CN104103065A - No-reference fuzzy image quality evaluation method based on singular value decomposition - Google Patents

Abstract

Disclosed is a no-reference fuzzy image quality evaluation method based on singular value decomposition. The method is characterized by comprising the following steps: (1), selecting an image as an image to be evaluated; (2), converting the image into a gray scale image; (3), performing singular value decomposition (SVD) on the image to be evaluated to obtain a singular value vector S1; (4), according to a singular value, establishing an image quality evaluation function Blur; and (5), calculating the value of the quality evaluation function Blur of the image to be evaluated and taking the value as a fuzzy index. By using the method provided by the invention, the fuzzy degree of a fuzzy image can be evaluated without any reference images, and the evaluation results accord with mankind visual subjective understanding results. The greater the Blur value is, the better the quality of the image to be evaluated is. The quality of the fuzzy image can be rapidly evaluated without any reference signals, the consistency with mankind subjective visual perception is quite good, and the method is simple.

Description

A kind of based on svd without with reference to blurred picture quality evaluating method

Technical field

The present invention relates to image processing field, specifically, is a kind of method that does not need reference picture to carry out blurred picture quality assessment.

Background technology

Universal along with electronic products such as mobile phone, digital cameras, has produced a large amount of digital images, in the process of taking pictures, because the shake of focal length and hand can cause image blurring.How from these digital images, automatical is selected up-to-standard image, gives up which underproof blurred picture, just need to evaluate picture quality.

According to evaluation procedure, need how many original reference image informations, objective image quality evaluating method can be divided into three major types again: complete in (Full-Reference, FR) image quality evaluation method, partial reference (Reduced-Reference, RR) image quality evaluation method and without with reference to (No-Reference, NR) image quality evaluation method.Full reference and partial reference image quality appraisement method need all or part of information of reference picture, and maybe cannot not obtain all or part of information of reference picture in a lot of application scenarios, so non-reference picture quality appraisement method is more practical.

Summary of the invention

The object of the present invention is to provide a kind of Fast Fuzzy image quality evaluating method, can, in the situation that not needing reference picture, evaluate the quality of blurred picture quality.

To achieve these goals, technical scheme of the present invention is as follows: a kind of based on svd without with reference to blurred picture quality evaluating method, its key is to carry out as follows:

(1) select piece image for being evaluated image;

(2) this image is converted into gray level image;

(3) to being evaluated image, do singular value (SVD) decomposition, obtain singular value vector S ₁;

Any one gray level image can be regarded real number matrix A ∈ R as ^{m * n}, there is quadrature (or tenth of the twelve Earthly Branches) matrix U ∈ R ^{m * m}with V ∈ R ^{n * n}make A=USV ^t(1)

In formula $S=\left[\begin{array}{cc}{S}_1& 0\\ 0& 0\end{array}\right],$ S ₁=diag (σ ₁, σ ₂..., σ _r), and number σ ₁, σ ₂..., σ _rbe all non-zero singular values of matrix A, r=rank (A), the left singular value vector that the column vector of U is matrix A, the right singular value vector that the column vector of V is A, claims that (1) formula is the svd of matrix A.By svd formula, calculate singular value vector S ₁.

(4) according to singular value, set up image quality evaluation function Blur;

$\mathrm{Blur}=\left\{\begin{array}{c}\frac{\underset{S1\left(i\right)>c1}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r>512\\ \frac{\underset{S1\left(i\right)>c2}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r\≤512\end{array}\right.---\left(2\right)$

In formula, S ₁for the singular value vector of image, r is singular value number, and we need to set a threshold value for singular value, if singular value number is greater than 512, and the value that c1 is 500, if singular value number is less than or equal to 512, the value that c2 is 70;

(5) calculate the value of the quality assessment function Blur that is evaluated image, as fuzzy indicator.

Advantage of the present invention is: the present invention does not need the quality that reference picture just can Fast Evaluation one width blurred picture, and better with mankind's subjective vision perception consistance, and method is simple.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention.

Embodiment

Below in conjunction with drawings and Examples, further the present invention is illustrated.

Embodiment:

As shown in Figure 1: a kind of based on svd without with reference to blurred picture quality evaluating method, it is characterized in that carrying out as follows:

(1) select piece image for being evaluated image;

(2) this image is converted into gray level image;

(3) to being evaluated image, do singular value (SVD) decomposition, obtain singular value vector S ₁;

Any one gray level image can be regarded real number matrix A ∈ R as ^{m * n}, there is quadrature (or tenth of the twelve Earthly Branches) matrix U ∈ R ^{m * m}with V ∈ R ^{n * n}make A=USV ^t(1)

In formula $S=\left[\begin{array}{cc}{S}_1& 0\\ 0& 0\end{array}\right],$ S ₁=diag (σ ₁, σ ₂..., σ _r), and number σ ₁, σ ₂..., σ _rbe all non-zero singular values of matrix A, r=rank (A), the left singular value vector that the column vector of U is matrix A, the right singular value vector that the column vector of V is A, claims that (1) formula is the svd of matrix A.By svd formula, calculate singular value vector S ₁.

(4) according to singular value, set up image quality evaluation function Blur;

$\mathrm{Blur}=\left\{\begin{array}{c}\frac{\underset{S1\left(i\right)>c1}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r>512\\ \frac{\underset{S1\left(i\right)>c2}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r\≤512\end{array}\right.---\left(2\right)$

In formula, S ₁for the singular value vector of image, r is singular value number, and we need to set a threshold value for singular value, if singular value number is greater than 512, and the value that c1 is 500, if singular value number is less than or equal to 512, the value that c2 is 70;

(5) calculate the value of the quality assessment function Blur that is evaluated image, as fuzzy indicator, more key diagram picture is more clear for Blur value.

In order to verify the superiority of the inventive method, on true fuzzy database, test below.True blurred picture database (BID-Blurred Image Database[Online]

Available:http: //www.lps.ufrj.br/profs/eduardo/ImageDatabase.htm), have 585 width images, pixel coverage from 1280 * 960 to 2272 * 1704, image in database is directly taken from client's camera, and these images are divided into 5 types: not fuzzy 204 width, lose fuzzy 142 width of focus, fuzzy 57 width of simple motion, fuzzy 63 width of compound movement and other 119 width.True fuzzy database provides rough MOS to obtain score value.In order to test the consistance of the present invention and subjective perception, we have selected two kinds of measurement criterions: (1) Spearman rank order coefficient of relationship (SROCC), the monotonicity of reflection objective evaluating prediction achievement; (2) related coefficient (CC), the accuracy of reflection objective evaluating.Final testing result is presented at table 1, as can be seen from the table, the method that the present invention proposes is better than document [1] (Ferzli Rony and Karam L J.A No-Reference Objective Image Sharpness Metric Based on the Notion of Just Noticeable Blur (JNB) .IEEE Transactions on Image Processing on true fuzzy database, 2009, 18 (4): 717-728) and document [2] (Li C F, Yuan W, Bovik A C and Wu X.No-reference blur index using blur comparisons.Electronics Letters, 2011, 47 (17): 962-963), there are extraordinary performance index.

Table 1 different without with reference to blurred picture evaluation method, the Performance Ratio on True Data storehouse is

Claims (3)

1. Based on svd without with reference to a blurred picture quality evaluating method, it is characterized in that carrying out as follows:

   (1) select piece image for being evaluated image;

   (2) this image is converted into gray level image;

   (3) to being evaluated image, do singular value (SVD) decomposition, obtain singular value vector S ₁;

   (4) according to singular value, set up image quality evaluation function Blur;

   (5) calculate the value of the quality assessment function Blur that is evaluated image, as fuzzy indicator.
2. According to claim 1 based on svd without with reference to blurred picture quality evaluating method, it is characterized in that: in step (three), svd formula is A=USV ^t, in formula $S=\left[\begin{array}{cc}{S}_1& 0\\ 0& 0\end{array}\right],$ S ₁=diag (σ ₁, σ ₂..., σ _r), and number σ ₁, σ ₂..., σ _rbe all non-zero singular values of matrix A, r=rank (A), the left singular value vector that the column vector of U is matrix A, the right singular value vector that the column vector of V is A.
3. According to claim 1 based on svd without with reference to blurred picture quality evaluating method, it is characterized in that: in step (four), image quality evaluation function is $\mathrm{Blur}=\left\{\begin{array}{c}\frac{\underset{S1\left(i\right)>c1}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r>512\\ \frac{\underset{S1\left(i\right)>c2}{\mathrm{\Σ}}(1/S_1\left(i\right))}{r},r\≤512\end{array}\right.,$ In formula, S ₁for the singular value vector of image, r is singular value number, and we need to set a threshold value for singular value, if singular value number is greater than 512, and the value that c1 is 500, if singular value number is less than or equal to 512, the value that c2 is 70.