}| 1≤t'≤M ₁' in t' proper vector, y _{t', test}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _testand R _testin belong to total number of the sub-block of occlusion area, z _{t', test}for { z _{t', test}| 1≤t'≤M ₂' in t' proper vector, z _{t', test}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _testand R _testin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter;

Equally, by L _organd R _orgin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', org}| 1≤t'≤M ₁', and by L _organd R _orgin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', org}| 1≤t'≤M ₂', wherein, y _{t', org}for { y _{t', org}| 1≤t'≤M ₁' in t' proper vector, y _{t', org}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _organd R _orgin belong to total number of the sub-block of occlusion area, z _{t', org}for { z _{t', org}| 1≤t'≤M ₂' in t' proper vector, z _{t', org}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _organd R _orgin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter;

2.-5, according to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, obtain { y _{t', test}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector and { y _{t', org}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector, by y _{t', test}sparse coefficient matrix be designated as x _{t', test}, x _{t', test}=(D _nc) ^-1y _{t', test}, by y _{t', org}sparse coefficient matrix be designated as x _{t', org}, x _{t', org}=(D _nc) ^-1y _{t', org}, wherein, (D _nc) ^-1for D _ncinverse matrix;

According to { the S that the training stage obtains _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, obtain { z _{t', test}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector and { z _{t', org}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector, by z _{t', test}sparse coefficient matrix be designated as f _{t', test}, f _{t', test}=(D _bf) ^-1z _{t', test}, by z _{t', org}sparse coefficient matrix be designated as f _{t', org}, f _{t', org}=(D _bf) ^-1z _{t', org}, wherein, (D _bf) ^-1for D _bfinverse matrix;

2.-6, L is calculated _testand R _testin belong to the local objective evaluation metric of each sub-block of occlusion area, by L _testand R _testin the local objective evaluation metric of t' sub-block that belongs in all sub-blocks of occlusion area be designated as q _{t', test}, wherein, (x _{t', test}) ^tfor x _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters;

And calculate L _testand R _testin belong to the local objective evaluation metric of each sub-block of matching area, by L _testand R _testin belong to t' sub-block of matching area local objective evaluation metric be designated as p _{t', test}, wherein, (f _{t', test}) ^tfor f _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters;

2.-7, S is calculated _testmonocular image Objective Quality Assessment predicted value, be designated as and calculate S _testbinocular image Objective Quality Assessment predicted value, be designated as

2.-8, S is calculated _testpicture quality objective evaluation predicted value, be designated as Q, Q=w _nc× Q _nc+ (1-w _nc) × Q _bf, wherein, w _ncfor Q _ncweights proportion.

2. described step gets C=0.02 in-6.

2. described step gets w in-8 _nc=0.2.

Compared with prior art, the invention has the advantages that:

1) the inventive method is in the training stage, obtained in undistorted stereo-picture by method for detecting area and belong to the pixel of occlusion area and belong to the pixel of matching area, and by unsupervised learning mode to training plan image set structure monocular vision dictionary table and binocular vision dictionary table, computation vision dictionary table is not more needed at test phase, this avoid complicated machine learning training process, reduce computation complexity.

2) the inventive method is at test phase, according to the monocular vision dictionary table constructed and binocular vision dictionary table, the sparse coefficient matrix of each sub-block of occlusion area and matching area is belonged in the test stereo-picture of distortion estimator and the undistorted stereo-picture of its correspondence, estimate that the sparse coefficient matrix obtained can reflect eyefidelity characteristic well according to monocular vision dictionary table and binocular vision dictionary table, and calculate monocular image Objective Quality Assessment predicted value and binocular image Objective Quality Assessment predicted value by sparse coefficient matrix, the picture quality objective evaluation predicted value obtained and subjective assessment value maintain good consistance.

Accompanying drawing explanation

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

Fig. 2 is that the picture quality objective evaluation predicted value of every width distortion stereo-picture in the University Of Ningbo's stereo-picture storehouse utilizing the inventive method to obtain and mean subjective are marked the scatter diagram of difference;

Fig. 3 is that the picture quality objective evaluation predicted value of every width distortion stereo-picture in the LIVE stereo-picture storehouse utilizing the inventive method to obtain and mean subjective are marked the scatter diagram of difference.

Embodiment

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

The objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity that the present invention proposes, it totally realizes block diagram as shown in Figure 1, and it comprises training stage and test phase two processes, and the training stage comprises the following steps:

-1 1., choose N original undistorted stereo-picture composing training image set, be designated as { S _{i, org}| 1≤i≤N}, wherein, N>1, S _{i, org}represent { S _{i, org}| i-th in 1≤i≤N} original undistorted stereo-picture, symbol " { } " is set expression symbol.

In the specific implementation, the width number that original undistorted stereo-picture is chosen should be suitable, if the value of N is larger, then by training the precision of the visual dictionary table obtained also higher, but computation complexity is also higher, therefore gets N=10 in the present embodiment.

-2 1., utilize Region detection algorithms, judge { S _{i, org}| each pixel in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image belongs to occlusion area and still belongs to matching area.

In this particular embodiment, the detailed process of the Region detection algorithms of step 1. in-2 is:

A1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1.

A2, the anaglyph adopting the calculating of existing Block Matching Algorithm to work as front left visual point image and work as between front right visual point image, be designated as { d _{i, org}(x, y) }, wherein, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, d _{i, org}(x, y) represents { d _{i, org}(x, y) } in coordinate position be the pixel value of the pixel of (x, y).

A3, basis { d _{i, org}(x, y) } in the pixel value of each pixel, judge still to belong to matching area when each pixel in front left visual point image belongs to occlusion area; For being the pixel of (x, y) when coordinate position in front left visual point image, if d _{i, org}(x, y)=255, then judge when coordinate position in front left visual point image belongs to occlusion area, if d as the pixel of (x, y) _{i, org}(x, y) ≠ 255, then judge when coordinate position in front left visual point image belongs to matching area as the pixel of (x, y).

A4, when in front right visual point image, the pixel corresponding with when each pixel belonging to matching area in front left visual point image is judged to belong to matching area; Then be judged to belong to occlusion area by when all pixels in front right visual point image except belonging to matching area.

A5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step a2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

-3 1., adopt Gabor filter, obtain { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image.Then according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture.Equally, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture.Further, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, obtains { S _{i, org}| the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image.

In this particular embodiment, step 1.-3 detailed process be:

B1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1.

B2, employing Gabor filter carry out filtering process to when front left visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front left visual point image $G_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})=e_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})+{\mathrm{jo}}_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;}).$ Equally, Gabor filter is adopted to carry out filtering process to when front right visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})=e_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})+{\mathrm{jo}}_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;}).$ Above-mentioned, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, ω represents the centre frequency of adopted Gabor filter, ω ∈ Ω _ω, Ω _ωrepresent the set of the centre frequency of the Gabor filter adopted, Ω _ω={ 1.74,2.47,3.49,4.93,6.98}, θ represent the direction factor of adopted Gabor filter, θ ∈ Ω _θ, Ω _θrepresent the set of the direction factor of the Gabor filter adopted, ${\mathrm{\&Omega;}}_{\mathrm{\&theta;}}=\{0,\frac{\mathrm{\&pi;}}{4},\frac{\mathrm{\&pi;}}{2},\frac{3\mathrm{\&pi;}}{4},\mathrm{\&pi;},\frac{5\mathrm{\&pi;}}{4},\frac{3\mathrm{\&pi;}}{2},\frac{7\mathrm{\&pi;}}{4}\},$ for real part, for imaginary part, for real part, for imaginary part, j is imaginary unit.

B3, basis are when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front left visual point image $E_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix.Equally, according to the frequency response of each pixel under different center frequency and the different directions factor of working as in front right visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $E_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix.

B4, basis are when front left visual point image with when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, calculate when front left visual point image with when the amplitude of each pixel in front right visual point image under different center frequency, the skew of different directions Summing Factor out of phase, by work as front left visual point image and when front right visual point image separately middle coordinate position be (x, y) pixel centre frequency be ω, the direction factor amplitude that to be θ and phase offset be under Δ ψ is designated as $E_{i,\mathrm{org}}^{\mathrm{LR}}(x,y;\mathrm{\&omega;},\mathrm{\&theta;},\mathrm{\&Delta;\&psi;})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})+e^{\mathrm{j\&Delta;\&psi;}}\&times;G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, Δ ψ ∈ Ω _{Δ ψ}, Ω _{Δ ψ}represent the set of the phase offset of the Gabor filter adopted, ${\mathrm{\&Omega;}}_{\mathrm{\&Delta;\&psi;}}=\{-\mathrm{\&pi;},-\frac{7\mathrm{\&pi;}}{8},-\frac{3\mathrm{\&pi;}}{4},-\frac{5\mathrm{\&pi;}}{8},-\frac{\mathrm{\&pi;}}{2},-\frac{3\mathrm{\&pi;}}{8},-\frac{\mathrm{\&pi;}}{4},-\frac{\mathrm{\&pi;}}{8},0\},$ Symbol " || || ₂" for asking for the 2-norm sign of matrix, e represents nature radix, j is imaginary unit.

B5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step b2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

1.-4, to { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture carry out non-overlapped point sub-block process.Then K-SVD method is adopted, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _ncin a jth visual dictionary, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.Equally, adopt K-SVD method, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _bfin a jth visual dictionary, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.

At this, the value due to K is excessive there will be cluster phenomenon, and the value of K is too small there will be deficient cluster phenomenon, therefore gets K=128 in the present embodiment.

In this particular embodiment, step 1.-4 detailed process be:

C1, by { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture are divided into respectively individual size is the sub-block of the non-overlapping copies of 8 × 8, wherein, and W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}.

C2, determine { S _{i, org}| belong to all sub-blocks of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area.

And by { S _{i, org}| all sub-blocks in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image except the sub-block belonging to occlusion area are defined as belonging to matching area.

C3, acquisition { S _{i, org}| belong to the proper vector of each sub-block of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.

And obtain { S _{i, org}| belong to the proper vector of each sub-block of matching area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.

C4, by { S _{i, org}| all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { y _t| 1≤t≤M ₁, wherein, y _tfor { y _t| 1≤t≤M ₁in t proper vector, y _tdimension be 64 × N _ω× N _θ, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N in the present embodiment _ω=5, N _θ=8, M ₁represent { S _{i, org}| belong to total number of the sub-block of occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image, $M_1<\frac{W\&times;H\&times;N}{64}.$

And by { S _{i, org}| the proper vector belonging to all sub-blocks of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { z _t| 1≤t≤M ₂, wherein, z _tfor { z _t| 1≤t≤M ₂in t proper vector, z _tdimension be 64 × N _ω× N _θ× N _{Δ ψ}, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter, N in the present embodiment _ω=5, N _θ=8, N _{Δ ψ}=9, M ₂represent { S _{i, org}| belong to total number of the sub-block of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image,

C5, employing K-SVD method are to { y _t| 1≤t≤M ₁carry out dictionary training and operation, obtain { y _t| 1≤t≤M ₁visual dictionary table, and by { y _t| 1≤t≤M ₁visual dictionary table as { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, d _ncsolved by K-SVD method obtain, constraint condition be: || x _t|| ₀≤ τ, wherein, D _ncdimension be (64 × N _ω× N _θ) × K, K represents total number of the dictionary of setting, K>=1, represent D _ncin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, y _ncdimension be (64 × N _ω× N _θ) × M ₁, y ₁represent { y _t| 1≤t≤M ₁in the 1st proper vector, y _trepresent { y _t| 1≤t≤M ₁in t proper vector, represent { y _t| 1≤t≤M ₁in M ₁individual proper vector, X _ncrepresent sparse matrix, x _ncdimension be K × M ₁, x ₁represent X _ncin the 1st row, x _trepresent X _ncin t row, represent X _ncin M ₁row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient, gets τ=0.1 in the present embodiment.

And adopt K-SVD method to { z _t| 1≤t≤M ₂carry out dictionary training and operation, obtain { z _t| 1≤t≤M ₂visual dictionary table, and by { z _t| 1≤t≤M ₂visual dictionary table as { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, d _bfsolved by K-SVD method obtain, constraint condition be: || f _t|| ₀≤ τ, wherein, D _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × K, K represents total number of the dictionary of setting, K>=1, represent D _bfin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, z _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × M ₂, z ₁for { z _t| 1≤t≤M ₂in the 1st proper vector, z _tfor { z _t| 1≤t≤M ₂in t proper vector, for { z _t| 1≤t≤M ₂in M ₂individual proper vector, F _bfrepresent sparse matrix, f _bfdimension be K × M ₂, f ₁represent F _bfin the 1st row, f _trepresent F _bfin t row, represent F _bfin M ₂row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient, gets τ=0.1 in the present embodiment.

Test phase comprises the following steps:

2. for the distortion stereo-picture S of any one secondary test _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org.According to step 1.-2 to step 1.-3 process, with identical operation obtain S _testand S _orgthe amplitude of each pixel under different center frequency and the different directions factor in respective left visual point image and right visual point image, and obtain S _testand S _orgthe amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in respective left visual point image and right visual point image.By S _testleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, by S _orgleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; By S _testleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, by S _orgleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.According to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition occlusion area is correlated with; And { the S obtained according to the training stage _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition matching area is correlated with.Two the proper vectors sparse coefficient matrix separately corresponding according to position in two proper vector set that occlusion area is correlated with, obtains S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block; And two proper vectors sparse coefficient matrix separately that in two proper vector set of being correlated with according to matching area, position is corresponding, obtain S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block.According to S _testleft visual point image and right visual point image in belong to local objective evaluation metric and the S of each sub-block of occlusion area _testleft visual point image and right visual point image in belong to the local objective evaluation metric of each sub-block of matching area, obtain S _testpicture quality objective evaluation predicted value.

In this particular embodiment, step detailed process is 2.:

2.-1, by S _testleft visual point image be designated as L _test, by S _testright visual point image be designated as R _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org, by S _orgleft visual point image be designated as L _org, by S _orgright visual point image be designated as R _org.

-2 2., according to step 1.-2 process, judge L with identical operation _testand R _testin each pixel belong to occlusion area and still belong to matching area, and judge L _organd R _orgin each pixel belong to occlusion area and still belong to matching area.With L _testand R _testfor example, judge L _testand R _testin each pixel belong to the detailed process that occlusion area still belongs to matching area and be: calculate L _testwith R _testbetween anaglyph, be designated as { d _test(x, y) }, d _test(x, y) represents { d _test(x, y) } in coordinate position be the pixel value of the pixel of (x, y); For L _testmiddle coordinate position is the pixel of (x, y), if d _test(x, y)=255, then judge L _testmiddle coordinate position is that the pixel of (x, y) belongs to occlusion area, if d _test(x, y) ≠ 255, then judge L _testmiddle coordinate position is that the pixel of (x, y) belongs to matching area; At R _testin, will with L _testin belong to each pixel of matching area corresponding pixel be judged to belong to matching area, by R _testin all pixels except belonging to matching area be judged to belong to occlusion area.

-3 2., according to step 1.-3 process, obtain L with identical operation _testin amplitude, the R of each pixel under different center frequency and the different directions factor _testin amplitude, the L of each pixel under different center frequency and the different directions factor _testand R _testin the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase, and obtain L _orgin amplitude, the R of each pixel under different center frequency and the different directions factor _orgin amplitude, the L of each pixel under different center frequency and the different directions factor _organd R _orgin each pixel different center frequency, different directions Summing Factor out of phase skew under amplitude.With L _testand R _testfor example, the detailed process obtaining the amplitude of pixel is: adopt Gabor filter to L _testand R _testcarry out filtering process respectively, correspondence obtains L _testand R _testin the frequency response of each pixel under different center frequency and the different directions factor; Calculate L _testin the amplitude of each pixel under different center frequency and the different directions factor, by L _testmiddle coordinate position is that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ $E_{\mathrm{test}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{\mathrm{test}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ represent L _testmiddle coordinate position is the pixel of (x, y) be ω and direction factor be θ in centre frequency under frequency response; Calculate R _testin the amplitude of each pixel under different center frequency and the different directions factor, by R _testmiddle coordinate position is that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ $E_{\mathrm{test}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{\mathrm{test}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ represent R _testmiddle coordinate position is the pixel of (x, y) be ω and direction factor be θ in centre frequency under frequency response; Calculate L _testand R _testin each pixel different center frequency, different directions Summing Factor out of phase skew under amplitude, by L _testand R _testmiddle coordinate position be (x, y) pixel centre frequency be ω, direction factor be θ and phase offset be Δ ψ under amplitude correspondence be designated as $E_{\mathrm{test}}^{\mathrm{LR}}(x,y;\mathrm{\&omega;},\mathrm{\&theta;},\mathrm{\&Delta;\&psi;})={\left|\right|G_{\mathrm{test}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})+e^{\mathrm{j\&Delta;\&psi;}}\&times;G_{\mathrm{test}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2.$

2.-4, to L _testand R _testand L _organd R _orgcarry out non-overlapped point sub-block process respectively, by L _testand R _testand L _organd R _orgbe divided into respectively individual size is the sub-block of the non-overlapping copies of 8 × 8, S _testsize consistent with the size of the undistorted stereo-picture selected by the training stage.Then L is determined _testand R _testand L _organd R _orgin belong to all sub-blocks of occlusion area, for L _testand R _testand L _organd R _orgin any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area; And by L _testand R _testand L _organd R _orgin all sub-blocks except the sub-block belonging to occlusion area be defined as belonging to matching area.Then L is obtained _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of occlusion area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; And obtain L _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of matching area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block.Afterwards by L _testand R _testin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', test}| 1≤t'≤M ₁', and by L _testand R _testin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', test}| 1≤t'≤M ₂', wherein, y _{t', test}for { y _{t', test}| 1≤t'≤M ₁' in t' proper vector, y _{t', test}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _testand R _testin belong to total number of the sub-block of occlusion area, z _{t', test}for { z _{t', test}| 1≤t'≤M ₂' in t' proper vector, z _{t', test}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _testand R _testin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter, N in the present embodiment _ω=5, N _θ=8, N _{Δ ψ}=9.Equally, by L _organd R _orgin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', org}| 1≤t'≤M ₁', and by L _organd R _orgin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', org}| 1≤t'≤M ₂', wherein, y _{t', org}for { y _{t', org}| 1≤t'≤M ₁' in t' proper vector, y _{t', org}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _organd R _orgin belong to total number of the sub-block of occlusion area, z _{t', org}for { z _{t', org}| 1≤t'≤M ₂' in t' proper vector, z _{t', org}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _organd R _orgin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter, N in the present embodiment _ω=5, N _θ=8, N _{Δ ψ}=9.

2.-5, according to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, obtain { y _{t', test}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector and { y _{t', org}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector, by y _{t', test}sparse coefficient matrix be designated as x _{t', test}, x _{t', test}=(D _nc) ^-1y _{t', test}, by y _{t', org}sparse coefficient matrix be designated as x _{t', org}, x _{t', org}=(D _nc) ^-1y _{t', org}, wherein, (D _nc) ^-1for D _ncinverse matrix.

According to { the S that the training stage obtains _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, obtain { z _{t', test}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector and { z _{t', org}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector, by z _{t', test}sparse coefficient matrix be designated as f _{t', test}, f _{t', test}=(D _bf) ^-1z _{t', test}, by z _{t', org}sparse coefficient matrix be designated as f _{t', org}, f _{t', org}=(D _bf) ^-1z _{t', org}, wherein, (D _bf) ^-1for D _bfinverse matrix.

2.-6, L is calculated _testand R _testin belong to the local objective evaluation metric of each sub-block of occlusion area, by L _testand R _testin the local objective evaluation metric of t' sub-block that belongs in all sub-blocks of occlusion area be designated as q _{t', test}, wherein, (x _{t', test}) ^tfor x _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters, gets C=0.02 in the present embodiment.

And calculate L _testand R _testin belong to the local objective evaluation metric of each sub-block of matching area, by L _testand R _testin belong to t' sub-block of matching area local objective evaluation metric be designated as p _{t', test}, wherein, (f _{t', test}) ^tfor f _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters, gets C=0.02 in the present embodiment.

2.-7, S is calculated _testmonocular image Objective Quality Assessment predicted value, be designated as and calculate S _testbinocular image Objective Quality Assessment predicted value, be designated as Q _bf,

2.-8, S is calculated _testpicture quality objective evaluation predicted value, be designated as Q, Q=w _nc× Q _nc+ (1-w _nc) × Q _bf, wherein, w _ncfor Q _ncweights proportion, get w in the present embodiment _nc=0.2.

For further illustrating feasibility and the validity of the inventive method, test assessment is carried out to the inventive method.

Here, utilize 2 of evaluate image quality evaluating method conventional objective parameters as evaluation index, namely Pearson correlation coefficient (the Pearson linear correlation coefficient under non-linear regression condition, PLCC), Spearman related coefficient (Spearman rank order correlation coefficient, SRCC), PLCC reflects the accuracy of the objective evaluation result of distortion stereo-picture, and SRCC reflects its monotonicity.

Utilize the inventive method to calculate the picture quality objective evaluation predicted value of the every width distortion stereo-picture in the picture quality objective evaluation predicted value of the every width distortion stereo-picture in University Of Ningbo's stereo-picture storehouse and LIVE stereo-picture storehouse, recycle the mean subjective scoring difference that existing subjective evaluation method obtains the every width distortion stereo-picture in the mean subjective scoring difference of the every width distortion stereo-picture in University Of Ningbo's stereo-picture storehouse and LIVE stereo-picture storehouse.The picture quality objective evaluation predicted value of the distortion stereo-picture calculated by the inventive method is done five parameter Logistic function nonlinear fittings, PLCC and SRCC value is higher, illustrates that the objective evaluation result of the inventive method and mean subjective difference correlativity of marking is better.Table 1 and table 2 give the picture quality objective evaluation predicted value of distortion stereo-picture that adopts the inventive method to obtain and mean subjective and to mark Pearson correlation coefficient between difference and Spearman related coefficient.As can be seen from Table 1 and Table 2, final picture quality objective evaluation predicted value and the mean subjective correlativity of marking between difference of the distortion stereo-picture adopting the inventive method to obtain are very high, the result indicating objective evaluation result and human eye subjective perception is more consistent, is enough to the validity that the inventive method is described.

Fig. 2 gives the scatter diagram that the picture quality objective evaluation predicted value of the every width distortion stereo-picture in the University Of Ningbo's stereo-picture storehouse utilizing the inventive method to obtain and mean subjective mark difference, Fig. 3 gives the scatter diagram that the picture quality objective evaluation predicted value of the every width distortion stereo-picture in the LIVE stereo-picture storehouse utilizing the inventive method to obtain and mean subjective mark difference, loose point is more concentrated, illustrates that the consistance of objective evaluation result and subjective perception is better.As can be seen from Fig. 2 and Fig. 3, adopt the scatter diagram that obtains of the inventive method more concentrated, and the goodness of fit between subjective assessment data is higher.

The Pearson correlation coefficient that picture quality objective evaluation predicted value and the mean subjective of the distortion stereo-picture that table 1 utilizes the inventive method to obtain are marked between difference compares

The Spearman related coefficient that picture quality objective evaluation predicted value and the mean subjective of the distortion stereo-picture that table 2 utilizes the inventive method to obtain are marked between difference compares

Claims (8)

1. an objective evaluation method for quality of stereo images for view-based access control model fidelity, it is characterized in that comprising training stage and test phase two processes, the described training stage comprises the following steps:

-1 1., choose N original undistorted stereo-picture composing training image set, be designated as { S _{i, org}| 1≤i≤N}, wherein, N>1, S _{i, org}represent { S _{i, org}| i-th in 1≤i≤N} original undistorted stereo-picture;

-2 1., utilize Region detection algorithms, judge { S _{i, org}| each pixel in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image belongs to occlusion area and still belongs to matching area;

-3 1., adopt Gabor filter, obtain { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image;

Then according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture;

Equally, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture, obtains { S _{i, org}| the amplitude of each pixel under different center frequency and the different directions factor in the right visual point image of every in 1≤i≤N} original undistorted stereo-picture;

Further, according to { S _{i, org}| the frequency response of each pixel under different center frequency and the different directions factor in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, obtains { S _{i, org}| the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image;

1.-4, to { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture carry out non-overlapped point sub-block process;

Then K-SVD method is adopted, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _ncin a jth visual dictionary, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Equally, adopt K-SVD method, to by { S _{i, org}| dictionary training and operation is carried out in the proper vector set that all sub-blocks proper vector separately belonging to matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image is formed, and obtains { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, wherein, the initial value of j is total number that 1, K represents the dictionary of setting, K>=1, represent D _bfin a jth visual dictionary, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Described test phase comprises the following steps:

2. for any one secondary test stereo-picture S _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org;

According to step 1.-2 to step 1.-3 process, with identical operation obtain S _testand S _orgthe amplitude of each pixel under different center frequency and the different directions factor in respective left visual point image and right visual point image, and obtain S _testand S _orgthe amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase in respective left visual point image and right visual point image;

By S _testleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, by S _orgleft visual point image and right visual point image in belong to occlusion area all sub-blocks proper vector separately form a proper vector set, any one sub-block belonging to occlusion area is have the pixel belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of occlusion area is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; By S _testleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, by S _orgleft visual point image and right visual point image in belong to all sub-blocks of matching area proper vector form a proper vector set, any one sub-block belonging to matching area is the pixel not belonging to occlusion area in this sub-block, and the proper vector belonging to any one sub-block of matching area is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

According to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition occlusion area is correlated with; And { the S obtained according to the training stage _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, the sparse coefficient matrix of each proper vector in two proper vector set that acquisition matching area is correlated with;

Two the proper vectors sparse coefficient matrix separately corresponding according to position in two proper vector set that occlusion area is correlated with, obtains S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block; And two proper vectors sparse coefficient matrix separately that in two proper vector set of being correlated with according to matching area, position is corresponding, obtain S _testleft visual point image and right visual point image in the local objective evaluation metric of corresponding sub-block;

According to S _testleft visual point image and right visual point image in belong to local objective evaluation metric and the S of each sub-block of occlusion area _testleft visual point image and right visual point image in belong to the local objective evaluation metric of each sub-block of matching area, obtain S _testpicture quality objective evaluation predicted value.

2. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 1, is characterized in that the detailed process of the Region detection algorithms of described step 1. in-2 is:

A1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1;

A2, the anaglyph adopting Block Matching Algorithm calculating to work as front left visual point image and work as between front right visual point image, be designated as { d _{i, org}(x, y) }, wherein, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, d _{i, org}(x, y) represents { d _{i, org}(x, y) } in coordinate position be the pixel value of the pixel of (x, y);

A3, basis { d _{i, org}(x, y) } in the pixel value of each pixel, judge still to belong to matching area when each pixel in front left visual point image belongs to occlusion area; For being the pixel of (x, y) when coordinate position in front left visual point image, if d _{i, org}(x, y)=255, then judge when coordinate position in front left visual point image belongs to occlusion area, if d as the pixel of (x, y) _{i, org}(x, y) ≠ 255, then judge when coordinate position in front left visual point image belongs to matching area as the pixel of (x, y);

A4, when in front right visual point image, the pixel corresponding with when each pixel belonging to matching area in front left visual point image is judged to belong to matching area; Then be judged to belong to occlusion area by when all pixels in front right visual point image except belonging to matching area;

A5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step a2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

3. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 1 and 2, it is characterized in that described step 1.-3 detailed process be:

B1, by { S _{i, org}| current pending i-th original undistorted stereo-picture S in 1≤i≤N} _{i, org}be defined as current stereo-picture, by current stereo-picture S _{i, org}left visual point image be defined as when front left visual point image, by current stereo-picture S _{i, org}right visual point image be defined as when front right visual point image, wherein, the initial value of i is 1;

B2, employing Gabor filter carry out filtering process to when front left visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front left visual point image

Equally, Gabor filter is adopted to carry out filtering process to when front right visual point image, obtain when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as in the centre frequency frequency response that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;});$

Above-mentioned, 1≤x≤W, 1≤y≤H, W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N}, ω represents the centre frequency of adopted Gabor filter, ω ∈ Ω _ω, Ω _ωrepresent the set of the centre frequency of the Gabor filter adopted, θ represents the direction factor of adopted Gabor filter, θ ∈ Ω _θ, Ω _θrepresent the set of the direction factor of the Gabor filter adopted;

B3, basis are when the frequency response of each pixel under different center frequency and the different directions factor in front left visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front left visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front left visual point image $E_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix;

Equally, according to the frequency response of each pixel under different center frequency and the different directions factor of working as in front right visual point image, calculate when the amplitude of each pixel under different center frequency and the different directions factor in front right visual point image, be that the pixel of (x, y) is designated as at the centre frequency amplitude that to be ω and direction factor be under θ by working as coordinate position in front right visual point image $E_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})={\left|\right|G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, symbol " || || ₂" for asking for the 2-norm sign of matrix;

B4, basis are when front left visual point image with when the frequency response of each pixel under different center frequency and the different directions factor in front right visual point image, calculate when front left visual point image with when the amplitude of each pixel in front right visual point image under different center frequency, the skew of different directions Summing Factor out of phase, by work as front left visual point image and when front right visual point image separately middle coordinate position be (x, y) pixel centre frequency be ω, the direction factor amplitude that to be θ and phase offset be under Δ ψ is designated as $E_{i,\mathrm{org}}^{\mathrm{LR}}(x,y;\mathrm{\&omega;},\mathrm{\&theta;},\mathrm{\&Delta;\&psi;})={\left|\right|G_{i,\mathrm{org}}^L(x,y;\mathrm{\&omega;},\mathrm{\&theta;})+e^{\mathrm{j\&Delta;\&psi;}}\&times;G_{i,\mathrm{org}}^R(x,y;\mathrm{\&omega;},\mathrm{\&theta;})\left|\right|}_2^2,$ Wherein, Δ ψ ∈ Ω _{Δ ψ}, Ω _{Δ ψ}represent the set of the phase offset of the Gabor filter adopted, symbol " || || ₂" for asking for the 2-norm sign of matrix, e represents nature radix, j is imaginary unit;

B5, make i=i+1, by { S _{i, org}| in 1≤i≤N}, next pending original undistorted stereo-picture is as current stereo-picture, using the left visual point image of current stereo-picture as working as front left visual point image, using the right visual point image of current stereo-picture as working as front right visual point image, then return step b2 to continue to perform, until { S _{i, org}| till all original undistorted stereo-picture in 1≤i≤N} is disposed, wherein, "=" in i=i+1 is assignment.

4. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 3, it is characterized in that described step 1.-4 detailed process be:

C1, by { S _{i, org}| left visual point image and the right visual point image of every in 1≤i≤N} original undistorted stereo-picture are divided into respectively individual size is the sub-block of the non-overlapping copies of 8 × 8, wherein, and W represents { S _{i, org}| the width of the original undistorted stereo-picture in 1≤i≤N}, H represents { S _{i, org}| the height of the original undistorted stereo-picture in 1≤i≤N};

C2, determine { S _{i, org}| belong to all sub-blocks of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area;

And by { S _{i, org}| all sub-blocks in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image except the sub-block belonging to occlusion area are defined as belonging to matching area;

C3, acquisition { S _{i, org}| belong to the proper vector of each sub-block of occlusion area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

And obtain { S _{i, org}| belong to the proper vector of each sub-block of matching area in the left visual point image of every in 1≤i≤N} original undistorted stereo-picture and right visual point image, for S _{i, org}left visual point image and right visual point image in belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

C4, by { S _{i, org}| all sub-blocks proper vector separately belonging to occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { y _t| 1≤t≤M ₁, wherein, y _tfor { y _t| 1≤t≤M ₁in t proper vector, y _tdimension be 64 × N _ω× N _θ, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, M ₁represent { S _{i, org}| belong to total number of the sub-block of occlusion area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image,

And by { S _{i, org}| the proper vector belonging to all sub-blocks of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image forms a proper vector set, is designated as { z _t| 1≤t≤M ₂, wherein, z _tfor { z _t| 1≤t≤M ₂in t proper vector, z _tdimension be 64 × N _ω× N _θ× N _{Δ ψ}, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter, M ₂represent { S _{i, org}| belong to total number of the sub-block of matching area in the left visual point image of all original undistorted stereo-picture in 1≤i≤N} and right visual point image, $M_2<\frac{W\&times;H\&times;N}{64};$

C5, employing K-SVD method are to { y _t| 1≤t≤M ₁carry out dictionary training and operation, obtain { y _t| 1≤t≤M ₁visual dictionary table, and by { y _t| 1≤t≤M ₁visual dictionary table as { S _{i, org}| the monocular vision dictionary table of 1≤i≤N}, is designated as D _nc, d _ncsolved by K-SVD method obtain, constraint condition be: wherein, D _ncdimension be (64 × N _ω× N _θ) × K, K represents total number of the dictionary of setting, K>=1, represent D _ncin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, y _ncdimension be (64 × N _ω× N _θ) × M ₁, y ₁represent { y _t| 1≤t≤M ₁in the 1st proper vector, y _trepresent { y _t| 1≤t≤M ₁in t proper vector, represent { y _t| 1≤t≤M ₁in M ₁individual proper vector, X _ncrepresent sparse matrix, x _ncdimension be K × M ₁, x ₁represent X _ncin the 1st row, x _trepresent X _ncin t row, represent X _ncin M ₁row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient;

And adopt K-SVD method to { z _t| 1≤t≤M ₂carry out dictionary training and operation, obtain { z _t| 1≤t≤M ₂visual dictionary table, and by { z _t| 1≤t≤M ₂visual dictionary table as { S _{i, org}| the binocular vision dictionary table of 1≤i≤N}, is designated as D _bf, d _bfsolved by K-SVD method obtain, constraint condition be: wherein, D _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × K, K represents total number of the dictionary of setting, K>=1, represent D _bfin a jth dictionary, min () for getting minimum value function, symbol " || || ₂" for asking for the 2-norm sign of matrix, z _bfdimension be (64 × N _ω× N _θ× N _{Δ ψ}) × M ₂, z ₁for { z _t| 1≤t≤M ₂in the 1st proper vector, z _tfor { z _t| 1≤t≤M ₂in t proper vector, for { z _t| 1≤t≤M ₂in M ₂individual proper vector, F _bfrepresent sparse matrix, f _bfdimension be K × M ₂, f ₁represent F _bfin the 1st row, f _trepresent F _bfin t row, represent F _bfin M ₂row, symbol " [] " is vector representation symbol, represent existence t, symbol " || || ₀" for asking for the 0-norm sign of matrix, τ is error coefficient.

5. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 4, is characterized in that getting τ=0.1 in described step c5.

6. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 4, is characterized in that described step detailed process is 2.:

2.-1, by S _testleft visual point image be designated as L _test, by S _testright visual point image be designated as R _test, by S _testcorresponding original undistorted stereo-picture is designated as S _org, by S _orgleft visual point image be designated as L _org, by S _orgright visual point image be designated as R _org;

-2 2., according to step 1.-2 process, judge L with identical operation _testand R _testin each pixel belong to occlusion area and still belong to matching area, and judge L _organd R _orgin each pixel belong to occlusion area and still belong to matching area;

-3 2., according to step 1.-3 process, obtain L with identical operation _testin amplitude, the R of each pixel under different center frequency and the different directions factor _testin amplitude, the L of each pixel under different center frequency and the different directions factor _testand R _testin the amplitude of each pixel under different center frequency, the skew of different directions Summing Factor out of phase, and obtain L _orgin amplitude, the R of each pixel under different center frequency and the different directions factor _orgin amplitude, the L of each pixel under different center frequency and the different directions factor _organd R _orgin each pixel different center frequency, different directions Summing Factor out of phase skew under amplitude;

2.-4, to L _testand R _testand L _organd R _orgcarry out non-overlapped point sub-block process respectively;

Then L is determined _testand R _testand L _organd R _orgin belong to all sub-blocks of occlusion area, for L _testand R _testand L _organd R _orgin any one sub-block, if there is the pixel belonging to occlusion area in this sub-block, then determine that this sub-block belongs to occlusion area; And by L _testand R _testand L _organd R _orgin all sub-blocks except the sub-block belonging to occlusion area be defined as belonging to matching area;

Then L is obtained _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of occlusion area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of occlusion area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies and direction factor in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block; And obtain L _testand R _testand L _organd R _orgin belong to the proper vector of each sub-block of matching area, for L _testand R _testand L _organd R _orgin belong to any one sub-block of matching area, the proper vector of this sub-block is that the amplitude of all pixels under all centre frequencies, direction factor and phase offset in this sub-block is formed by the sequencing arrangement of the coordinate position of each pixel in this sub-block;

Afterwards by L _testand R _testin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', test}| 1≤t'≤M ₁' _,and by L _testand R _testin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', test}| 1≤t'≤M ₂', wherein, y _{t', test}for { y _{t', test}| 1≤t'≤M ₁' in t' proper vector, y _{t', test}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _testand R _testin belong to total number of the sub-block of occlusion area, z _{t', test}for { z _{t', test}| 1≤t'≤M ₂' in t' proper vector, z _{t', test}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _testand R _testin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter;

Equally, by L _organd R _orgin belong to occlusion area all sub-blocks proper vector separately form a proper vector set, be designated as { y _{t', org}| 1≤t'≤M ₁', and by L _organd R _orgin belong to all sub-blocks of matching area proper vector form a proper vector set, be designated as { z _{t', org}| 1≤t'≤M ₂', wherein, y _{t', org}for { y _{t', org}| 1≤t'≤M ₁' in t' proper vector, y _{t', org}dimension be 64 × N _ω× N _θ, at this M ₁' represent L _organd R _orgin belong to total number of the sub-block of occlusion area, z _{t', org}for { z _{t', org}| 1≤t'≤M ₂' in t' proper vector, z _{t', org}dimension be 64 × N _ω× N _θ× N _{Δ ψ}, at this M ₂' represent L _organd R _orgin belong to total number of the sub-block of matching area, N _ωrepresent total number of the centre frequency of Gabor filter, N _θrepresent total number of the direction factor of Gabor filter, N _{Δ ψ}represent total number of the phase offset of Gabor filter;

2.-5, according to { the S that the training stage obtains _{i, org}| the monocular vision dictionary table D of 1≤i≤N} _nc, obtain { y _{t', test}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector and { y _{t', org}| 1≤t'≤M ₁' in the sparse coefficient matrix of each proper vector, by y _{t', test}sparse coefficient matrix be designated as x _{t', test}, x _{t', test}=(D _nc) ^-1y _{t', test}, by y _{t', org}sparse coefficient matrix be designated as x _{t', org}, x _{t', org}=(D _nc) ^-1y _{t', org}, wherein, (D _nc) ^-1for D _ncinverse matrix;

According to { the S that the training stage obtains _{i, org}| the binocular vision dictionary table D of 1≤i≤N} _bf, obtain { z _{t', test}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector and { z _{t', org}| 1≤t'≤M ₂' in the sparse coefficient matrix of each proper vector, by z _{t', test}sparse coefficient matrix be designated as f _{t', test}, f _{t', test}=(D _bf) ^-1z _{t', test}, by z _{t', org}sparse coefficient matrix be designated as f _{t', org}, f _{t', org}=(D _bf) ^-1z _{t', org}, wherein, (D _bf) ^-1for D _bfinverse matrix;

2.-6, L is calculated _testand R _testin belong to the local objective evaluation metric of each sub-block of occlusion area, by L _testand R _testin the local objective evaluation metric of t' sub-block that belongs in all sub-blocks of occlusion area be designated as q _{t', test}, wherein, (x _{t', test}) ^tfor x _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters;

And calculate L _testand R _testin belong to the local objective evaluation metric of each sub-block of matching area, by L _testand R _testin belong to t' sub-block of matching area local objective evaluation metric be designated as p _{t', test}, wherein, (f _{t', test}) ^tfor f _{t', test}transposed matrix, symbol " || || ₂" for asking for the 2-norm sign of matrix, C is controling parameters;

2.-7, S is calculated _testmonocular image Objective Quality Assessment predicted value, be designated as Q _nc, and calculate S _testbinocular image Objective Quality Assessment predicted value, be designated as Q _bf,

2.-8, S is calculated _testpicture quality objective evaluation predicted value, be designated as Q, Q=w _nc× Q _nc+ (1-w _nc) × Q _bf, wherein, w _ncfor Q _ncweights proportion.

7. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 6, is characterized in that 2. described step gets C=0.02 in-6.

8. the objective evaluation method for quality of stereo images of a kind of view-based access control model fidelity according to claim 7, is characterized in that 2. described step gets w in-8 _nc=0.2.