CN108416770A

CN108416770A - A kind of image quality evaluating method of view-based access control model conspicuousness

Info

Publication number: CN108416770A
Application number: CN201810185626.2A
Authority: CN
Inventors: 王靖宇; 王霰禹; 张科; 姜海旭; 张彦华; 王景鹏
Original assignee: Northwestern Polytechnical University
Current assignee: Northwestern Polytechnical University
Priority date: 2018-03-07
Filing date: 2018-03-07
Publication date: 2018-08-17
Anticipated expiration: 2038-03-07
Also published as: CN108416770B

Abstract

The present invention provides a kind of image quality evaluating methods of view-based access control model conspicuousness, it is related to technical field of image processing, the present invention first calculates phase equalization, the gradient similarity graph between shade of gray figure and shade of gray figure is calculated again, distribution consistency degree is asked to gradient similarity graph, calculate mass fraction, as quality evaluation score of the distorted image with respect to reference picture.The beneficial effects of the present invention are vision significance concept is introduced in image quality evaluation, pass through the vision significance of distribution consistency degree description disturbance in the picture of Analysis interference in the picture, to embody subjective impact degree of the interference in image to people's vision, more accurate evaluation picture quality is come with this.

Description

A kind of image quality evaluating method of view-based access control model conspicuousness

Technical field

The present invention relates to technical field of image processing, especially a kind of image quality evaluating method.

Background technology

Basic technology one of of the image quality evaluation in image processing field, compares Algorithm Analysis and system performance is commented It is by comparing the difference between distorted image and original image that estimate etc., which has important role, full reference image quality appraisement, A kind of method that the different quality to distorted image is evaluated is widely used in evaluation image coding and compresses, at guide image In the fields such as adjustment method and the monitoring of the picture quality of user terminal.

Wen Xin, Zhang Wanyi etc. (《The full reference image quality evaluation algorithm of view-based access control model perception》, electronic surveying and instrument Report, 2016,30 (11):1780-1789) on the basis of traditional images quality evaluating method, using human eye for image border Extremely sensitive vision system perception characteristics, proposition describe the subjective perception difference of people with edge feature, and this method is only for few There is higher performance in the evaluation of number compression algorithm, still not accurate enough, shortage subjectivity is generally described to the subjective perception of human eye Consistency.

Invention content

For overcome the deficiencies in the prior art, the present invention is proposed through the aobvious of local structural variation dense degree description disturbance Work property, in conjunction with local structural variation intensity, variation total amount, variation pixel and contour line relationship, overall merit picture quality, to carry The accuracy rate of high praise.

The technical solution adopted by the present invention to solve the technical problems is as follows：

Testing image and reference picture are respectively converted into gray level image I by step 1_distAnd I_ref, to two gray level images I_distAnd I_refIts phase equalization PC is calculated separately under frequency domain_distAnd PC_ref；

Step 2, shade of gray figure calculation formula are as follows：

Wherein G represents gradient map, d_xAnd d_yFor the gradient template of horizontal and vertical both direction：

Wherein, a₁And a₂Meet：

I is gray level image, and gray level image I can be acquired respectively according to formula (1)_distAnd I_refShade of gray figure G_distWith G_ref；

Step 3 calculates two width shade of gray figure G_distAnd G_refBetween gradient similarity graph S_G, gradient similarity graph S_GIn The value S of each pixel p_G(p) it is：

G_dist(p) and G_ref(p) that respectively represent is figure G_distAnd G_refThe gray value of middle pixel p；Wherein, T₂＞ 0, from And acquire the Weighted Similarity S of two width shade of gray figures_GPCFor：

Wherein, max (PC_dist,PC_ref) it is PC_distAnd PC_refThe matrix that middle corresponding element maximum value is constituted, S_GBy formula (2) it calculates and determines；

Step 4, to gradient similarity graph S_GSeek distribution consistency degree S_D：

First by gradient similarity graph S_GDown-sampling, by its successive segmentation at the square piecemeal B that the n length of side is even number_i, Wherein i=1,2,3 ... n, each piecemeal B_iThe maximum value of middle pixel gray value is as corresponding position pixel after down-sampling Gray value, to obtain down-sampling figure S_G(max), size n_x×n_y, the gray value of each pixel is p in down-sampling figure_x,y, X and y indicates pixel in S_G(max)In abscissa and ordinate, calculate its distribution consistency degree S_D：

S_D=| S₁+S₂|+|S₃+S₄|

Wherein S₁,S₂,S₃,S₄For intermediate variable：

Step 5：Calculate mass fractionWherein 0 ＜ σ ＜ 10, S_qualityAs distorted image The quality evaluation score of opposite reference picture.

The beneficial effects of the present invention are vision significance concept is introduced in image quality evaluation, existed by Analysis interference The vision significance of distribution consistency degree description disturbance in the picture in image, to embody the interference in image to people's vision Subjective impact degree, more accurate evaluation picture quality is come with this.

Description of the drawings

Fig. 1 is the flow chart of the present invention.

Specific implementation mode

Present invention will be further explained below with reference to the attached drawings and examples.

Human visual perception system is more sensitive to high-frequency information, therefore the sky of picture material caused by various disturbance factors Between structural changed significance degree, undoubtedly influence picture quality subjective feeling central factor.Have at present complete Reference image quality appraisement method is only started with from image difference specific analysis, this concept of concern conspicuousness.

The present invention is based on the image quality evaluating method flow charts of vision significance as shown in Figure 1, specific implementation mode packet Include following steps：

Step 1：For original reference picture and the testing image that distortion processing obtains is carried out by it and is converted to gray-scale map As I_distAnd I_ref, two images are subjected to Fourier transformation and calculate its phase equalization PC_distAnd PC_ref。

Step 2：Calculate image gradient figure.Define horizontal and vertical gradient template d_xAnd d_yIt is as follows：

Convolutional calculation gradient map is carried out to image using gradient templateWhen I takes respectively I_distAnd I_refWhen, respectively obtain G_distAnd G_ref。

Step 3：Calculate Weighted Similarity S_GPC.Gradient similarity S is calculated first_G, for each pixel in gradient map P acquires its similarityWherein take T₂=160.

Then Weighted Similarity S is calculated_GPC：

Wherein max (PC_dist,PC_ref) it is PC_distAnd PC_refThe matrix that middle corresponding element maximum value is constituted.

Step 4：Seek distribution consistency degree S_D。

First by similarity graph S_GDown-sampling, by its successive segmentation at the piecemeal of several 8 × 8 pixel sizes, each piecemeal Gray value of the maximum value of middle pixel gray value as corresponding position pixel in down-sampling figure, obtains down-sampling figure S_G(max), Its size is n_x×n_y, the gray value of each pixel is p in figure_x,_y, x and y indicate pixel in S_G(max)In abscissa and Ordinate calculates its distribution consistency degree S_D：

S_D=| S₁+S₂|+|S₃+S₄|

Wherein, S₁,S₂,S₃,S₄For intermediate variable：

Step 5：Calculate quality evaluation scoreWherein σ values are 4, S_qualityAs it is distorted Quality evaluation score of the image with respect to reference picture.

Claims

1. a kind of image quality evaluating method of view-based access control model conspicuousness, it is characterised in that include the following steps：

Testing image and reference picture are respectively converted into gray level image I by step 1_distAnd I_ref, to two gray level image I_distWith I_refIts phase equalization PC is calculated separately under frequency domain_distAnd PC_ref；

Step 2, shade of gray figure calculation formula are as follows：

Wherein, a₁And a₂Meet：

I is gray level image, and gray level image I can be acquired respectively according to formula (1)_distAnd I_refShade of gray figure G_distAnd G_ref；

Step 3 calculates two width shade of gray figure G_distAnd G_refBetween gradient similarity graph S_G, gradient similarity graph S_GEach of The value S of pixel p_G(p) it is：

G_dist(p) and G_ref(p) that respectively represent is figure G_distAnd G_refThe gray value of middle pixel p；Wherein, T₂＞ 0, to ask Obtain the Weighted Similarity S of two width shade of gray figures_GPCFor：

Wherein, max (PC_dist,PC_ref) it is PC_distAnd PC_refThe matrix that middle corresponding element maximum value is constituted, S_GIt is counted by formula (2) It calculates and determines；

First by gradient similarity graph S_GDown-sampling, by its successive segmentation at the square piecemeal B that the n length of side is even number_i, wherein i =1,2,3 ... n, each piecemeal B_iGray scale of the maximum value of middle pixel gray value as corresponding position pixel after down-sampling Value, to obtain down-sampling figure S_G(max), size n_x×n_y, the gray value of each pixel is p in down-sampling figure_x,y, x and y Indicate pixel in S_G(max)In abscissa and ordinate, calculate its distribution consistency degree S_D：

S_D=| S₁+S₂|+|S₃+S₄|

Wherein S₁,S₂,S₃,S₄For intermediate variable：

Step 5：Calculate mass fractionWherein 0 ＜ σ ＜ 10, S_qualityAs distorted image is opposite The quality evaluation score of reference picture.