CN109584242A - Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method - Google Patents
Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method Download PDFInfo
- Publication number
- CN109584242A CN109584242A CN201811411545.6A CN201811411545A CN109584242A CN 109584242 A CN109584242 A CN 109584242A CN 201811411545 A CN201811411545 A CN 201811411545A CN 109584242 A CN109584242 A CN 109584242A
- Authority
- CN
- China
- Prior art keywords
- image
- entropy
- divergence
- maximum entropy
- intensity
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to field of image processings, to propose that a kind of no reference contrast distorted image quality evaluating method, this method and subjective assessment score have higher consistency, can effectively evaluate contrast distortion image.For this reason, the technical scheme adopted by the present invention is that maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method, steps are as follows: (1) maximum entropy calculates；(2) mean intensity histogram calculation；(3) change of intensity distribution is measured；(4) quality evaluation.Present invention is mainly applied to image procossing occasions.
Description
Technical field
The invention belongs to field of image processings, more particularly, to a kind of reference-free quality evaluation side of contrast distortion image
Method.More particularly to maximum entropy and KL divergence without reference contrast distorted image quality evaluating method.
Background technique
Image is a kind of presentation mode of information, plays key player in people's daily life.In recent years, with
Electronic image pickup apparatus it is increasing, image data is civilian and industrial all there is be widely applied.However, actually obtaining
During obtaining image, usually due to low cost, low-quality imaging sensor, very poor lighting condition and user's operation are not
When etc. factors, contrast image quality and the visual demand for being unsatisfactory for people.Other than the subjectivity of influence people is aesthetic, medical treatment, mesh
Mark detects and the application environments such as content description have harsh requirement to picture quality, and bad picture quality can be to production, life
It is living etc. to cause serious harm.In order to restore the details of image, scholars propose many contrast enhancement algorithms, at certain
The influence of contrast distortion is alleviated in kind degree.But this contrast distortion is inevitable, and therefore, finds one kind
The quality evaluation algorithm of contrast distortion image is an extremely challenging task.Image quality evaluation can according to Appraising subject
To be divided into subjective quality assessment and evaluating objective quality, the former is dependent on subject main body, although being capable of fundamentally quantitatively evaluating
Problem, but need to take considerable time and energy, and the latter simply and readily realizes, more mainstream.Evaluating objective quality is again
Full reference (Full Reference, FR), half can be divided into reference to (Reduced Reference, RR) and without with reference to (No
Reference, NR) three types.Existing algorithm is divided into general distortion mostly and is distorted specific two kinds, contrast distortion category
In the latter, rarely have scholar to make biggish performance boost in this field, therefore the present invention is based on maximum entropies and KL divergence, propose one
Kind is without reference contrast distorted image quality evaluating method.
Summary of the invention
In order to overcome the deficiencies of the prior art, for picture contrast problem of dtmf distortion DTMF, the present invention is directed to propose a kind of no reference
Contrast distortion image quality evaluating method, this method and subjective assessment score have higher consistency, can effectively comment
Valence contrast distortion image.For this reason, the technical scheme adopted by the present invention is that maximum entropy and KL divergence are without reference contrast distortion map
Image quality evaluation method, steps are as follows:
(1) maximum entropy calculates
An image I is given, be divided into n x n size first is not overlapped image block, under each piece of entropy passes through
Formula calculates:
Wherein, τiThe probability for indicating i-th of intensity value, after obtaining each piece of entropy, maximum entropy is indicated are as follows:
Wherein, EmIndicate maximum entropy, Z indicates the block number of every image segmentation, according to the size of each piece of entropy from greatly to
Small sequence, Ez(I) indicate numerical ranks before z image block entropy mean value；
(2) mean intensity histogram calculation
Natural scene image shows certain statistical law, in order to indicate the statistical property of a large amount of high quality graphics, chooses
DUT-OMRON data set contains the high quality natural scene image of a large amount of classifications, calculates all figures as reference, the data set
The intensity histogram of picture simultaneously takes mean value, obtains mean intensity histogram H0, as intensity distribution change reference information；
(3) change of intensity distribution is measured
A contrast distortion image is given, its intensity histogram H is calculated1, it is straight that intensity is measured using KL divergence
Square figure priori H0With test image intensity histogram H1Distance, to measure the change of intensity distribution, formula is as follows:
K=- ∫ H1(t)logH0(t)dt+∫H1(t)logH1(t)dt
=ξ (H1-H0)-E(H1) (3)
Wherein, K indicates KL divergence, ξ (H1-H0) indicate H1And H0Cross entropy, indicate are as follows:
Wherein, pjAnd qjIt is histogram H respectively1And H0J-th of bin, J is the total bin quantity of intensity histogram, be arranged J
=26, E (H1) it is intensity histogram H1Entropy；
(4) quality evaluation
Final picture quality Q is obtained to this two progress linear combinations in conjunction with KL divergence and maximum entropy:
Q=α K+ (1- α) Em (5)
Wherein, α is a constant-weight, for regulating and controlling the relative importance of KL divergence and maximum entropy two.
The features of the present invention and beneficial effect are:
The present invention proposes a kind of quality evaluating method for contrast distortion image, without necessarily referring to the intervention of image,
Contrast distortion picture quality can effectively be evaluated.
Detailed description of the invention:
Fig. 1 algorithm frame.
Fig. 2 mean intensity histogram H0。
Specific embodiment
The invention proposes a kind of no reference contrast distorted image quality evaluating methods, specifically includes the following steps:
1, maximum entropy calculates
An image I is given, be divided into 64 x, 64 size is not overlapped image block, calculates the entropy of each block of image
And sort from large to small, take the mean value of the entropy of 10% image block before numerical ordering to be denoted as maximum entropy Em。
2, mean intensity histogram calculation
DUT-OMRON data set is chosen as reference, the intensity histogram of all 5172 images is calculated and takes mean value, obtain
To mean intensity histogram H0。
3, the change of intensity distribution is measured
A contrast distortion image is given, its intensity histogram H is calculated1, intensity histogram is measured using KL divergence
Priori H0With test image intensity histogram H1Distance, be denoted as K.
4, quality evaluation
Maximum entropy Em and KL divergence K linear combination is obtained into final objective quality scores Q, in CSIQ, TID2008 and
It is verified on the contrast distortion figure of TID2013 database, the property of the method for the present invention is finally judged using four kinds of image evaluation criterion
Energy.
The present invention proposes that a kind of no reference contrast distorted image quality evaluating method, frame are as shown in Figure 1.
(1) maximum entropy calculates
An image I is given, be divided into n x n size first is not overlapped image block, and each piece of entropy can lead to
Cross following formula calculating:
Wherein, τiIndicate the probability of i-th of intensity value, n=64 is arranged in the present invention.After obtaining each piece of entropy, maximum entropy
It can indicate are as follows:
Wherein, EmIndicate maximum entropy, Z indicates the block number of every image segmentation, according to the size of each piece of entropy from greatly to
Small sequence, Ez(I) mean value of the entropy of z image block before numerical ranks, here z=η %Z, setting η=10 are indicated.
(2) mean intensity histogram calculation
Natural scene image shows certain statistical law, in order to indicate the statistical property of a large amount of high quality graphics, chooses
DUT-OMRON data set is as reference.The data set contains the high quality natural scene image of a large amount of classifications, calculates all
The intensity histogram of 5172 images simultaneously takes mean value, obtains mean intensity histogram H0, as intensity distribution change reference believe
Breath, mean intensity histogram are as shown in Figure 2.
(3) change of intensity distribution is measured
A contrast distortion image is given, its intensity histogram H is calculated1, it is straight that intensity is measured using KL divergence
Square figure priori H0With test image intensity histogram H1Distance, to measure the change of intensity distribution, formula is as follows:
K=- ∫ H1(t)logH0(t)dt+∫H1(t)logH1(t)dt
=ξ (H1-H0)-E(H1) (8)
Wherein, K indicates KL divergence, ξ (H1-H0) indicate H1And H0Cross entropy, can indicate are as follows:
Wherein, pjAnd qjIt is histogram H respectively1And H0J-th of bin, J is the total bin quantity of intensity histogram, be arranged J
=26, E (H1) it is intensity histogram H1Entropy.
(4) quality evaluation
Low picture quality, KL divergence is smaller, and maximum entropy is also small.In conjunction with KL divergence and maximum entropy, to this two into
Row linear combination obtains final picture quality Q:
Q=α K+ (1- α) Em (10)
Wherein, α is that a constant-weight is set as regulating and controlling the relative importance of KL divergence and maximum entropy two
0.5。
For verification algorithm validity, three databases comprising contrast distortion, respectively CSIQ, TID2008 are selected
And TID2013, the Database details of selection are shown in Table 1 (distortion map is the contrast distortion image chosen).
1 Database details of table
It is that Pearson is linear respectively using four kinds of Performance Evaluation criterion for the performance of checking image quality evaluation algorithm
Related coefficient (Pearson Linear Correlation Coefficient, PLCC), Spearman rank correlation coefficient
(Spearman Rank-order Correlation Coefficient, SRCC), Ken Deer rank correlation coefficient (Kendall ' s
Rank Correlation Coefficient, KRCC) and root-mean-square error (Root Mean-Squared Error, RMSE).
Wherein, PLCC, SRCC and KRCC value are bigger, and RMSE value is smaller, show that the test image of evaluation has relatively high quality.
Table 2 is performance scores of this method on three databases.
2 algorithm performance of table
As can be seen from Table 2, PLCC, the SRCC being calculated in three databases are generally higher than 0.83, KRCC
It is below 0.50 in 0.64 or more, RMSE, is illustrated related between the evaluating objective quality predicted value of this method and subjective scoring
Property it is high, show that the method for the present invention and human visual system have preferable consistency.
Claims (1)
1. a kind of maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method, characterized in that steps are as follows:
(1) maximum entropy calculates
An image I is given, be divided into nxn size first is not overlapped image block, and each piece of entropy is calculate by the following formula:
Wherein, τiThe probability for indicating i-th of intensity value, after obtaining each piece of entropy, maximum entropy is indicated are as follows:
Wherein, EmIndicate maximum entropy, Z indicates the block number of every image segmentation, arranges from big to small according to the size of each piece of entropy
Sequence, Ez(I) indicate numerical ranks before z image block entropy mean value；
(2) mean intensity histogram calculation
Natural scene image shows certain statistical law, in order to indicate the statistical property of a large amount of high quality graphics, chooses DUT-
OMRON data set contains the high quality natural scene image of a large amount of classifications, calculates all images as reference, the data set
Intensity histogram simultaneously takes mean value, obtains mean intensity histogram H0, as intensity distribution change reference information；
(3) change of intensity distribution is measured
A contrast distortion image is given, its intensity histogram H is calculated1, intensity histogram elder generation is measured using KL divergence
Test H0With test image intensity histogram H1Distance, to measure the change of intensity distribution, formula is as follows:
K=- ∫ H1(t)logH0(t)dt+∫H1(t)logH1(t)dt
=ξ (H1-H0)-E(H1) (3)
Wherein, K indicates KL divergence, ξ (H1-H0) indicate H1And H0Cross entropy, indicate are as follows:
Wherein, pjAnd qjIt is histogram H respectively1And H0J-th of bin, J is the total bin quantity of intensity histogram, be arranged J=26,
E(H1) it is intensity histogram H1Entropy；
(4) quality evaluation
Final picture quality Q is obtained to this two progress linear combinations in conjunction with KL divergence and maximum entropy:
Q=α K+ (1- α) Em (5)
Wherein, α is a constant-weight, for regulating and controlling the relative importance of KL divergence and maximum entropy two.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811411545.6A CN109584242A (en) | 2018-11-24 | 2018-11-24 | Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811411545.6A CN109584242A (en) | 2018-11-24 | 2018-11-24 | Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109584242A true CN109584242A (en) | 2019-04-05 |
Family
ID=65923885
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811411545.6A Pending CN109584242A (en) | 2018-11-24 | 2018-11-24 | Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109584242A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112419300A (en) * | 2020-12-04 | 2021-02-26 | 清华大学深圳国际研究生院 | Underwater image quality evaluation method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104282019A (en) * | 2014-09-16 | 2015-01-14 | 电子科技大学 | Blind image quality evaluation method based on natural scene statistics and perceived quality propagation |
| CN106447646A (en) * | 2016-06-28 | 2017-02-22 | 中国人民解放军陆军军官学院 | Quality blind evaluation method for unmanned aerial vehicle image |
| CN107767367A (en) * | 2017-09-26 | 2018-03-06 | 天津大学 | It is a kind of for HDR figures without reference mass method for objectively evaluating |
| CN108717694A (en) * | 2018-04-24 | 2018-10-30 | 天津大学 | Electrical impedance tomography image quality evaluating method based on fuzzy C-means clustering |
-
2018
- 2018-11-24 CN CN201811411545.6A patent/CN109584242A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104282019A (en) * | 2014-09-16 | 2015-01-14 | 电子科技大学 | Blind image quality evaluation method based on natural scene statistics and perceived quality propagation |
| CN106447646A (en) * | 2016-06-28 | 2017-02-22 | 中国人民解放军陆军军官学院 | Quality blind evaluation method for unmanned aerial vehicle image |
| CN107767367A (en) * | 2017-09-26 | 2018-03-06 | 天津大学 | It is a kind of for HDR figures without reference mass method for objectively evaluating |
| CN108717694A (en) * | 2018-04-24 | 2018-10-30 | 天津大学 | Electrical impedance tomography image quality evaluating method based on fuzzy C-means clustering |
Non-Patent Citations (1)
| Title |
|---|
| GUANGHUI YUE ET AL: "Effective and Efficient Blind Quality Evaluator for Contrast Distorted Images", 《HTTPS://IEEEXPLORE.IEEE.ORG/DOCUMENT/8467549》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112419300A (en) * | 2020-12-04 | 2021-02-26 | 清华大学深圳国际研究生院 | Underwater image quality evaluation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106780434A (en) | Underwater picture visual quality evaluation method | |
| CN107396095B (en) | A kind of no reference three-dimensional image quality evaluation method | |
| CN106462771A (en) | 3D image significance detection method | |
| CN108289221B (en) | The non-reference picture quality appraisement model and construction method of rejecting outliers | |
| CN108335289A (en) | A kind of full image method for evaluating objective quality with reference to fusion | |
| WO2016145571A1 (en) | Method for blind image quality assessment based on conditional histogram codebook | |
| CN109829905A (en) | It is a kind of face beautification perceived quality without reference evaluation method | |
| CN116843692B (en) | Regenerated active carbon state detection method based on artificial intelligence | |
| WO2023125456A1 (en) | Multi-level variational autoencoder-based hyperspectral image feature extraction method | |
| CN109584242A (en) | Maximum entropy and KL divergence are without reference contrast distorted image quality evaluating method | |
| CN109754390A (en) | A kind of non-reference picture quality appraisement method based on mixing visual signature | |
| CN111145205B (en) | Pig body temperature detection method based on infrared image under multiple pig scenes | |
| CN109523508A (en) | A kind of intensive light field quality evaluating method | |
| CN102419791A (en) | Method for estimating genetic risk of human common diseases | |
| TW201508696A (en) | Analytic system of wafer bin map and analytic method thereof | |
| CN105139373A (en) | Objective non-reference image quality evaluation method based on independent subspace analysis | |
| CN106295135B (en) | Piecewise regression catastrophe point determines shallow lake water quality management goal approach | |
| CN108022241B (en) | Consistency enhancement quality evaluation method for underwater image set | |
| Zewdie et al. | A new pooling strategy for image quality metrics: Five number summary | |
| CN108428232A (en) | A kind of blind appraisal procedure of cartoon image quality | |
| CN102769772B (en) | Method and device for evaluating video sequence distortion | |
| CN104794716B (en) | Method for objectively evaluating image quality based on image fractal characteristic | |
| Varadarajan et al. | A no-reference perceptual texture regularity metric | |
| CN108171255A (en) | Picture association intensity ratings method and device based on image identification | |
| CN113344823A (en) | Three-dimensional roughness characterization method for morphology of ablation area of silver linear contact |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190405 |
|
| WD01 | Invention patent application deemed withdrawn after publication |