CN107578395A - The image quality evaluating method that a kind of view-based access control model perceives - Google Patents
The image quality evaluating method that a kind of view-based access control model perceives Download PDFInfo
- Publication number
- CN107578395A CN107578395A CN201710771347.XA CN201710771347A CN107578395A CN 107578395 A CN107578395 A CN 107578395A CN 201710771347 A CN201710771347 A CN 201710771347A CN 107578395 A CN107578395 A CN 107578395A
- Authority
- CN
- China
- Prior art keywords
- image
- access control
- view
- control model
- reference picture
- 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
Landscapes
- Image Analysis (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The present invention provides the image quality evaluating method that a kind of view-based access control model perceives, and comprises the following steps:Reconstructed image is adjusted to and reference picture identical size using local SIFT methods of estimation, GMS figures are obtained by calculating the gradient amplitude of reference picture and reconstructed image, the Saliency maps of reference picture are obtained with the method for vision noticing mechanism, have studied the convergence strategy based on average value respectively, view-based access control model pays attention to the strategy of weight, the convergence strategy based on mean square error, finally integrate various convergence strategies obtain the overall evaluation method of image.Vision noticing mechanism is applied in reconstructed image quality evaluation by algorithm, is indicated the conspicuousness local message saved in reconstructed image in how many reference chart, is more met the method for human subject's evaluation image.Present invention can apply in the quality evaluation of all reconstructed images.
Description
Technical field
The present invention relates to technical field of computer vision, and in particular to the image quality evaluation side that a kind of view-based access control model perceives
Method.
Background technology
With developing rapidly for mechanics of communication, image is constantly transmitted repeatedly in different equipment, and machine vision
Requirement of the treatment technology to input picture also more and more higher, hence it is imperative that image can be measured in transmitting procedure by finding one kind
The method of middle loss amount, image quality evaluation are exactly to grow up on this basis, mainly by carrying out characteristic point to image
Analysis research, then evaluate the distortion level of image.Image quality evaluation is in image processing system, for Algorithm Analysis ratio
Play the role of compared with, System Performance Analysis etc. important.Recently, multimedia reconfiguration technique draws in the research of figure and vision
Played great concern, the reconfiguration technique of image and video on different display screens in different sizes or length-width ratio is shown
Original scene, therefore the quality for evaluating reconstructed image is significant.
Common picture appraisal algorithm, as Y-PSNR (PSNR) and structural similarity (SSIM) are not suitable for reconstruct image
Picture, because these algorithms need to refer to image and reconstructed image and have identical size.Existing reconstruction quality evaluation method master
Including:Color description (CL), edge histogram (EH), two-way torsion resistance (BDW), two-way similitude (BDS).CL and EH are
Based on the two methods of MPEG-7 standards, CL uses distribution of color histogram, and as characterization image, EH uses marginal information conduct
Characterization image, their estimating using the distance between two histograms as two image similarity.BDW and BDS lead to respectively
The pairing crossed between Monotone Mappings and two-way non-monotonic mapping progress image block, defined in the average distance of all image blocks
BDW distances and BDS distances.
Although the above method has obtained preferable evaluation result to a certain extent, they are still based on reference chart
As and reconstructed image consider that but the mankind are only the final evaluator of picture quality in itself, picture appraisal should add mankind master
The factor of sight, so researcher is seeking the automatic evaluation method consistent with subjective assessment always.
The content of the invention
In view of this, a kind of good the embodiment provides evaluation result, addition human subject's factor base
In the image quality evaluating method of visually-perceptible.
The present invention provides the image quality evaluating method that a kind of view-based access control model perceives, and comprises the following steps:
S1, reference picture and reconstructed image are obtained, be adjusted to reconstructed image and reference picture by local SIFT estimations
Identical size, obtain the first reference picture and the first reconstructed image;
S2, by calculating the direction gradient amplitude of first reference picture and the first reconstructed image obtain gradient amplitude phase
Scheme like degree;
S3, view-based access control model attention mechanism structure visual attention model, reference picture is obtained according to the visual attention model
Saliency maps;
S4, according to the Saliency maps and gradient amplitude similarity graph existing convergence strategy is assessed, assessed
As a result;
S5, according to the assessment result, build the quality score of reconstructed image.
Further, in the step S2, the gradient of image is calculated using Scharr operators, the Scharr operators are along dampening
Square it can be write as respectively to vertical direction:
Further, the gradient amplitude calculating formula of similarity is:
Further, in the calculation formula of the gradient amplitude similarity, c is to maintain the constant term of numerical stability, wherein
mref(x, y) is the first reference picture SrefIn the direction gradient amplitude of position (x, y), mret(x, y) is the first reconstructed image Sret
In the direction gradient amplitude of position (x, y)..
Further, in the step S3, the visual attention model is to pay attention to mould in classical Itti model-based visions
The visual attention model of textural characteristics is added on the basis of type.
Further, in the step S4, the existing convergence strategy includes:Convergence strategy based on average value;Base
In the convergence strategy of vision attention weight;Convergence strategy based on mean square error.
Further, in the step S5, the calculation formula of the quality score is:Its
Middle GMSS is the image quality evaluation index of the convergence strategy of the vision attention weight, and GMSD is to be described based on mean square error
The image quality evaluation index of convergence strategy, d are to maintain the constant of numerical stability.
The present invention by vision noticing mechanism be applied to reconstructed image quality evaluation in, indicate saved in reconstructed image it is more
Conspicuousness local message in few reference chart, more meet the method for human subject's evaluation image;Picture quality provided by the invention
Evaluation method is with a wide range of applications, and evaluation result is good.
Brief description of the drawings
Fig. 1 is the flow chart for the image quality evaluating method that a kind of view-based access control model of the present invention perceives.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing to embodiment party of the present invention
Formula is further described.
Fig. 1 is refer to, the embodiment provides the image quality evaluating method that a kind of view-based access control model perceives, the party
Method comprises the following steps:
Step S1, reference picture and reconstructed image are obtained, passes through local SIFT (Scale-invariant feature
Transform, Scale invariant features transform operator) estimation by reconstructed image be adjusted to reference picture identical size, obtain
First reference picture and the first reconstructed image.
Specifically, in step S1, when the size difference of reference picture and reconstructed image, original gradient amplitude similarity
Figure can not be used directly the similitude calculated between image, so needing the matching algorithm established between two images pixel;
By the present invention in that with SIFT stream method find reconstructed image in point (x, y) corresponding to point (x ', y '), will pass through SIFT flow
Reference picture and reconstructed image after conversion are designated as the first reference picture S respectivelyrefWith the first reconstructed image Sret。
Step S2, by calculating the first reference picture SrefWith the first reconstructed image SretDirection gradient amplitude obtain gradient
Amplitude similarity graph (Gradient Magnitude Similarity, GMS);
In Digital Image Processing, gradient information usually is represented with convolution mask such as linear filtering, gradient magnitude is by edge
The square root of the Grad of two vertical direction to represent, in order to strengthen response of the image to gradient and speed-up computation, this hair
The bright gradient that image is calculated using Scharr operators, Scharr operators along can horizontally and vertically write respectively
Into:
First reference picture SrefWith the first reconstructed image SretBy the convolution with horizontal direction h and vertical direction v respectively,
The gradient of horizontal and vertical directions can be respectively obtained, remembers the first reference picture SrefIn the first direction ladder of position (x, y)
Degree amplitude is mref(x, y), remember the first reconstructed image SretIt is m in the second direction gradient amplitude of position (x, y)ret(x, y), the
One direction gradient amplitude mref(x, y) and second direction gradient amplitude mretThe calculation formula of (x, y) is:
In formula,Represent convolution algorithm;
According to first direction gradient amplitude mref(x, y) and second direction gradient amplitude mret(x, y) calculates gradient amplitude phase
Scheme GMS like degree, gradient amplitude similarity graph GMS calculation formula is:
In formula, c is to maintain the constant term of numerical stability.
Step S3, visual attention model is built, the Saliency maps of reference picture are obtained according to visual attention model;
Specifically, when people observe the world, the part rather than entirety of scene are often selectively observed,
Vision attention is a kind of psychological phenomena that attentional selection is carried out using the visual information obtained, the mankind and other has vision system
Animal residing for living environment it is complicated and changeable, while enter their visuals field visual information be magnanimity, and and it is not all
Information is all of equal importance, and this, which allows for them, must distribute to limited resource priority in a few significant region.
During very long biological evolution, for the needs of race's existence and procreation, it is desirable to which vision system is to acquisition
Information progress is selective, is handled in real time, and vision noticing mechanism requires precisely in order to adapting to external survival environment,
The product formed during long-term biological evolution, Itti models are classical visual attention models.
Vision attention is incorporated into the subjective assessment more one of the objective evaluation and image that can make image in image evaluation
Cause.Because image gradient is more sensitive to image texture, therefore invention increases the Itti moulds that textural characteristics extend classics
Type visual attention model.
Step S4, existing convergence strategy is assessed according to the Saliency maps and gradient amplitude similarity graph, obtained
Assessment result;
Existing convergence strategy includes following 3 kinds:
S4a, the convergence strategy based on average value:
Specifically, local quality figure can represent the local quality of each image block, therefore can be by merging local quality
The strategy of figure carrys out the total quality of evaluation image.Common convergence strategy is average method, by the scoring of all topographies
Addition averagely obtains final quality score, is designated as mean value method (GMSM):
In formula, M and N represent the line number and columns of reference picture respectively, and higher GMSM scorings mean reconstructed image
Higher quality, however, part all in image is regarded as of equal importance by mean value method, it is clear that do not meet the one of evaluation image
As method.
S4b, view-based access control model pay attention to the convergence strategy of weight:
Specifically, the effect of image quality evaluation can be lifted by using the data of eyes tracking.Therefore, using notable
Property figure predicts the position of human eye sight, and the value for making the conspicuousness at position (x, y) place is S (x, y), and GMS value is GMS (x, y),
Define the fusion method (GMSS) that a view-based access control model pays attention to weight:
S4c, the convergence strategy based on mean square error:
Specifically, in general, natural image can include various possible partial structurtes, so different space structures exists
It may understand during reconstruct in gradient by different degrees of degraded.For example, reconstructed image includes obstruction, distortion and obscured
Deng being blocked in smooth region causes more degradeds, and the fuzzy more degenerations that can cause texture region.However, GMSS ignores
In image the fact that diverse location quality degradation difference, overall image quality scoring ought to response diagram picture all positions degeneration
Situation, so the present invention is using GMS mean square deviations (GMSD) as a kind of index of quality evaluation:
Above-mentioned formula reflects the degree of the degraded in reconstructed image, and GMSD scores are higher, more serious, picture quality of degenerating
It is poorer.
Step S5, based on the assessment result in step S4, the quality score of reconstructed image is built.
Specifically, in summary, GMSM does not account for saliency information, it is believed that all parts of image are heavy on an equal basis
Will, it is that a kind of evenness of picture quality is measured.Vision significance is regarded as weights by GMSS, is fused in GMS figures, and what is obtained comments
Valency method contains visually-perceptible mechanism, meets the custom of mankind's evaluation image.GMSD considers the degraded in reconstructed image
Degree, using GMS mean square deviations as a kind of index of quality evaluation.The present invention proposes a kind of fusion GMSS and GMSD new matter
Appraisal procedure is measured, obtains a kind of scoring (QS) of new reconstructed image:
Wherein d is equally to maintain the constant of numerical stability.
The present invention by vision noticing mechanism be applied to reconstructed image quality evaluation in, indicate saved in reconstructed image it is more
Conspicuousness local message in few reference chart, more meet the method for human subject's evaluation image;Picture quality provided by the invention
Evaluation method is with a wide range of applications, and evaluation result is good.
In the case where not conflicting, the feature in embodiment and embodiment herein-above set forth can be combined with each other.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.
Claims (6)
- A kind of 1. image quality evaluating method that view-based access control model perceives, it is characterised in that:Comprise the following steps:S1, reference picture and reconstructed image are obtained, estimate reconstructed image being adjusted to identical with reference picture by local SIFT Size, obtain the first reference picture and the first reconstructed image;S2, by calculating the direction gradient amplitude of first reference picture and the first reconstructed image obtain gradient amplitude similarity Figure;S3, view-based access control model attention mechanism structure visual attention model, the aobvious of reference picture is obtained according to the visual attention model Work property figure;S4, according to the Saliency maps and gradient amplitude similarity graph existing convergence strategy is assessed, obtain assessment result;S5, according to the assessment result, build the quality score of reconstructed image.
- 2. the image quality evaluating method that a kind of view-based access control model as claimed in claim 1 perceives, it is characterised in that:In the step In rapid S2, the gradient of image is calculated using Scharr operators, the Scharr operators are along horizontally and vertically dividing It can not write as:
- 3. the image quality evaluating method that a kind of view-based access control model as claimed in claim 2 perceives, it is characterised in that:The gradient The calculation formula of amplitude similarity isThe gradient amplitude similarity Calculation formula in, c is to maintain the constant term of numerical stability, wherein mref(x, y) is the first reference picture SrefAt position (x, y) Direction gradient amplitude, mret(x, y) is the first reconstructed image SretIn the direction gradient amplitude of position (x, y).
- 4. the image quality evaluating method that a kind of view-based access control model as claimed in claim 1 perceives, it is characterised in that:In the step In rapid S3, the visual attention model is to add regarding for textural characteristics on the basis of the Itti model-based vision attention models of classics Feel attention model.
- 5. the image quality evaluating method that a kind of view-based access control model as claimed in claim 1 perceives, it is characterised in that:In the step In rapid S4, the existing convergence strategy includes:Convergence strategy based on average value;View-based access control model pays attention to the convergence strategy of weight; Convergence strategy based on mean square error.
- 6. the image quality evaluating method perceived as weighed a kind of view-based access control model as claimed in claim 6, it is characterised in that:Institute State in step S5, the calculation formula of the quality score is:Wherein GMSS weighs for the vision attention The image quality evaluation index of the convergence strategy of weight, GMSD are the image quality evaluation of the convergence strategy based on mean square error Index, d are to maintain the constant of numerical stability.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710771347.XA CN107578395B (en) | 2017-08-31 | 2017-08-31 | Image quality evaluation method based on visual perception |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710771347.XA CN107578395B (en) | 2017-08-31 | 2017-08-31 | Image quality evaluation method based on visual perception |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107578395A true CN107578395A (en) | 2018-01-12 |
CN107578395B CN107578395B (en) | 2019-12-17 |
Family
ID=61030324
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710771347.XA Active CN107578395B (en) | 2017-08-31 | 2017-08-31 | Image quality evaluation method based on visual perception |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107578395B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416770A (en) * | 2018-03-07 | 2018-08-17 | 西北工业大学 | A kind of image quality evaluating method of view-based access control model conspicuousness |
CN108549872A (en) * | 2018-04-17 | 2018-09-18 | 福州大学 | A kind of vision attention fusion method being suitable for redirecting image quality measure |
CN109389591A (en) * | 2018-09-30 | 2019-02-26 | 西安电子科技大学 | Color image quality evaluation method based on colored description |
CN109711352A (en) * | 2018-12-28 | 2019-05-03 | 中国地质大学(武汉) | Vehicle front road environment based on geometry convolutional neural networks has an X-rayed cognitive method |
CN110858394A (en) * | 2018-08-20 | 2020-03-03 | 深圳云天励飞技术有限公司 | Image quality evaluation method and device, electronic equipment and computer readable storage medium |
CN114399495A (en) * | 2022-01-14 | 2022-04-26 | 平安普惠企业管理有限公司 | Image definition calculation method, device, equipment and storage medium |
CN114511518A (en) * | 2022-01-20 | 2022-05-17 | 重庆大学 | Method and device for evaluating visual security of image, electronic equipment and storage medium |
CN111462028B (en) * | 2020-03-16 | 2023-06-27 | 中国地质大学(武汉) | Infrared and visible light image fusion method based on phase consistency and target enhancement |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104574399A (en) * | 2015-01-06 | 2015-04-29 | 天津大学 | Image quality evaluation method based on multi-scale vision significance and gradient magnitude |
CN105354835A (en) * | 2015-10-16 | 2016-02-24 | 浙江工业大学 | Method for evaluating medical image quality in combination with phase consistency, gradient magnitude and structural prominence |
US20160104054A1 (en) * | 2014-10-08 | 2016-04-14 | Adobe Systems Incorporated | Saliency Map Computation |
CN106920232A (en) * | 2017-02-22 | 2017-07-04 | 武汉大学 | Gradient similarity graph image quality evaluation method and system based on conspicuousness detection |
-
2017
- 2017-08-31 CN CN201710771347.XA patent/CN107578395B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160104054A1 (en) * | 2014-10-08 | 2016-04-14 | Adobe Systems Incorporated | Saliency Map Computation |
CN104574399A (en) * | 2015-01-06 | 2015-04-29 | 天津大学 | Image quality evaluation method based on multi-scale vision significance and gradient magnitude |
CN105354835A (en) * | 2015-10-16 | 2016-02-24 | 浙江工业大学 | Method for evaluating medical image quality in combination with phase consistency, gradient magnitude and structural prominence |
CN106920232A (en) * | 2017-02-22 | 2017-07-04 | 武汉大学 | Gradient similarity graph image quality evaluation method and system based on conspicuousness detection |
Non-Patent Citations (1)
Title |
---|
QIANGQIANG ZHOU ETC.: ""Saliency-Based Image Quality Assessment Metric"", 《THE 2016 3RD INTERNATIONAL CONFERENCE ON SYSTEMS AND INFORMATICS》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416770A (en) * | 2018-03-07 | 2018-08-17 | 西北工业大学 | A kind of image quality evaluating method of view-based access control model conspicuousness |
CN108416770B (en) * | 2018-03-07 | 2021-06-11 | 西北工业大学 | Image quality evaluation method based on visual saliency |
CN108549872A (en) * | 2018-04-17 | 2018-09-18 | 福州大学 | A kind of vision attention fusion method being suitable for redirecting image quality measure |
CN108549872B (en) * | 2018-04-17 | 2022-03-22 | 福州大学 | Visual attention fusion method suitable for quality evaluation of redirected image |
CN110858394A (en) * | 2018-08-20 | 2020-03-03 | 深圳云天励飞技术有限公司 | Image quality evaluation method and device, electronic equipment and computer readable storage medium |
CN110858394B (en) * | 2018-08-20 | 2021-03-05 | 深圳云天励飞技术有限公司 | Image quality evaluation method and device, electronic equipment and computer readable storage medium |
CN109389591A (en) * | 2018-09-30 | 2019-02-26 | 西安电子科技大学 | Color image quality evaluation method based on colored description |
CN109389591B (en) * | 2018-09-30 | 2020-11-20 | 西安电子科技大学 | Color descriptor-based color image quality evaluation method |
CN109711352A (en) * | 2018-12-28 | 2019-05-03 | 中国地质大学(武汉) | Vehicle front road environment based on geometry convolutional neural networks has an X-rayed cognitive method |
CN111462028B (en) * | 2020-03-16 | 2023-06-27 | 中国地质大学(武汉) | Infrared and visible light image fusion method based on phase consistency and target enhancement |
CN114399495A (en) * | 2022-01-14 | 2022-04-26 | 平安普惠企业管理有限公司 | Image definition calculation method, device, equipment and storage medium |
CN114511518A (en) * | 2022-01-20 | 2022-05-17 | 重庆大学 | Method and device for evaluating visual security of image, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN107578395B (en) | 2019-12-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107578395A (en) | The image quality evaluating method that a kind of view-based access control model perceives | |
EP3631570A2 (en) | Method, device and computer program for virtually adjusting a spectacle frame | |
CN110879982B (en) | Crowd counting system and method | |
CN108830145A (en) | A kind of demographic method and storage medium based on deep neural network | |
CN108389192A (en) | Stereo-picture Comfort Evaluation method based on convolutional neural networks | |
CN101071473A (en) | Feature point detector and its method | |
CN107886507B (en) | A kind of salient region detecting method based on image background and spatial position | |
CN109711268B (en) | Face image screening method and device | |
CN105825499A (en) | Reference plane determination method and determination system | |
CN108470178B (en) | Depth map significance detection method combined with depth credibility evaluation factor | |
CN111292336B (en) | Omnidirectional image non-reference quality evaluation method based on segmented spherical projection format | |
CN113313732A (en) | Forward-looking scene depth estimation method based on self-supervision learning | |
CN108052989A (en) | A kind of image classification method based on Spline convolution neutral net | |
CN103218812A (en) | Method for rapidly acquiring tree morphological model parameters based on photogrammetry | |
WO2022126674A1 (en) | Method and system for evaluating quality of stereoscopic panoramic image | |
CN113298129B (en) | Polarized SAR image classification method based on superpixel and graph convolution network | |
WO2022213761A1 (en) | Image processing method and apparatus, electronic device, and storage medium | |
CN104751406A (en) | Method and device used for blurring image | |
WO2022088572A1 (en) | Model training method, image processing and alignment method, apparatus, device, and medium | |
CN103679662B (en) | Based on the right super-resolution image restoration method of classification priori non-negative sparse coding dictionary | |
CN108259893A (en) | Virtual reality method for evaluating video quality based on double-current convolutional neural networks | |
CN106709862B (en) | A kind of image processing method and device | |
CN103971365B (en) | A kind of image saliency map extracting method | |
CN115564915A (en) | Map construction method and device for environment digital area of transformer substation | |
CN110415211B (en) | Blind reference three-dimensional grid quality evaluation method based on map features and space features |
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 |