CN109523506A - The complete of view-based access control model specific image feature enhancing refers to objective evaluation method for quality of stereo images - Google Patents
The complete of view-based access control model specific image feature enhancing refers to objective evaluation method for quality of stereo images Download PDFInfo
- Publication number
- CN109523506A CN109523506A CN201811109216.6A CN201811109216A CN109523506A CN 109523506 A CN109523506 A CN 109523506A CN 201811109216 A CN201811109216 A CN 201811109216A CN 109523506 A CN109523506 A CN 109523506A
- Authority
- CN
- China
- Prior art keywords
- image
- view
- stereo
- channel
- right view
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
Complete the invention discloses a kind of enhancing of view-based access control model specific image feature refers to objective evaluation method for quality of stereo images.First, stereo-picture is converted into YIQ color space from RGB color, luminance components are extracted from the channel Y, obtain disparity map and vision significance figure, then image co-registration is carried out to the left and right view in the channel Y and obtains intermediate image, and then edge/texture and depth information feature that vision significantly increases are extracted, similarity measurement is carried out, corresponding Measure Indexes are obtained.Secondly, extracting corresponding color information feature from the channel I and Q of stereo-picture, binocular fusion and similarity measurement are carried out, the Measure Indexes for the color information that vision significantly increases are obtained.All Measure Indexes are finally supported vector regression training prediction, obtain objective quality scores.Experiment shows that stereo image quality proposed by the invention is objectively evaluated with subjective assessment with good consistency, and performance is better than most of existing stereo image quality evaluation method.
Description
Technical field
The invention belongs to image processing techniques and computer vision field more particularly to a kind of view-based access control model specific image are special
The complete of sign enhancing refers to objective evaluation method for quality of stereo images.
Background technique
Video and image always occur in image sampling, compression, transmission and reconstruction processes various distortions and are made
The quality of video and image reduces.Nowadays the mankind are higher and higher to the quality requirement of video and image information, to video and image
Transmission process carry out picture quality real-time monitoring requirement it is also more more and more urgent.And with the hair of three-dimensional video-frequency technology
Exhibition, is measured in real time stereo image quality and also becomes a urgent problem to be solved.Stereo image quality assessment technique exists
It is particularly important under this overall situation.It is each figure since subjective stereo image quality evaluation method requires human viewer
As being offered to subjective mass fraction.The defects of these methods are time consuming nature and weak projectivity, can not be applied to actual end and arrives
In the middle of the video detection system at end, so very it is necessary to develop objective effective stereo image quality evaluation method, with reality
Quality that is now automatic, efficiently, objectively evaluating stereo-picture.One good stereo image quality evaluation method should have well
Prediction stereo image quality ability, and with the result of subjective measurement have high consistency.
Objective stereo image quality method can be divided into three kinds of classifications according to the difference of object of reference: full reference, half refer to
With no reference.Complete original reference image is needed to carry out quality as object of reference with reference to stereo image quality evaluation method entirely
Evaluation, the object of reference of half reference image quality appraisement only has the Partial Feature information of original image, and non-reference picture quality is commented
Valence algorithm needs not refer to object, only analyzes distorted image, obtains its mass fraction.If divided according to experimental considerations,
Stereo image quality evaluation can be divided into three classes again: the first kind is that the algorithm of existing 2D is directly applied directly to 3D rendering matter
In amount evaluation；Second class is to take into account the exclusive depth information of 3D, is first carried out respectively to the left and right view of stereo-picture special
Sign is extracted, then carries out Fusion Features；Third class method is to carry out the fused image of binocular vision, then carry out feature extraction and matter
Measure the acquisition of score.The first kind is most simple in these three types of algorithms, and effect is also worst, and third class algorithm framework is big due to meeting the mankind
Binocular vision mechanism in brain and be widely studied, obtain pretty good result.But due to stereo-picture vision mode system
It is still not perfect, therefore three-dimensional image objective quality evaluation is still the hot and difficult issue studied now.
Summary of the invention
The invention discloses a kind of visual saliency maps as that feature enhances is complete with reference to objective evaluation method for quality of stereo images.
The purpose is to utilize vision significance model, the edge, texture and the color character that extract intermediate stereo-picture are assisted, to realize
Measurement and evaluation to stereo image quality are completed in mapping to stereo image quality.
The technical solution adopted by the present invention is that:
Firstly, the left and right view of stereo pairs is converted into YIQ color space from RGB color, wherein logical to Y
The gray scale stereo pairs (luminance components) that road obtains are handled to obtain corresponding disparity map, then respectively to gray scale perspective view
As pair left and right view carry out image co-registration obtain middle reference and distorted image, utilize based on spectrum residual error vision significance mould
Type obtains left and right view Saliency maps respectively, and integrates and obtain 3D mesopic vision Saliency maps, from middle reference and distorted image
It extracts edge/texture information feature and extracts depth information feature from the disparity map of stereo pairs, carry out similarity measurements
Amount, obtains the Measure Indexes for each visual information feature that vision significantly increases.Secondly because color information is also composition image
Important information carries out double so corresponding color information feature is extracted in the channel I and Q (colour component) for stereo pairs
Mesh fusion and similarity measurement, obtain the Measure Indexes of color information.Finally by the Measure Indexes of all features be supported to
Regression training prediction is measured, objective quality scores are obtained, realizes the mapping to stereo image quality, is completed to stereo image quality
Evaluation.
The technical solution adopted by the present invention to solve the technical problems is as follows:
Step (1) input is with reference to stereo pairs and distortion stereo pairs, wherein each stereo pairs respectively include
Left view and right view image；
Step (2) converts the color space of the stereo pairs in step (1), is converted into from rgb color space
YIQ color space, wherein the channel Y shows that the gray component of image, I and Q indicate the colour component of image.Specific conversion formula
It is as follows:
Step (3) constructs Log Gabor filter model, to the gray scale perspective view obtained in step (2) by the channel Y
As to convolution algorithm processing is carried out, respectively obtaining reference and being distorted the energy response figure of stereo image pair or so view；
Log Gabor filter hLGExpression formula it is as follows:
Wherein, f0And θ0Indicate centre frequency and the azimuth of Log Gabor filter, σθAnd σfRespectively represent filter
Azimuth bandwidth and radial bandwidth, f and θ respectively represent radial coordinate and the azimuth of filter；
By Log Gabor filter and after referring to and being distorted stereo image pair or so view progress convolution, obtain corresponding
Energy response figure F (x, y), expression formula is as follows:
Wherein, I (x, y) is the left view or right view of reference and distortion stereo pairs gray component,For convolution fortune
It calculates；
Step (4) extracts parallax with distortion stereo pairs to the gray reference stereo pairs that step (2) obtains respectively
Scheme Dref(x, y) and Ddis(x, y) utilizes the left and right view for the gray reference stereo pairs that step (2) obtains residual based on composing
The conspicuousness model of difference extracts left and right Saliency maps SL respectivelysr(x, y) and SRsr(x,y)；
Step (5) constructs 3D vision significance figure S3D(x, y), specific expression formula are as follows:
S3D(x, y)=ω1×SLsr(x,y)+ω2×SRsr(x,y)+ω3×CB(x,y)+ω4×Dref(x,y) (5-2)
Wherein ω1-ω4For different weight factors,
CB (x, y) indicates central point offsetting mechanism.
The right view for the gray scale stereo pairs that step (6) obtains step (2) is according to the parallax obtained in step (4)
The level that the parallax value of figure carries out pixel moves to right, the calibration right view I that construction and left view pixel coordinate pair are answeredR((x+d),
Y), it is then based on the available left view of Log Gabor filter model described in step (3) and calibrates the energy sound of right view
Ying Tu calculates normalized left view weight map WL(x, y) and calibration right view weight map WR((x+d), y), expression
It is as follows:
Wherein, FL(x, y) and FR((x+d), y) is respectively the energy sound of the left view that step (3) obtains and calibration right view
Ying Tu, d are the disparity map D that step (4) are calculatedrefThe parallax value of respective coordinates in (x, y)；
The left view and step (6) of gray reference and distortion stereo pairs that step (7) is based in step (2) obtain
Reference and be distorted stereo pairs calibration right view and normalized left view weight map and calibration right view weight map,
It is realized using binocular view Fusion Model to the image co-registration of stereo-picture, respectively obtains reference and distortion intermediate image；
The formula of binocular view fusion is as follows:
CI (x, y)=WL(x,y)×IL(x,y)+WR((x+d),y)×IR((x+d),y) (7-1)
Wherein, CI (x, y) is the fused intermediate image of binocular view, IL(x, y) and IR((x+d), y) is respectively ash
Spend the left view and calibration right view of stereo pairs；
Step (8) is referred to the middle gray that step (7) obtains and distorted image extracts edge and textural characteristics respectively；
The extraction of marginal information feature carries out process of convolution by Sobel operator and by altimetric image, obtains comprising edge contour
The gradient map of information, the expression formula using the marginal information feature of Sobel operator extraction middle reference and distorted image are as follows:
Wherein, f (x, y) is the left/right view of stereo pairs,For convolution algorithm, GxAnd GyIt is 3 × 3 Sobel water
Flat die plate and vertical formwork are respectively intended to the horizontal edge and vertical edge of detection image, and template expression formula is as follows:
The extraction of texture information feature.Using local binary patterns LBP, the expression formula of LBP is as follows:
Wherein, gcIt is the gray value of the central pixel point of image, gpIt is the gray value of the neighbor pixel of image, from center
The front-right of pixel rotates counterclockwise is followed successively by 0,1,2 ..., and P, x and y represent the coordinate value of central pixel point, and P indicates adjacent
The number of pixel, sgn (x) are jump functions；
The visual information feature and step (5) for the middle reference and distorted image that step (9) extracts step (8) are established
The multiplication put pixel-by-pixel of vision significance figure, obtain the visual information feature of vision significance enhancing, expression
It is as follows:
GMSR(x, y)=GMR(x,y)*S3D(x,y)GMSD(x, y)=GMD(x,y)*S3D(x,y) (9-1)
TISR(x, y)=TIR(x,y)*S3D(x,y)TISD(x, y)=TID(x,y)*S3D(x,y) (9-2)
Wherein, GMR(x, y) and TIR(x, y) is the edge and texture feature information of middle reference image, GM respectivelyD(x,y)
And TID(x, y) is the edge and texture feature information of intermediate distorted image respectively；S3DView after the integration obtained for step (5)
Feel Saliency maps；
Step (10) carries out similarity measurement, expression to the visual information feature for the conspicuousness enhancing extracted in step (9)
Formula is as follows:
Wherein, GMSR(x, y) and TISR(x, y) indicates the edge and texture information of the conspicuousness enhancing of middle reference image
Feature, GMSD(x, y) and TISD(x, y) indicates the edge and texture information of the conspicuousness enhancing of intermediate distorted image, WithRespectively the edge of middle reference/distorted image conspicuousness enhancing and texture information feature is equal
Value, M and N indicate the length and wide pixel number of image, Index1And Index2Respectively represent edge and texture information feature
Similarity measurements figureofmerit；
It is special that step (11) extracts color information by the color stereo image pair that the channel I and Q obtains from step (2)
Sign carries out similarity measurement to the left and right view of color stereo-picture respectively, obtains the color similarity figure under respective channel, table
It is as follows up to formula:
Wherein, ILR(x, y) and QLR(x, y) indicates the color under the channel I and the channel Q with reference to stereo pairs left view
Hum pattern, ILD(x, y) and QLD(x, y) indicates the color information figure under the channel I and the channel Q of distortion stereo pairs left view,
SIL(x, y) and SQL(x, y) respectively indicates the color similarity figure under the channel I and the channel Q of stereo pairs left view.Right view
The method that the color similarity figure of figure obtains is consistent with left view color similarity figure.T1And T2For constant, preventing denominator is zero.
The color similarity figure for the left and right view that step (12) obtains step (11) according to step (6) and (7) fusion
Method carries out binocular fusion, obtains the color similarity figure SI (x, y) and SQ (x, y) in the intermediate channel I and the channel Q.
The channel I of middle reference and distorted image that step (13) obtains step (12) and the color similarity in the channel Q
Figure is multiplied with what the stereoscopic vision Saliency maps that step (5) are established were put pixel-by-pixel, obtains the color of vision significance enhancing
Information Saliency maps, and then obtain the Color Similarity Measurement index Index in the channel I and the channel Q3And Index4, expression
It is as follows:
Wherein, M and N indicates the length and wide pixel number of image.
The disparity map of reference stereo pairs and distortion stereo pairs that step (14) is obtained using step (4) extracts
Depth characteristic information, and measurement is made to the distortion level of the disparity map of distortion stereo pairs；Using the side of pixel domain error
Method extracts the similitude of the depth characteristic information of reference and distortion stereo pairs, as reaction distortion stereo pairs in parallax
Figure is improved quality the index of distortion level, and expression formula is as follows:
Wherein, Dref(x, y) represents the disparity map of reference picture, Ddis(x, y) represents the disparity map of distorted image, mean ()
It is mean function, Index5Indicate the similarity measurements figureofmerit of depth characteristic information；
Step (15) integration step (10), the Measure Indexes Index of 5 visual correlations obtained in (13) and (14)1-
Index5, it is supported vector machine SVR training prediction, obtains optimum prediction model, and be mapped as objectively evaluating for picture quality
Score.
Wherein, the step (4) disparity map extracting method includes the following steps:
Step (4.1) will be referred to respectively and the right view all pixels point level of distortion stereo image pair moves to right n times,
The step-length moved every time is s pixel, the k width amendment right view I after obtaining horizontal move to rightR((x+i*s), y), (i=1,2 ...
K), then k=n/s, each width amendment right view is corresponding marked as i, and (i=1,2 ... k)；
Step (4.2) calculates separately the left view of stereo image pair using structural similarity algorithm SSIM and k width is corrected
The structural similarity of right view obtains k width structural similarity figure, (Z.Wang, A.C.Bovik, H.R.Sheikh, and
E.P.Simoncelli,“Image quality assessment:from error visibility to structural
similarity,” IEEE Transactions on Image Processing,vol.13,no.4,pp.600-612,
2004), SSIM algorithm expression formula is as follows:
SSIM (x, y)=[l (x, y)]α[c(x,y)]β[s(x,y)]γ (4-1)
Wherein, μxAnd μyRespectively indicate a corresponding image in the left view and amendment right view image of stereo pairs
Mean value in block；,σxAnd σyRespectively indicate a corresponding image block in the left view and amendment right view image of stereo pairs
Interior variance yields；σxyIn covariance between the left view of stereo pairs and an image block of amendment right view image
Covariance, C1=(k1L)2、C2=(k2L)2、C3=C2/2、k1、k2It is the positive number much smaller than 1, L is the maximum gray scale of image
Grade, preventing denominator is zero, and α, β, γ indicate the coefficient of weight between three functions, is all larger than zero.l(x,y),c(x,y),s(x,y)
Respectively luminance function, contrast function and structural similarity function of the stereo-picture in an image block；
Step (4.3) takes in its k width structural similarity figure and locally ties for each pixel p (x, y) of left view
The maximum width of structure similarity is corresponded to marked as i, (i=1,2 ... k), then i is the corresponding parallax value of p (x, y) pixel,
It is recorded as d (x, y), parallax value can be constructed for each pixel, to form disparity map D.
As a preferred solution of the present invention, in step (4) vision significance figure extracting method specifically:
Vision significance figure extracting method using spectrum residual error vision significance model (SR) (X.Hou and L.Zhang,
“Saliency detection:A spectral residual approach,”in Proc.20th IEEE
Conf.Comput. Vis.Pattern Recognit., Minneapolis, MN, USA, pp.1-8, Jun.2007), it is specific interior
Hold as follows:
Given piece image I (x, y), has:
Wherein, F () and F-1() is two-dimensional Fourier transform and its inverse transformation, and Re () expression takes real part operation, Angle ()
Expression takes argument operation, and A (f) is magnitude function, and P (f) is phase value function, and S (x, y) is significant to be obtained by spectrum residual error method
Property figure.In addition to this, g (x, y) is gauss low frequency filter, hn(f) it is local mean value filter, expression formula is as follows:
Wherein, σ is the standard deviation in probability distribution；
The left and right view of reference picture pair is regarded by the left and right that the method for spectrum residual error can respectively obtain reference picture pair
Feel Saliency maps.
Wherein, the step (15) is trained regression forecasting using the method for support vector regression (SVR), obtains most
Good prediction model specifically:
The method that SVR training prediction technique specifically uses 5- folding cross validation is trained and tests to model, specific side
Case is as follows:
Sample is divided into mutually disjoint five parts by step (15.1) at random, selects wherein four parts of progress SVR training to obtain
Best model is obtained, then remaining portion is applied on the model and is tested, corresponding objective quality value is obtained and comes to master
Appearance quality is predicted that then five parts of samples can carry out five training predictions；
Step (15.2) repeats the operation of step (15.1) 1000 times, and the intermediate value of all experimental results is taken to carry out table
Levy the performance of proposed model；
Expression is as follows:
Q=SVR (Index1,Index2,…,Indexn) (15-1)
Wherein, Q is evaluating objective quality score.
Beneficial effects of the present invention:
The present invention assists to extract edge, texture and color information feature by 3D vision significance figure, to realize to solid
The mapping of picture quality is realized and is objectively evaluated to distortion stereo pairs quality.The experimental results showed that being mentioned based on the present invention
Method has good consistency to the evaluation performance of stereo image quality and subjective assessment out, better than many typical perspective views
Image quality evaluation method.
Detailed description of the invention
Fig. 1 is that the present invention is based on visual saliency maps as that feature enhances is complete with reference to objective evaluation method for quality of stereo images
Schematic diagram.
Specific embodiment
The method of the present invention is described further with reference to the accompanying drawing.
Step (1) successively reads in the 3D LIVE image data of texas,U.S university Austin using Matlab software
The reference stereo pairs of stage I and stage II and corresponding distortion stereo pairs in library, wherein each stereo pairs point
It Bao Kuo not left and right view image.
Step (2) converts the color space of the stereo pairs in step (1), is converted into from rgb color space
YIQ color space, wherein the channel Y shows that the gray component of image, I and Q indicate the colour component of image.Specific conversion formula
It is as follows:
Step (3) constructs Log Gabor filter model, to the gray scale perspective view obtained in step (2) by the channel Y
As to convolution algorithm processing is carried out, respectively obtaining reference and being distorted the energy response figure of stereo image pair or so view；
Log Gabor filter hLGExpression formula it is as follows:
Wherein, f0And θ0Indicate centre frequency and the azimuth of Log Gabor filter, σθAnd σfRespectively represent filter
Azimuth bandwidth and radial bandwidth, f and θ respectively represent radial coordinate and the azimuth of filter.Wherein, σθ=π/18, σf=
0.75, f0=1/6, θ0=0, f=0, π/4, π/3,3 π/4, θ=0, π/5,2 π/5,3 π/5,4 π/5.Thus 4 × 5=20 is obtained
The local energy of a LoG Gabor filter energy response figure, the response of Log Gabor filter is defined as the energy between each scale
The maximum value of amount, and the local energy in each scale is defined as each azimuth and corresponds to the sum of local energy；
By Log Gabor filter and after referring to and being distorted stereo image pair or so view progress convolution, obtain corresponding
Energy response figure F (x, y), expression formula is as follows:
Wherein, I (x, y) is the left view or right view of reference and distortion stereo pairs gray component,For convolution fortune
It calculates；
Step (4) extracts parallax with distortion stereo pairs to the gray reference stereo pairs that step (2) obtains respectively
Scheme Dref(x, y) and Ddis(x, y) utilizes the left and right view for the gray reference stereo pairs that step (2) obtains residual based on composing
The conspicuousness model of difference extracts left and right Saliency maps SL respectivelysr(x, y) and SRsr(x,y)；
Step (5) constructs 3D vision significance figure S3D(x, y), specific expression formula are as follows:
S3D(x, y)=ω1×SLsr(x,y)+ω2×SRsr(x,y)+ω3×CB(x,y)+ω4×Dref(x,y) (5-2)
Wherein ω1-ω4For different weight factors,
CB (x, y) indicates central point offsetting mechanism, wherein ω1=ω2=0.45, ω3=ω4=0.05.
The right view for the gray scale stereo pairs that step (6) obtains step (2) is according to the parallax obtained in step (4)
The level that the parallax value of figure carries out pixel moves to right, the calibration right view I that construction and left view pixel coordinate pair are answeredR((x+d),
Y), it is then based on the available left view of Log Gabor filter model described in step (3) and calibrates the energy sound of right view
Ying Tu calculates normalized left view weight map WL(x, y) and calibration right view weight map WR((x+d), y), expression
It is as follows:
Wherein, FL(x, y) and FR((x+d), y) is respectively the energy sound of the left view that step (3) obtains and calibration right view
Ying Tu, d are the disparity map D that step (4) are calculatedrefThe parallax value of respective coordinates in (x, y)；
The left view and step (6) of gray reference and distortion stereo pairs that step (7) is based in step (2) obtain
Reference and be distorted stereo pairs calibration right view and normalized left view weight map and calibration right view weight map,
It is realized using binocular view Fusion Model to the image co-registration of stereo-picture, respectively obtains reference and distortion intermediate image；
The formula of binocular view fusion is as follows:
CI (x, y)=WL(x,y)×IL(x,y)+WR((x+d),y)×IR((x+d),y) (7-1)
Wherein, CI (x, y) is the fused intermediate image of binocular view, IL(x, y) and IR((x+d), y) is respectively ash
Spend the left view and calibration right view of stereo pairs；
Step (8) is referred to the middle gray that step (7) obtains and distorted image extracts edge and textural characteristics respectively；
The extraction of marginal information feature carries out process of convolution by Sobel operator and by altimetric image, obtains comprising edge contour
The gradient map of information, the expression formula using the marginal information feature of Sobel operator extraction middle reference and distorted image are as follows:
Wherein, f (x, y) is the left/right view of stereo pairs,For convolution algorithm, GxAnd GyIt is 3 × 3 Sobel water
Flat die plate and vertical formwork are respectively intended to the horizontal edge and vertical edge of detection image, and template expression formula is as follows:
The extraction of texture information feature.Using local binary patterns LBP, the expression formula of LBP is as follows:
Wherein, gcIt is the gray value of the central pixel point of image, gpIt is the gray value of the neighbor pixel of image, from center
The front-right of pixel rotates counterclockwise is followed successively by 0,1,2 ..., and P, x and y represent the coordinate value of central pixel point, and P indicates adjacent
The number of pixel, P=8, sgn (x) are jump functions；
The visual information feature and step (5) for the middle reference and distorted image that step (9) extracts step (8) are established
The multiplication put pixel-by-pixel of vision significance figure, obtain the visual information feature of vision significance enhancing, expression
It is as follows:
GMSR(x, y)=GMR(x,y)*S3D(x,y)GMSD(x, y)=GMD(x,y)*S3D(x,y) (9-1)
TISR(x, y)=TIR(x,y)*S3D(x,y)TISD(x, y)=TID(x,y)*S3D(x,y) (9-2)
Wherein, GMR(x, y) and TIR(x, y) is the edge and texture feature information of middle reference image, GM respectivelyD(x,y)
And TID(x, y) is the edge and texture feature information of intermediate distorted image respectively；S3DView after the integration obtained for step (5)
Feel Saliency maps；
Step (10) carries out similarity measurement, expression to the visual information feature for the conspicuousness enhancing extracted in step (9)
Formula is as follows:
Wherein, GMSR(x, y) and TISR(x, y) indicates the edge and texture information of the conspicuousness enhancing of middle reference image
Feature, GMSD(x, y) and TISD(x, y) indicates the edge and texture information of the conspicuousness enhancing of intermediate distorted image, WithRespectively the edge of middle reference/distorted image conspicuousness enhancing and texture information feature is equal
Value, M and N indicate the length and wide pixel number of image, Index1And Index2Respectively represent edge and texture information feature
Similarity measurements figureofmerit；
It is special that step (11) extracts color information by the color stereo image pair that the channel I and Q obtains from step (2)
Sign carries out similarity measurement to the left and right view of color stereo-picture respectively, obtains the color similarity figure under respective channel, table
It is as follows up to formula:
Wherein, ILR(x, y) and QLR(x, y) indicates the color under the channel I and the channel Q with reference to stereo pairs left view
Hum pattern, ILD(x, y) and QLD(x, y) indicates the color information figure under the channel I and the channel Q of distortion stereo pairs left view,
SIL(x, y) and SQL(x, y) respectively indicates the color similarity figure under the channel I and the channel Q of stereo pairs left view.Right view
The method that the color similarity figure of figure obtains is consistent with left view color similarity figure, T1And T2For constant, preventing denominator is zero.
Here, T1And T2It is 0.5.
The color similarity figure for the left and right view that step (12) obtains step (11) according to step (6) and (7) fusion
Method carries out binocular fusion, obtains the color similarity figure SI (x, y) and SQ (x, y) in the intermediate channel I and the channel Q.
The channel I of middle reference and distorted image that step (13) obtains step (12) and the color similarity in the channel Q
Figure is multiplied with what the stereoscopic vision Saliency maps that step (5) are established were put pixel-by-pixel, obtains the color of vision significance enhancing
Information Saliency maps, and then obtain the Color Similarity Measurement index Index in the channel I and the channel Q3And Index4, expression
It is as follows:
Wherein, M and N indicates the length and wide pixel number of image.
The disparity map of reference stereo pairs and distortion stereo pairs that step (14) is obtained using step (4) extracts
Depth characteristic information, and measurement is made to the distortion level of the disparity map of distortion stereo pairs；Using the side of pixel domain error
Method extracts the similitude of the depth characteristic information of reference and distortion stereo pairs, as reaction distortion stereo pairs in parallax
Figure is improved quality the index of distortion level, and expression formula is as follows:
Wherein, Dref(x, y) represents the disparity map of reference picture, Ddis(x, y) represents the disparity map of distorted image, mean ()
It is mean function, Index5Indicate the similarity measurements figureofmerit of depth characteristic information；
Step (15) integration step (10), the Measure Indexes Index of 5 visual correlations obtained in (13) and (14)1-
Index5, it is trained prediction with support vector regression (SVR), obtains optimum prediction model, and be mapped as picture quality
Objective assessment score.
Wherein, the complete of view-based access control model specific image feature enhancing according to claim 1 refers to stereo image quality
Method for objectively evaluating, it is characterised in that described step (4) the disparity map extracting method includes the following steps:
Step (4.1) will be referred to respectively and the right view all pixels point level of distortion stereo image pair moves to right n times,
The step-length moved every time is s pixel, the k width amendment right view I after obtaining horizontal move to rightR((x+i*s), y), (i=1,2 ...
K), then k=n/s, in this s=1, n=25, then k=25.Each width amendment right view is corresponding marked as i, (i=1,2 ...
k)；
Step (4.2) calculates separately the left view of stereo image pair using structural similarity algorithm SSIM and k width is corrected
The structural similarity of right view obtains k width structural similarity figure (Z.Wang, A.C.Bovik, H.R.Sheikh, and
E.P.Simoncelli,“Image quality assessment:from error visibility to structural
similarity,” IEEE Transactions on Image Processing,vol.13,no.4,pp.600-612,
2004), SSIM algorithm expression formula is as follows:
SSIM (x, y)=[l (x, y)]α[c(x,y)]β[s(x,y)]γ (4-1)
Wherein, μxAnd μyRespectively indicate a corresponding image in the left view and amendment right view image of stereo pairs
Mean value in block；,σxAnd σyRespectively indicate a corresponding image block in the left view and amendment right view image of stereo pairs
Interior variance yields；σxyIn covariance between the left view of stereo pairs and an image block of amendment right view image
Covariance, C1=(k1L)2、C2=(k2L)2、C3=C2/2、k1、k2It is the positive number much smaller than 1, L is the maximum gray scale of image
Grade, preventing denominator is zero, and α, β, γ indicate the coefficient of weight between three functions, zero is all larger than, in this α=β=γ=1, C1=
6.5025 C2=58.5225, C2=29.2612.L (x, y), c (x, y), s (x, y) are respectively stereo-picture in an image block
In luminance function, contrast function and structural similarity function；
Step (4.3) takes in its k width structural similarity figure and locally ties for each pixel p (x, y) of left view
The maximum width of structure similarity is corresponded to marked as i, (i=1,2 ... k), then i is the corresponding parallax value of p (x, y) pixel,
It is recorded as d (x, y), parallax value can be constructed for each pixel, to form disparity map D.
The complete of view-based access control model specific image feature according to claim 1 enhancing is commented with reference to stereo image quality is objective
Valence method, it is characterised in that the extracting method of vision significance figure in the step (4) specifically:
Vision significance figure extracting method using spectrum residual error vision significance model (SR) (X.Hou and L.Zhang,
“Saliency detection:A spectral residual approach,”in Proc.20th IEEE
Conf.Comput. Vis.Pattern Recognit., Minneapolis, MN, USA, pp.1-8, Jun.2007), it is specific interior
Hold as follows:
Given piece image I (x, y), has:
Wherein, F () and F-1() is two-dimensional Fourier transform and its inverse transformation, and Re () expression takes real part operation, Angle ()
Expression takes argument operation, and A (f) is magnitude function, and P (f) is phase value function, and S (x, y) is significant to be obtained by spectrum residual error method
Property figure.In addition to this, g (x, y) is gauss low frequency filter, hn(f) it is local mean value filter, expression formula is as follows:
Wherein, σ is the standard deviation in probability distribution, σ=1.5；
The left and right view of reference picture pair is regarded by the left and right that the method for spectrum residual error can respectively obtain reference picture pair
Feel Saliency maps.
Wherein, the complete of view-based access control model specific image feature enhancing according to claim 1 refers to stereo image quality
Method for objectively evaluating, it is characterised in that the step (15) is trained recurrence using the method for support vector regression (SVR)
Prediction obtains optimum prediction model specifically:
The method that SVR training prediction technique specifically uses 5- folding cross validation is trained and tests to model, specific side
Case is as follows:
Sample is divided into mutually disjoint five parts by step (15.1) at random, selects wherein four parts of progress SVR training to obtain
Best model is obtained, then remaining portion is applied on the model and is tested, corresponding objective quality value is obtained and comes to master
Appearance quality is predicted that then five parts of samples can carry out five training predictions；
Step (15.2) repeats the operation of step (15.1) 1000 times, and the intermediate value of all experimental results is taken to carry out table
Levy the performance of proposed model；
Expression is as follows:
Q=SVR (Index1,Index2,…,Indexn) (15-1)
Wherein, Q is evaluating objective quality score.
The complete of view-based access control model specific image feature according to claim 1 enhancing is commented with reference to stereo image quality is objective
Valence method has carried out following experiment to verify the superior function of algorithm of the present invention, is labeled as step (16).
Database is the LIVE stereo image quality rating database stage using two large database concepts being widely adopted at present
I and stage II (http://live.ece.utexas.edu/research/Quality/live_3dimage.html).Stage
I includes 20 width reference pictures pair, and the symmetrical different distorted image of 365 width is then derived from this 20 width reference picture pair
Right, stage II includes that 8 width reference pictures pair are symmetrically lost by this 8 width image to 120 width are derived unlike stage I
True image to and 240 asymmetrical image faults pair, this is to test proposed stereo image quality evaluation algorithms pair
In the evaluation and test effect of asymmetric distorted image.Either stage I or stage II, the type for the distorted image that they are included are
Five kinds below: JPEG compression (JPEG), JPEG2000 compression (JP2K), Gaussian Blur (GB), white Gaussian noise (WN) and fast
It declines (FF).Performance evaluation can be carried out by the image to every kind of type of distortion respectively in experimentation, it finally can be to all distortion maps
As carrying out total performance evaluation.One good three-dimensional Environmental Evaluation Model will not only require the overall performance on all distorted images
It is good, and also to be got well in the performance of every kind of type of distortion image.
Experimental index selection Pearson correlation coefficients (Pearson ' s linear correlation coefficient,
PLCC), Spearman's correlation coefficient (Spearman ' s rank ordered correlation coefficient,
SROCC) and root-mean-square error (root-mean-squared error, RMSE), the objective quality proposed for inspection institute
The performance of evaluation method.Wherein PLCC and RMSE is the accuracy for predicting Objective image quality evaluation method, and SRCC is then
It is the monotonicity for predicting method for objectively evaluating.The PLCC and SRCC of one algorithm are higher, and RMSE is lower, represent the objective matter
The algorithm of amount evaluation has better accuracy and robustness.The calculation formula of PLCC, SRCC and RMSE are as follows:
Wherein, n is total number of images amount, xiAnd yiRespectively subjective quality scores and prediction evaluating objective quality score, Xi
And YiRespectively xiAnd yiRanking in subjective quality scores and objective quality scores.
Last experimental result is listed in table, and table 1 indicates the method for the invention in LIVE 3D database stage I and rank
Overall performance on section II, table 2 list the performance of PLCC, SROCC and RMSE for different type of distortion.Experimental result
Show either on LIVE 3D database stage I or stage II, algorithm of the invention all achieves prediction effect well
Fruit has preferable subjective consistency.
Overall performance of 1 the method for the invention of table on LIVE 3D database
PLCC, SROCC and RMSE of 2 the method for the invention of table each type of distortion on LIVE 3D database
Claims (4)
1. a kind of the complete of view-based access control model specific image feature enhancing refers to objective evaluation method for quality of stereo images, it is characterised in that
The following steps are included:
Step (1) input is with reference to stereo pairs and distortion stereo pairs, wherein each stereo pairs respectively include left view
Figure and right view image；
Step (2) converts the color space of the stereo pairs in step (1), is converted into YIQ from rgb color space
Color space, wherein the channel Y shows that the gray component of image, I and Q indicate the colour component of image；Specific conversion formula is such as
Shown in lower:
Step (3) constructs Log Gabor filter model, to the gray scale stereo pairs obtained in step (2) by the channel Y
Convolution algorithm processing is carried out, reference is respectively obtained and is distorted the energy response figure of stereo image pair or so view；
Log Gabor filter hLGExpression formula it is as follows:
Wherein, f0And θ0Indicate centre frequency and the azimuth of Log Gabor filter, σθAnd σfRespectively represent the orientation of filter
Angle bandwidth and radial bandwidth, f and θ respectively represent radial coordinate and the azimuth of filter；
By Log Gabor filter and after referring to and being distorted stereo image pair or so view progress convolution, corresponding energy is obtained
It measures response diagram F (x, y), expression formula is as follows:
Wherein, I (x, y) is the left view or right view of reference and distortion stereo pairs gray component,For convolution algorithm；
Step (4) extracts disparity map with distortion stereo pairs to the gray reference stereo pairs that step (2) obtains respectively
Dref(x, y) and Ddis(x, y) utilizes based on spectrum residual error the left and right view for the gray reference stereo pairs that step (2) obtains
Conspicuousness model extract left and right Saliency maps SL respectivelysr(x, y) and SRsr(x,y)；
Step (5) constructs 3D vision significance figure S3D(x, y), specific expression formula are as follows:
S3D(x, y)=ω1×SLsr(x,y)+ω2×SRsr(x,y)+ω3×CB(x,y)+ω4×Dref(x,y) (5-2)
Wherein ω1-ω4For different weight factors, CB (x, y) indicates central point offsetting mechanism；
The right view for the gray scale stereo pairs that step (6) obtains step (2) is according to the disparity map obtained in step (4)
The level that parallax value carries out pixel moves to right, the calibration right view I that construction and left view pixel coordinate pair are answeredR((x+d), y), so
Energy response figure based on the available left view of Log Gabor filter model described in step (3) and calibration right view afterwards,
Calculate normalized left view weight map WL(x, y) and calibration right view weight map WR((x+d), y), expression is as follows:
Wherein, FL(x, y) and FR((x+d), y) is respectively the energy response of the left view that step (3) obtains and calibration right view
Figure, d are the disparity map D that step (4) are calculatedrefThe parallax value of respective coordinates in (x, y)；
The ginseng that the left view and step (6) of gray reference and distortion stereo pairs that step (7) is based in step (2) obtain
Examine and be distorted stereo pairs calibration right view and normalized left view weight map and calibration right view weight map, utilize
Binocular view Fusion Model is realized to the image co-registration of stereo-picture, and reference and distortion intermediate image is respectively obtained；Binocular
The formula of view fusion is as follows:
CI (x, y)=WL(x,y)×IL(x,y)+WR((x+d),y)×IR((x+d),y) (7-1)
Wherein, CI (x, y) is the fused intermediate image of binocular view, IL(x, y) and IR((x+d), y) is respectively that gray scale is vertical
The left view and calibration right view of body image pair；
Step (8) is referred to the middle gray that step (7) obtains and distorted image extracts edge and textural characteristics respectively；
The extraction of marginal information feature: process of convolution is carried out by Sobel operator and by altimetric image, is obtained comprising edge contour information
Gradient map, the expression formula using the marginal information feature of Sobel operator extraction middle reference and distorted image is as follows:
Wherein, f (x, y) is the left/right view of stereo pairs,For convolution algorithm, GxAnd GyIt is 3 × 3 horizontal mould of Sobel
Plate and vertical formwork are respectively intended to the horizontal edge and vertical edge of detection image, and template expression formula is as follows:
The extraction of texture information feature: using local binary patterns LBP, and the expression formula of LBP is as follows:
Wherein, gcIt is the gray value of the central pixel point of image, gpIt is the gray value of the neighbor pixel of image, from center pixel
The front-right of point rotates counterclockwise is followed successively by 0,1,2 ..., and P, x and y represent the coordinate value of central pixel point, and P indicates neighbor pixel
Number, sgn (x) is jump function；
The view that the visual information feature and step (5) for the middle reference and distorted image that step (9) extracts step (8) are established
Feel the multiplication that Saliency maps are put pixel-by-pixel, obtain the visual information feature of vision significance enhancing, expression is as follows:
GMSR(x, y)=GMR(x,y)*S3D(x,y) GMSD(x, y)=GMD(x,y)*S3D(x,y) (9-1)
TISR(x, y)=TIR(x,y)*S3D(x,y) TISD(x, y)=TID(x,y)*S3D(x,y) (9-2)
Wherein, GMR(x, y) and TIR(x, y) is the edge and texture feature information of middle reference image, GM respectivelyD(x, y) and TID
(x, y) is the edge and texture feature information of intermediate distorted image respectively；S3DVision after the integration obtained for step (5) is significant
Property figure；
Step (10) carries out similarity measurement to the visual information feature for the conspicuousness enhancing extracted in step (9), and expression formula is such as
Under:
Wherein, GMSR(x, y) and TISR(x, y) indicates the edge and texture information feature of the conspicuousness enhancing of middle reference image,
GMSD(x, y) and TISD(x, y) indicates the edge and texture information of the conspicuousness enhancing of intermediate distorted image, WithRespectively the edge of middle reference/distorted image conspicuousness enhancing and texture information feature is equal
Value, M and N indicate the length and wide pixel number of image, Index1And Index2Respectively represent edge and texture information feature
Similarity measurements figureofmerit；
Step (11) extracts color information feature by the color stereo image pair that the channel I and Q obtains from step (2), point
The other left and right view to color stereo-picture carries out similarity measurement, obtains the color similarity figure under respective channel, expression formula
It is as follows:
Wherein, ILR(x, y) and QLR(x, y) indicates the color information under the channel I and the channel Q with reference to stereo pairs left view
Figure, ILD(x, y) and QLD(x, y) indicates the color information figure under the channel I and the channel Q of distortion stereo pairs left view, SIL
(x, y) and SQL(x, y) respectively indicates the color similarity figure under the channel I and the channel Q of stereo pairs left view, right view
The obtained method of color similarity figure it is consistent with left view color similarity figure, T1And T2For constant, preventing denominator is zero；
The color similarity figure for the left and right view that step (12) obtains step (11) according to step (6) and (7) fusion method
Binocular fusion is carried out, the color similarity figure SI (x, y) and SQ (x, y) in the intermediate channel I and the channel Q are obtained；
The color similarity figure in the channel I of middle reference and distorted image that step (13) obtains step (12) and the channel Q and
The multiplication that the stereoscopic vision Saliency maps that step (5) is established are put pixel-by-pixel obtains the color information of vision significance enhancing
Saliency maps, and then obtain the Color Similarity Measurement index Index in the channel I and the channel Q3And Index4, expression is such as
Under:
Wherein, M and N indicates the length and wide pixel number of image；
The disparity map of reference stereo pairs and distortion stereo pairs that step (14) is obtained using step (4) extracts depth
Characteristic information, and measurement is made to the distortion level of the disparity map of distortion stereo pairs；It is mentioned using the method for pixel domain error
It takes reference and is distorted the similitude of the depth characteristic information of stereo pairs, as reaction distortion stereo pairs on disparity map
The index of quality distortion degree, expression formula are as follows:
Wherein, Dref(x, y) represents the disparity map of reference picture, Ddis(x, y) represents the disparity map of distorted image, and mean () is equal
Value function, Index5Indicate the similarity measurements figureofmerit of depth characteristic information；
Step (15) integration step (10), the Measure Indexes Index of 5 visual correlations obtained in (13) and (14)1-
Index5, it is supported vector regression (SVR) and is trained prediction, obtains optimum prediction model, and be mapped as picture quality
Objective assessment score.
2. the complete of view-based access control model specific image feature enhancing according to claim 1 is objectively evaluated with reference to stereo image quality
Method, it is characterised in that described step (4) the disparity map extracting method includes the following steps:
Step (4.1) will be referred to respectively and the right view all pixels point level of distortion stereo image pair moves to right n times, every time
Mobile step-length is s pixel, the k width amendment right view I after obtaining horizontal move to rightR((x+i*s), y), wherein i=1,2 ...
K, then k=n/s, each width amendment right view are corresponding marked as i, i=1,2 ... k；
Step (4.2) calculates separately the left view of stereo image pair using structural similarity algorithm SSIM and k width corrects right view
The structural similarity of figure obtains k width structural similarity figure, and SSIM algorithm expression formula is as follows:
SSIM (x, y)=[l (x, y)]α[c(x,y)]β[s(x,y)]γ (4-1)
Wherein, μxAnd μyIt respectively indicates in the left view and amendment right view image of stereo pairs in a corresponding image block
Mean value；,σxAnd σyIt respectively indicates in the left view and amendment right view image of stereo pairs in a corresponding image block
Variance yields；σxyThe association side in covariance between the left view of stereo pairs and an image block of amendment right view image
Difference, C1=(k1L)2、C2=(k2L)2、C3=C2/2、k1、k2It is the positive number much smaller than 1, L is the maximum gray scale of image, is prevented
Only denominator is zero, and α, β, γ indicate the coefficient of weight between three functions, is all larger than zero；L (x, y), c (x, y), s (x, y) are respectively
Luminance function, contrast function and structural similarity function of the stereo-picture in an image block；
Step (4.3) takes partial structurtes phase in its k width structural similarity figure for each pixel p (x, y) of left view
It is worth a maximum width like property, corresponding marked as i, i=1,2 ... k, then i is the corresponding parallax value of p (x, y) pixel, is recorded as d
(x, y) can construct parallax value for each pixel, to form disparity map D.
3. the complete of view-based access control model specific image feature enhancing according to claim 1 is objectively evaluated with reference to stereo image quality
Method, it is characterised in that the extracting method of described step (4) the vision significance figure specifically:
For vision significance figure extracting method using the vision significance model (SR) of spectrum residual error, particular content is as follows:
Given piece image I (x, y), has:
Wherein, F () and F-1() is two-dimensional Fourier transform and its inverse transformation, and Re () expression takes real part operation, and Angle () is indicated
Argument operation is taken, A (f) is magnitude function, and P (f) is phase value function, and S (x, y) is the conspicuousness obtained by composing residual error method
Figure；In addition to this, g (x, y) is gauss low frequency filter, hn(f) it is local mean value filter, expression formula is as follows:
Wherein, σ is the standard deviation in probability distribution；
The left and right vision that reference picture pair can be respectively obtained by the method for spectrum residual error to the left and right view of reference picture pair is aobvious
Work property figure.
4. the complete of view-based access control model specific image feature enhancing according to claim 1 is objectively evaluated with reference to stereo image quality
Method, it is characterised in that the step (15) is trained regression forecasting using the method for support vector regression (SVR), obtains
Optimum prediction model specifically:
The method that SVR training prediction technique specifically uses 5- folding cross validation is trained and tests to model, and concrete scheme is such as
Under:
Sample is divided into mutually disjoint five parts by step (15.1) at random, selects wherein four parts of progress SVR training to obtain most
Then remaining portion is applied on the model and is tested by good model, obtain corresponding objective quality value and come to subjective matter
Amount is predicted that then five parts of samples can carry out five training predictions；
Step (15.2) repeats the operation of step (15.1) 1000 times, takes the intermediate value of all experimental results to characterize
It is proposed the performance of model；
Expression is as follows:
Q=SVR (Index1,Index2,L,Indexn) (15-1)
Wherein, Q is evaluating objective quality score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811109216.6A CN109523506B (en) | 2018-09-21 | 2018-09-21 | Full-reference stereo image quality objective evaluation method based on visual salient image feature enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811109216.6A CN109523506B (en) | 2018-09-21 | 2018-09-21 | Full-reference stereo image quality objective evaluation method based on visual salient image feature enhancement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109523506A true CN109523506A (en) | 2019-03-26 |
| CN109523506B CN109523506B (en) | 2021-03-26 |
Family
ID=65771646
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811109216.6A Active CN109523506B (en) | 2018-09-21 | 2018-09-21 | Full-reference stereo image quality objective evaluation method based on visual salient image feature enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109523506B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110399887A (en) * | 2019-07-19 | 2019-11-01 | 合肥工业大学 | The representative Color Picking method of view-based access control model conspicuousness and statistics with histogram technology |
| CN110399881A (en) * | 2019-07-11 | 2019-11-01 | 深圳大学 | A kind of quality enhancement method and device based on binocular stereo image end to end |
| CN110930398A (en) * | 2019-12-09 | 2020-03-27 | 嘉兴学院 | Log-Gabor similarity-based full-reference video quality evaluation method |
| CN111179238A (en) * | 2019-12-24 | 2020-05-19 | 东华大学 | Subset confidence ratio dynamic selection method for subset-oriented guidance consistency enhancement evaluation |
| CN112132774A (en) * | 2019-07-29 | 2020-12-25 | 方玉明 | Quality evaluation method of tone mapping image |
| CN112581461A (en) * | 2020-12-24 | 2021-03-30 | 深圳大学 | No-reference image quality evaluation method and device based on generation network |
| CN112734733A (en) * | 2021-01-12 | 2021-04-30 | 天津大学 | Non-reference image quality monitoring method based on channel recombination and feature fusion |
| CN114067006A (en) * | 2022-01-17 | 2022-02-18 | 湖南工商大学 | Screen content image quality evaluation method based on discrete cosine transform |
| US11810530B2 (en) | 2019-04-02 | 2023-11-07 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for display-brightness adjustment and related devices |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130155192A1 (en) * | 2011-12-15 | 2013-06-20 | Industrial Technology Research Institute | Stereoscopic image shooting and display quality evaluation system and method applicable thereto |
| WO2013105720A1 (en) * | 2012-01-10 | 2013-07-18 | 에스케이플래닛 주식회사 | Device and method for analyzing quality of three-dimensional stereoscopic image |
| CN107330873A (en) * | 2017-05-05 | 2017-11-07 | 浙江大学 | Objective evaluation method for quality of stereo images based on multiple dimensioned binocular fusion and local shape factor |
| CN107578404A (en) * | 2017-08-22 | 2018-01-12 | 浙江大学 | The complete of view-based access control model notable feature extraction refers to objective evaluation method for quality of stereo images |
| CN107610093A (en) * | 2017-08-02 | 2018-01-19 | 西安理工大学 | Full-reference image quality evaluating method based on similarity feature fusion |
-
2018
- 2018-09-21 CN CN201811109216.6A patent/CN109523506B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130155192A1 (en) * | 2011-12-15 | 2013-06-20 | Industrial Technology Research Institute | Stereoscopic image shooting and display quality evaluation system and method applicable thereto |
| WO2013105720A1 (en) * | 2012-01-10 | 2013-07-18 | 에스케이플래닛 주식회사 | Device and method for analyzing quality of three-dimensional stereoscopic image |
| CN107330873A (en) * | 2017-05-05 | 2017-11-07 | 浙江大学 | Objective evaluation method for quality of stereo images based on multiple dimensioned binocular fusion and local shape factor |
| CN107610093A (en) * | 2017-08-02 | 2018-01-19 | 西安理工大学 | Full-reference image quality evaluating method based on similarity feature fusion |
| CN107578404A (en) * | 2017-08-22 | 2018-01-12 | 浙江大学 | The complete of view-based access control model notable feature extraction refers to objective evaluation method for quality of stereo images |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11810530B2 (en) | 2019-04-02 | 2023-11-07 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for display-brightness adjustment and related devices |
| CN110399881A (en) * | 2019-07-11 | 2019-11-01 | 深圳大学 | A kind of quality enhancement method and device based on binocular stereo image end to end |
| CN110399887B (en) * | 2019-07-19 | 2022-11-04 | 合肥工业大学 | Representative color extraction method based on visual saliency and histogram statistical technology |
| CN110399887A (en) * | 2019-07-19 | 2019-11-01 | 合肥工业大学 | The representative Color Picking method of view-based access control model conspicuousness and statistics with histogram technology |
| CN112132774A (en) * | 2019-07-29 | 2020-12-25 | 方玉明 | Quality evaluation method of tone mapping image |
| CN110930398A (en) * | 2019-12-09 | 2020-03-27 | 嘉兴学院 | Log-Gabor similarity-based full-reference video quality evaluation method |
| CN110930398B (en) * | 2019-12-09 | 2023-05-09 | 嘉兴学院 | Total reference video quality evaluation method based on Log-Gabor similarity |
| CN111179238A (en) * | 2019-12-24 | 2020-05-19 | 东华大学 | Subset confidence ratio dynamic selection method for subset-oriented guidance consistency enhancement evaluation |
| CN111179238B (en) * | 2019-12-24 | 2022-12-20 | 东华大学 | Subset confidence ratio dynamic selection method for underwater image set-oriented guidance consistency enhancement evaluation |
| CN112581461A (en) * | 2020-12-24 | 2021-03-30 | 深圳大学 | No-reference image quality evaluation method and device based on generation network |
| CN112734733A (en) * | 2021-01-12 | 2021-04-30 | 天津大学 | Non-reference image quality monitoring method based on channel recombination and feature fusion |
| CN112734733B (en) * | 2021-01-12 | 2022-11-01 | 天津大学 | Non-reference image quality monitoring method based on channel recombination and feature fusion |
| CN114067006B (en) * | 2022-01-17 | 2022-04-08 | 湖南工商大学 | Screen content image quality evaluation method based on discrete cosine transform |
| CN114067006A (en) * | 2022-01-17 | 2022-02-18 | 湖南工商大学 | Screen content image quality evaluation method based on discrete cosine transform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109523506B (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109523506A (en) | The complete of view-based access control model specific image feature enhancing refers to objective evaluation method for quality of stereo images | |
| CN107578404B (en) | View-based access control model notable feature is extracted complete with reference to objective evaluation method for quality of stereo images | |
| Battisti et al. | Objective image quality assessment of 3D synthesized views | |
| CN109255358B (en) | 3D image quality evaluation method based on visual saliency and depth map | |
| Shen et al. | Hybrid no-reference natural image quality assessment of noisy, blurry, JPEG2000, and JPEG images | |
| CN100389437C (en) | Method and apparatus for modeling film grain patterns in the frequency domain | |
| CN111563418A (en) | Asymmetric multi-mode fusion significance detection method based on attention mechanism | |
| CN109242834A (en) | It is a kind of based on convolutional neural networks without reference stereo image quality evaluation method | |
| CN106303507B (en) | Video quality evaluation without reference method based on space-time united information | |
| CN109919959A (en) | Tone mapping image quality evaluating method based on color, naturality and structure | |
| CN104123705B (en) | A kind of super-resolution rebuilding picture quality Contourlet territory evaluation methodology | |
| CN102333233A (en) | Stereo image quality objective evaluation method based on visual perception | |
| CN109831664B (en) | Rapid compressed stereo video quality evaluation method based on deep learning | |
| CN106023230B (en) | A kind of dense matching method of suitable deformation pattern | |
| CN105550989B (en) | The image super-resolution method returned based on non local Gaussian process | |
| CN102547368A (en) | Objective evaluation method for quality of stereo images | |
| CN103354617B (en) | Boundary strength compressing image quality objective evaluation method based on DCT domain | |
| CN109345502A (en) | A kind of stereo image quality evaluation method based on disparity map stereochemical structure information extraction | |
| Hsia | Improved depth image-based rendering using an adaptive compensation method on an autostereoscopic 3-D display for a Kinect sensor | |
| Zhou et al. | Reduced-reference quality assessment of point clouds via content-oriented saliency projection | |
| CN108447059A (en) | It is a kind of to refer to light field image quality evaluating method entirely | |
| CN116934592A (en) | Image stitching method, system, equipment and medium based on deep learning | |
| CN104637060A (en) | Image partition method based on neighbor-hood PCA (Principal Component Analysis)-Laplace | |
| CN117671384A (en) | Hyperspectral image classification method | |
| CN105721863B (en) | Method for evaluating video quality |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |