CN107635136B - View-based access control model perception and binocular competition are without reference stereo image quality evaluation method - Google Patents
View-based access control model perception and binocular competition are without reference stereo image quality evaluation method Download PDFInfo
- Publication number
- CN107635136B CN107635136B CN201711003045.4A CN201711003045A CN107635136B CN 107635136 B CN107635136 B CN 107635136B CN 201711003045 A CN201711003045 A CN 201711003045A CN 107635136 B CN107635136 B CN 107635136B
- Authority
- CN
- China
- Prior art keywords
- image
- view
- perception
- quality evaluation
- access control
- 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.)
- Active
Links
Landscapes
- Image Processing (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of stereo image quality evaluation method, in particular to a kind of view-based access control model perception, without reference stereo image quality evaluation method, belongs to art of image analysis with binocular competition.This method converts grayscale information for input stereo pairs first, obtains the simulation disparity map and uncertain figure of stereo pairs using matching algorithm to grayscale information, while synthesizing single eye images using grayscale information and its filter response and simulation disparity map correction.Secondly, obtained single eye images and uncertain figure are carried out difference of Gaussian processing on different scale space and frequency space, and extract nature scene statistics and visual perception feature vector.Then, feature is trained respectively using support vector machines and BP neural network, obtains prediction model, applied forecasting model and test and corresponding feature vector, carry out prediction of quality and assessment.This method has subjective consistency high, data base-independent is high, the high feature of stability, the effect of great competitiveness is all shown when handling Various Complex type of distortion, it can be embedded into the relevant application systems of stereoscopic visions content such as stereoscopic image/video processing, there is very strong application value.
Description
Technical field
The present invention relates to a kind of stereo image quality evaluation method, in particular to a kind of view-based access control model perception and binocular competition
Without reference stereo image quality evaluation method, belong to art of image analysis.
Background technique
In recent years, with the development of science and technology, the cost that stereo-picture is generated and propagated becomes lower and lower, this makes
Stereo-picture is obtained as a kind of medium that outstanding information is propagated, is become increasingly prevalent in our daily life, it is more next
It is more indispensable.However, stereo-picture is in scene acquisition, coding, network transmission, decoding, post-processing, compression storage and projection
Each stage all can inevitably introduce distortion, for example, in scene collection process due to apparatus parameter setting, camera lens shake
Distortion is obscured caused by the factors such as dynamic;Compression artefacts caused by compression of images stores etc..And the introducing being distorted can then drop significantly
The visual experience of low people, the serious physical and mental health for also affecting people.How the propagation of low quality stereo-picture is contained,
The visual experience for guaranteeing people, becomes a urgent problem to be solved.
The media for making stereo-picture generate and propagate have the ability of automatic Evaluation picture quality height, so as to improve media
The quality of output end image is of great significance for solving this problem.Specifically, this research has applies valence below
Value:
(1) it can be embedded in actual application system (such as projection system, network transmission system of video etc.), in real time
Monitoring image/video quality;
(2) can be used for evaluating various stereoscopic image/video Processing Algorithms, tool (such as stereo-picture compressed encoding,
Image/video sampling instrument etc.) superiority and inferiority;
(3) quality audit that can be used for stereoscopic image/video works prevents Poor Image product from endangering the body and mind of spectators
Health.
In conclusion for the research of objective no reference stereo image quality evaluation model have important theoretical value and
Realistic meaning.The invention proposes a kind of perception of view-based access control model and binocular perception without reference stereo image quality evaluation method,
Its existing theory and technology referred to is the visual impression that the visual perception theory that Kruger et al. is proposed and Joshi et al. are proposed
Know feature extraction theory.
(1) visual perception is theoretical
Kruger et al. proposes visual perception theory, and the research in relation to visual perception theory first has to consider human eye view
The perceptual phenomena of film.Photosensory cell in retina generates light conduction, and the signal that light conduction generates is in excitability or inhibition
Visual channel in transmit.There are low-pass filtering in Human Retinal Ganglion Cell for studies have shown that, go out in this background
An existing prominent features are exactly the center-of retina around received field [40].Center-is generally circular concentric around received field
Shape is (or inhibit) excited to optical signal in the central area of received field, and then press down to reception optical signal in field
System (or excited).This received field can be modeled by difference of Gaussian, and be similar to and filtered for the Laplce of edge detection
Device [41].Therefore it emphasizes the spatial variations of brightness, in addition, this received field is also sensitive to time change, and is therefore formed
To the basis of motion process.In addition, there is also handle different types of visual information (color, shape in human visual system
Shape, movement, texture, steric information) separation and height interconnection channel, this facilitate visual information expression efficiency and stabilization
Property.Under this visual perception mechanism, brain perceives the three-dimensional feature of stereo-picture, binocular vision by a large amount of depth information
Difference is one of most important one depth information.In view of there may be multiple spatial frequencys in retina, therefore to simulate this
The center-of a little frequencies needs to generate multiple standard deviations, and calculate difference diagram by difference of Gaussian around received field
Picture.
(2) visual perception feature extraction
On the basis of Joshi et al. is studied the problems such as perceiving to visual perception and retina, extraction is proposed
Method of the energy feature and edge feature of image as visual perception feature.
The extraction calculation formula of energy feature is as follows:
Wherein, the comentropy of H representative image, the quantity of m representative image grey level, plFirst of grey level is represented to go out
Existing probability correlation value.
The extraction calculation formula of edge feature is as follows:
Wherein, Canny represents the edge detection using Canny method progress image, and by qualified edge pixel
Point is indicated with numerical approach.
Summary of the invention
The purpose of the present invention is to solve human eye visual perception system simulation sides in the evaluation of no reference stereo image quality
Method is incomplete, in image visual perception information using insufficient, subjective consistency is poor, and data base-independent is poor, algorithm
The problem of stability difference etc. proposes that a kind of view-based access control model is perceived with binocular competition without reference stereo image quality evaluation method.
The method of the present invention is achieved through the following technical solutions.
View-based access control model is perceived with binocular competition without reference stereo image quality evaluation method, the specific steps of which are as follows:
Step 1: converting grayscale information for the stereo pairs to be tested of input.
Step 2: being handled using matching algorithm is further to grayscale information, simulation disparity map and uncertainty are obtained
Figure, while filtering to obtain the filter response of grayscale information using Gabor.
Step 3: utilizing grayscale information and its filter response and simulation disparity map correction synthesis single eye images.
Step 4: obtaining Gaussian difference component from the different scale space and frequency space of single eye images and uncertain figure
Picture, and complete natural scene statistics and visual perception feature extraction.
The calculation method of difference of Gaussian image is as follows:
σ2 ij=L* σ1 ij (3)
Wherein,Difference of Gaussian image is represented,WithRespectively represent to original image (single eye images or not really
Qualitative figure) carry out the image that the gaussian filtering under different convolution kernels obtains, σ1 ijAnd σ2 ijTwo different convolution kernels are respectively represented,
W and h represents the width and height of image to be processed under a certain scale, and f represents frequency, and i and j respectively represent some scale space and frequency
Rate space.
The extracting method of visual perception feature is as follows:
The extraction of energy feature:
Wherein, the comentropy of H representative image, the quantity of m representative image grey level, plFirst of grey level is represented to go out
Existing probability correlation value.
The extraction of edge feature:
Wherein, Canny represents the edge detection using Canny method progress image, and by qualified edge pixel
Point is indicated with numerical approach.
Step 5: using Step 1: Step 2: the method for step 3 and step 4 is vertical to each width colour in database
Each group of quality characteristic vector corresponding to stereo-picture is calculated to handling in body image;Then it utilizes based on study
Machine learning method is trained on training set, is tested on test set, quality characteristic vector is mapped as corresponding
Mass fraction;And then the superiority and inferiority of algorithm is assessed using existing algorithm performance index (SROCC, LCC etc.).
Beneficial effect
View-based access control model proposed by the present invention perception and binocular competition without reference stereo image quality evaluation method, and it is existing
Technology, which is compared, has the features such as subjective consistency is high, and data base-independent is high, and algorithm stability is high;It can be with stereo-picture/view
Frequency processing related application systematic collaboration uses, and has very strong application value.
Detailed description of the invention
Fig. 1 is the process without reference stereo image quality evaluation method of view-based access control model perception and binocular competition of the invention
Figure;
Fig. 2 is the box-like that the present invention and other stereo image quality evaluation methods are tested on LIVE database
Figure.
Specific embodiment
The embodiment of the method for the present invention is described in detail in the following with reference to the drawings and specific embodiments.
Embodiment
The process of this method is as shown in Figure 1, specific implementation process are as follows:
Step 1: converting grayscale information for the stereo pairs to be tested of input.
Step 2: being handled using matching algorithm is further to grayscale information, simulation disparity map and uncertainty are obtained
Figure, while filtering to obtain the filter response of grayscale information using Gabor.
Simulation disparity map matches to obtain by the structural similarity of left and right view grayscale information.
The calculation method of uncertain figure is as follows:
Wherein, l represents left view grayscale image, and r is represented by parallax compensation treated right view grayscale image, μ and σ difference
Represent the mean value and standard deviation of corresponding grayscale image, C1And C2Respectively represent constant term.Simulate disparity map and uncertain figure all
It will be used for subsequent difference of Gaussian image procossing and feature extraction.
Step 3: utilizing grayscale information and its filter response and simulation disparity map correction synthesis single eye images.
The calculation method of single eye images is as follows:
CI (x, y)=Wl(x,y)*Il(x,y)+Wr((x+d),y)*Ir((x+d),y) (2)
Wherein, (x, y) is coordinate, IlAnd IrSolid figure is respectively represented to the grayscale image of left and right view, d represents left and right
The parallax of correspondence mappings pixel between view, CI represent the single eye images of synthesis, WlAnd WrRepresentative image information weight, GElWith
GErRepresent the filter response summation of the left and right view indicated with numeric form.
Step 4: obtaining Gaussian difference component from the different scale space and frequency space of single eye images and uncertain figure
Picture, and complete natural scene statistics and visual perception feature extraction.
The calculation method of difference of Gaussian image is as follows:
σ2 ij=L* σ1 ij (7)
Wherein,Difference of Gaussian image is represented,WithRespectively represent to original image (single eye images or not really
Qualitative figure) carry out the image that the gaussian filtering under different convolution kernels obtains, σ1 ijAnd σ2 ijTwo different convolution kernels are respectively represented,
W and h represents the width and height of image to be processed under a certain scale, and f represents frequency, and i and j respectively represent some scale space and frequency
Rate space.
The extracting method of visual perception feature is as follows:
The extraction of energy feature:
Wherein, the comentropy of H representative image, the quantity of m representative image grey level, plFirst of grey level is represented to go out
Existing probability correlation value.
The extraction of edge feature:
Wherein, Canny represents the edge detection using Canny method progress image, and by qualified edge pixel
Point is indicated with numerical approach.
Step 5: using Step 1: Step 2: the method for step 3 and step 4 is vertical to each width colour in database
Each group of quality characteristic vector corresponding to stereo-picture is calculated to handling in body image;Then it utilizes based on study
Machine learning method is trained on training set, is tested on test set, quality characteristic vector is mapped as corresponding
Mass fraction;And then the superiority and inferiority of algorithm is assessed using existing algorithm performance index (SROCC, LCC etc.).
We implement our algorithm, including LIVE Phase II on three stereo image quality rating databases,
Waterloo IVC 3D Phase I and Phase II.The essential information of these databases is enumerated in table one.Meanwhile I
Have chosen six kinds of algorithms and disclose, the quality evaluation algorithm of excellent performance is compared with our method, including four kinds of 2D bases
Stereo image quality evaluation algorithms on plinth: PSNR, SSIM, MS-SSIM, BRISQUE.One kind is commented with reference to stereo image quality entirely
A kind of valence method C-FR and no reference stereo image quality evaluation method C-NR.In order to eliminate the shadow of training data and randomness
It rings, we have carried out the repetition test of 1000 times 80% training -20% test on the database, i.e., 80% data are for instructing
Practice, remaining 20% data are for testing, and there is no the overlappings of content for training data and test data.Finally using existing
Algorithm performance index (1000 repetition test SRCC, the intermediate value of PCC, RMSE) is assessed the superiority and inferiority of algorithm, experimental result
It is shown in Table two.
One database essential information of table
In conjunction with attached drawing 2, it can be seen that algorithm proposed by the present invention is not only shown in the test of four databases
The subjective consistency more more excellent than other non-reference picture quality appraisement algorithms and stability, in LIVE and TID2013 database
On, the quality evaluating method that even better than refers to entirely.
Algorithm performance compares on two or three databases of table
Claims (6)
1. view-based access control model perception and binocular competition are without reference stereo image quality evaluation method, it is characterised in that: its specific step
It is rapid as follows:
Step 1: converting grayscale information for the stereo pairs to be tested of input;
Step 2: being handled using matching algorithm is further to grayscale information, simulation disparity map and uncertain figure are obtained, together
Shi Liyong Gabor filters to obtain the filter response of grayscale information;
Step 3: utilizing grayscale information and its filter response and simulation disparity map correction synthesis single eye images;
Step 4: difference of Gaussian image is obtained from the different scale space and frequency space of single eye images and uncertain figure,
And complete natural scene statistics and visual perception feature extraction;
Step 5: using Step 1: Step 2: the method for step 3 and step 4 is to each width anaglyph in database
As each group of quality characteristic vector corresponding to stereo-picture is calculated to handling;Then the machine based on study is utilized
Learning method is trained on training set, is tested on test set, and quality characteristic vector is mapped as corresponding quality
Score;And then the superiority and inferiority of algorithm is assessed using existing algorithm performance index S ROCC, LCC.
2. view-based access control model according to claim 1 perception and binocular competition without reference stereo image quality evaluation method,
It is characterized by: colouring information converts to obtain using RGB color in the step 1.
3. view-based access control model according to claim 1 perception and binocular competition without reference stereo image quality evaluation method,
It matches to obtain by the structural similarity of left and right view grayscale information it is characterized by: simulating disparity map in the step 2;
The calculation method of uncertain figure is as follows in the step 2:
Wherein, l represents left view grayscale image, and r is represented by parallax compensation treated right view grayscale image, and μ and σ are respectively represented
The mean value and standard deviation of corresponding grayscale image, C1And C2Respectively represent constant term;Simulation disparity map and uncertain figure will all be used
In subsequent difference of Gaussian image procossing and feature extraction.
4. view-based access control model according to claim 1 perception and binocular competition without reference stereo image quality evaluation method,
It is characterized by: the calculation method of the step 3 median ocellus image is as follows:
CI (x, y)=Wl(x,y)*Il(x,y)+Wr((x+d),y)*Ir((x+d),y) (2)
Wherein, (x, y) is coordinate, IlAnd IrSolid figure is respectively represented to the grayscale image of left and right view, d represents left and right view
Between correspondence mappings pixel parallax, CI represent synthesis single eye images, WlAnd WrRepresentative image information weight, GElAnd GEr
Represent the filter response summation of the left and right view indicated with numeric form.
5. view-based access control model according to claim 1 perception and binocular competition without reference stereo image quality evaluation method,
It is characterized by: the calculation method of difference of Gaussian image is as follows in the step 4:
σ2 ij=L* σ1 ij (7)
Wherein,Difference of Gaussian image is represented,WithIt respectively represents and difference is carried out to single eye images or uncertain figure
The image that gaussian filtering under convolution kernel obtains, σ1 ijAnd σ2 ijTwo different convolution kernels are respectively represented, w and h represent a certain ruler
The width and height of image to be processed, f represent frequency under degree, and i and j respectively represent some scale space and frequency space;
The extracting method of visual perception feature is as follows in the step 4:
The extraction of energy feature:
Wherein, the comentropy of H representative image, the quantity of m representative image grey level, plRepresent the general of first of grey level appearance
Rate correlation;
The extraction of edge feature:
Wherein, Canny represents the edge detection that image is carried out using Canny method, and qualified edge pixel point is used
Numerical approach indicates.
6. view-based access control model according to claim 1 perception and binocular competition without reference stereo image quality evaluation method,
It is characterized by: the machine learning method in the step 5 includes using support vector machines (SVR), neural network machine study
Method.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710886018X | 2017-09-27 | ||
CN201710886018 | 2017-09-27 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107635136A CN107635136A (en) | 2018-01-26 |
CN107635136B true CN107635136B (en) | 2019-03-19 |
Family
ID=61106357
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711003045.4A Active CN107635136B (en) | 2017-09-27 | 2017-10-24 | View-based access control model perception and binocular competition are without reference stereo image quality evaluation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107635136B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108257131A (en) * | 2018-02-24 | 2018-07-06 | 南通大学 | A kind of 3D rendering quality evaluating method |
CN108520510B (en) * | 2018-03-19 | 2021-10-19 | 天津大学 | No-reference stereo image quality evaluation method based on overall and local analysis |
CN108648186B (en) * | 2018-05-11 | 2021-11-19 | 北京理工大学 | No-reference stereo image quality evaluation method based on primary visual perception mechanism |
CN109257593B (en) * | 2018-10-12 | 2020-08-18 | 天津大学 | Immersive virtual reality quality evaluation method based on human eye visual perception process |
CN109325550B (en) * | 2018-11-02 | 2020-07-10 | 武汉大学 | No-reference image quality evaluation method based on image entropy |
CN110517308A (en) * | 2019-07-12 | 2019-11-29 | 重庆邮电大学 | It is a kind of without refer to asymmetric distortion stereo image quality evaluation method |
CN110838120A (en) * | 2019-11-18 | 2020-02-25 | 方玉明 | Weighting quality evaluation method of asymmetric distortion three-dimensional video based on space-time information |
CN113269204B (en) * | 2021-05-17 | 2022-06-17 | 山东大学 | Color stability analysis method and system for color direct part marking image |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105338343A (en) * | 2015-10-20 | 2016-02-17 | 北京理工大学 | No-reference stereo image quality evaluation method based on binocular perception |
-
2017
- 2017-10-24 CN CN201711003045.4A patent/CN107635136B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105338343A (en) * | 2015-10-20 | 2016-02-17 | 北京理工大学 | No-reference stereo image quality evaluation method based on binocular perception |
Non-Patent Citations (3)
Title |
---|
New no-reference stereo image quality method for image communication;Wang Ying等;《2016 IEEE RIVF International Conference on Computing & Communication Technologies, Research, Innovation, and Vision for the Future (RIVF)》;20161229;全文 |
No-Reference Quality Assessment for Stereoscopic Images Based on Binocular Quality Perception;Seungchul Ryu等;《 IEEE Transactions on Circuits and Systems for Video Technology》;20130829;全文 |
No-Reference Quality Assessment of Natural Stereopairs;Ming-Jun Chen等;《 IEEE Transactions on Image Processing》;20130610;全文 |
Also Published As
Publication number | Publication date |
---|---|
CN107635136A (en) | 2018-01-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107635136B (en) | View-based access control model perception and binocular competition are without reference stereo image quality evaluation method | |
CN107578403B (en) | The stereo image quality evaluation method for instructing binocular view to merge based on gradient information | |
CN105338343B (en) | It is a kind of based on binocular perceive without refer to stereo image quality evaluation method | |
CN101610425B (en) | Method for evaluating stereo image quality and device | |
CN109523513B (en) | Stereoscopic image quality evaluation method based on sparse reconstruction color fusion image | |
CN105654465B (en) | A kind of stereo image quality evaluation method filtered between the viewpoint using parallax compensation | |
CN104811693B (en) | A kind of stereo image vision comfort level method for objectively evaluating | |
RU2009110511A (en) | SYSTEM AND METHOD FOR FORMING AND PLAYING THREE-DIMENSIONAL VIDEO IMAGES | |
CN109255358B (en) | 3D image quality evaluation method based on visual saliency and depth map | |
CN109242834A (en) | It is a kind of based on convolutional neural networks without reference stereo image quality evaluation method | |
CN108520510B (en) | No-reference stereo image quality evaluation method based on overall and local analysis | |
Geng et al. | A stereoscopic image quality assessment model based on independent component analysis and binocular fusion property | |
CN104954778A (en) | Objective stereo image quality assessment method based on perception feature set | |
CN110853027A (en) | Three-dimensional synthetic image no-reference quality evaluation method based on local variation and global variation | |
CN107360416A (en) | Stereo image quality evaluation method based on local multivariate Gaussian description | |
CN110246111A (en) | Based on blending image with reinforcing image without reference stereo image quality evaluation method | |
CN103338379A (en) | Stereoscopic video objective quality evaluation method based on machine learning | |
CN104853175B (en) | Novel synthesized virtual viewpoint objective quality evaluation method | |
CN114648482A (en) | Quality evaluation method and system for three-dimensional panoramic image | |
CN107018410B (en) | A kind of non-reference picture quality appraisement method based on pre- attention mechanism and spatial dependence | |
Appina et al. | A full reference stereoscopic video quality assessment metric | |
CN104243970A (en) | 3D drawn image objective quality evaluation method based on stereoscopic vision attention mechanism and structural similarity | |
CN109523508B (en) | Dense light field quality evaluation method | |
CN110691236A (en) | Panoramic video quality evaluation method | |
CN108648186B (en) | No-reference stereo image quality evaluation method based on primary visual perception mechanism |
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 |