CN105005990B - A kind of blind image quality evaluating method based on distinctiveness rarefaction representation - Google Patents
A kind of blind image quality evaluating method based on distinctiveness rarefaction representation Download PDFInfo
- Publication number
- CN105005990B CN105005990B CN201510381379.XA CN201510381379A CN105005990B CN 105005990 B CN105005990 B CN 105005990B CN 201510381379 A CN201510381379 A CN 201510381379A CN 105005990 B CN105005990 B CN 105005990B
- Authority
- CN
- China
- Prior art keywords
- dictionary
- mrow
- sub
- noise
- image
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of blind image quality evaluating method based on distinctiveness rarefaction representation, including：First, training obtains the sub- dictionary of feature and the sub- dictionary of noise from the natural image and noise image sample of noise-less pollution respectively, and two sub- dictionaries are merged into distinctiveness dictionary；Then, image to be evaluated is represented with the distinctiveness dictionary, obtains sparse coefficient corresponding to two sub- dictionaries；Finally, final picture quality scoring is obtained by counting the ratio of two parts sparse coefficient.Compared to other existing blind quality evaluating methods, the inventive method is high with human eye evaluation result degree of fitting, and realizes simply, is trained without artificial marking sample.
Description
Technical field
The present invention relates to a kind of blind image quality evaluating method, belong to image processing techniques, particularly perceive visual signal
Treatment technology.
Background technology
The problem of image quality evaluation is a basis in image processing field, in compression of images, recovery, reconstruction, increasing
By force, important application is suffered from the field such as identification and classification.Image quality evaluation can be divided into subjective assessment and objective evaluation
Two kinds.Subjective quality assessment is usually to ask one group of expert to image progress visual evaluation and carry out subjective marking.In many applications
In field, image is finally all read by people see, in this case, subjective quality assessment be it is unique accurately and reliably
Quality evaluation mode.But in a practical situation, subjective assessment is expensive due to time-consuming, it is inconvenient the shortcomings of be difficult to be applied.
Compared to subjective quality assessment, evaluating objective quality has speed fast, and cost is low, simple operation and other advantages,
As the emphasis in image quality evaluation research field.Evaluating objective quality can be divided into two kinds, and one kind needs undistorted ginseng
Examine image and carry out quality evaluation (reference image quality appraisement), another kind only needs the information of distorted image itself to carry out matter
Amount evaluation (blind quality evaluation).Reference image quality appraisement implements also phase by the development comparative maturity of decades
To simple.Comparing classical reference image quality appraisement method has PSNR, SSIM etc..But this kind of method is required to access without mistake
Genuine artwork, this requirement can not be met in many application environments.
It compared to reference image quality appraisement, can utilize, therefore be faced with more without reference to image in blind quality evaluation
Challenge.Existing blind quality evaluating method is mainly made up of two steps：Feature extraction and the model time based on human eye scoring
Return.The image that existing blind quality evaluating method usually requires a large amount of handmarkings is used to train, and the performance of algorithm is also to training
Sample is very sensitive；In addition, this kind of algorithm is typically necessary higher amount of calculation and complicated parameter Estimation and training, these are all
It significantly limit the application of blind quality evaluating method in practice.
The content of the invention
The technical problem to be solved by the invention is to provide one kind to realize simply, is trained without handmarking's image
And the high blind image quality evaluating method with human eye evaluation degree of fitting.
The present invention is solution above-mentioned technical problem the technical scheme adopted is that a kind of based on the blind of distinctiveness rarefaction representation
Image quality evaluating method, comprise the following steps：
(1) training obtains the sub- dictionary of feature and noise from the natural image and noise image sample of noise-less pollution respectively
Sub- dictionary, and two sub- dictionaries are merged into distinctiveness dictionary；
(2) image to be evaluated is represented with the distinctiveness dictionary, obtains sparse coefficient corresponding to two sub- dictionaries；
(3) final picture quality scoring is obtained by counting the ratio of two parts sparse coefficient.
In the step (1), specifically include：
(1.1) segment is extracted from the natural image of noise-less pollution as features training collection, is extracted from noise image
Segment is as noise training set；
(1.2) from two training sets of step (1.1), training obtains the sub- dictionary D of feature+With the sub- dictionary D of noise-, construction
Distinctiveness dictionary DdIt is characterized sub- dictionary D+With the sub- dictionary D of noise-Set.
In the step (2), specifically include：
(2.1) subgraph by picture breakdown to be evaluated for three Color Channels of red, green and blue；
(2.2) to the subgraph of each Color Channel, subgraph is split into and the segment size identical in training set
Segment, with distinctiveness dictionary DdSubgraph is represented, obtains sparse coefficient corresponding to each subgraph, the sparse coefficient includes spy
Levy sub- dictionary D+Represent the sparse coefficient and the sub- dictionary D of noise corresponding to image-Represent the sparse coefficient corresponding to image.
In the step (3), specifically include：
(3.1) sparse coefficient corresponding to the sub- dictionary of noise and the sub- word of feature in the subgraph by counting each Color Channel
The weight ratio of sparse coefficient corresponding to allusion quotation, obtain the scoring of corresponding subgraph；
(3.2) weighted sum scored according to the subgraph of three Color Channels, final picture quality scoring R is obtainedd,
RdIt is bigger, then it is assumed that picture quality is better.
With distinctiveness dictionary D in the step (2.2)dSubgraph is represented, obtains sparse coefficient corresponding to each subgraph
Specific method be：Equation below is solved with orthogonal matching pursuit algorithm, obtains sparse coefficient α：
Wherein, R is the operator that segment is extracted from subgraph y to be evaluated, and subscript i, j are the segment upper left corner in subgraph
Position coordinates, | | αij||0It is zero norm, limits αijThe number of middle nonzero element is no more than L；The sparse coefficient tried to achieve is by two
It is grouped into α=[α+, α-], wherein α+It is characterized sub- dictionary D+Represent the sparse coefficient corresponding to image, α-For the sub- dictionary D of noise-
Represent the sparse coefficient corresponding to image.
The scoring of the subgraph of each Color Channel in the step (3.1)Calculation formula be：
Wherein N is the segment number that subgraph decomposites,WithRespectively the sub- dictionary of c Color Channels feature and make an uproar
Phonon dictionary represents the sparse coefficient of k-th of segment, | | | |1For a norm, wkFor weight coefficient corresponding to k-th of segment.wk
It is defined as wk=σk+ M, wherein M are a constant, σkIt is the standard deviation of k-th of segment.
R in the step (3.2)dObtained according to the weighted sum that the subgraph of three Color Channels scores, each subgraph
Coefficient is converted into the conversion formula coefficient of Y passages in YIQ spaces with reference to rgb space as corresponding to scoring.
Beneficial effect：The blind image quality evaluating method based on distinctiveness rarefaction representation of the present invention, it is entitled to be based on area
Characteristic quantification (the Feature Quantification via Discriminative Sparse of other property rarefaction representation
Representation, FQ-DSR).The it is proposed of this method receives primary human vision cortex when carrying out picture quality to figure
As feature carries out the inspiration of rarefaction representation processing, it is proposed that carry out quantitative analysis to characteristics of image and carry out the side of quality evaluation
Method.Wherein, characteristics of image both includes normal picture structure, also the degenerative structure (noise) comprising image.It is therefore, of the invention
A kind of distinctiveness dictionary is pointedly devised, " just " the sub- dictionary of suitable expression normal picture structure had both been included in the dictionary
(the sub- dictionary of feature), and include " negative " sub- dictionary (the sub- dictionary of noise) of suitable expression picture noise.When with this distinctiveness word
When allusion quotation carries out rarefaction representation to image, it is possible to by being used for representing to scheme in the sub- dictionary of quantitative statisticses feature and the sub- dictionary of noise
The coefficient distribution of the atom of picture carrys out the quality of evaluation image quality.Compared to the blind quality image quality evaluation side of existing representativeness
Method, the inventive method realize that simply the image without handmarking is trained, and more consistent with manually giving a mark.
Brief description of the drawings
Fig. 1 is the flow chart of the inventive method.
Fig. 2 (a)-(g) is 7 undistorted natural images used in the sub- dictionary of training characteristics.
Fig. 2 (h) is the simulation Gaussian noise image used in the training sub- dictionary of noise.
Fig. 3 (a) is the sub- dictionary of feature trained using FDDL methods.
Fig. 3 (b) is the sub- dictionary of noise trained using FDDL methods.
Embodiment
With reference to specific embodiment, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention
Rather than limitation the scope of the present invention, after the present invention has been read, various equivalences of the those skilled in the art to the present invention
The modification of form falls within the application appended claims limited range.
As shown in figure 1, the blind image quality evaluating method master based on distinctiveness rarefaction representation disclosed in the embodiment of the present invention
To be made up of three steps：Distinctiveness dictionary is constructed, dictionary represents and comprehensive marking.First, from (no mistake of noise-less pollution
Very) dictionary corresponding to each self-training in natural image and noise image sample, and two sub- dictionaries are merged into distinctiveness dictionary.
Then, remove to represent image to be evaluated with the distinctiveness dictionary, obtain sparse coefficient corresponding to two sub- dictionaries.Finally, pass through
The ratio for counting two parts sparse coefficient obtains final image marking.Comprise the following steps that：
Step 1: construction distinctiveness dictionary, is comprised the following steps that：
1st step, substantial amounts of segment is extracted as features training collection from the natural image of some noise-less pollutions；Separately from mould
A large amount of segments are extracted in the Gaussian noise image of plan as noise training set.Analogue noise image can select Gaussian noise to carry out
Simulation, can also use other kinds of noise.Image to be evaluated is polluted by Gaussian noise in this experiment, therefore in order to obtain
Best evaluation effect, herein select corresponding to Gaussian noise analog image training noise dictionary.Such as Fig. 2 is selected in this experiment
(a) 7 undistorted natural images shown in-(g) and the simulation Gaussian noise image as shown in Fig. 2 (h) are as sample, wherein scheming
The size of block is 16 × 16.In concrete application scene can reasonable selection sample image and number of samples as the case may be,
And the size of segment.
In 2nd step, two training sets obtained from the 1st step, instructed using Fisher distinctiveness dictionary learning methods (FDDL)
Practise two sub- dictionaries that there is specific characteristic to represent ability：The sub- dictionary D of feature+With the sub- dictionary D of noise-.Wherein every sub- dictionary
Size is 256 × 512 (see Fig. 3).Purpose from FDDL methods is in order that the sub- dictionary of the feature trained is to natural image
Architectural feature have more preferable expression ability, have poor expression ability to noise；And the sub- dictionary of noise trained is then to making an uproar
Sound has more preferable expression ability, has poor expression ability to the architectural feature of image.
3rd step, construction distinctiveness dictionary DdIt is characterized the set of sub- dictionary and the sub- dictionary of noise
Dd=[D+,D-]。DdSize be 256 × 1024.
Step 2: dictionary represents, comprise the following steps that：
1st step, by picture breakdown it is red, the subgraph y of green and blue three passagesr, yg, yb。
2nd step, with distinctiveness dictionary DdRemove to represent the subgraph y to be evaluated of each passagec(c can be r, g or b, divide
Three passages of red, green, blue are not represented).Specific method is to solve following problem with orthogonal matching pursuit algorithm (OMP)：
Obtain sparse coefficient α.Wherein R is the operator that segment is extracted from subgraph y to be evaluated, and segment size is similarly 16
×16.Subscript i, j are the position coordinates of the segment upper left corner in the picture.Between segment at intervals of 16, i.e., segment and segment it
Between both do not overlapped, it is also very close to each other.||αij||0It is zero norm, limits αijThe number of middle nonzero element is no more than L, and L is set to
25.The sparse coefficient tried to achieve forms α=[α by two parts+, α-], wherein α+It is characterized sub- dictionary D+Represent dilute corresponding to image
Sparse coefficient, α-For the sub- dictionary D of noise-Represent the sparse coefficient corresponding to image.
Step 3: comprehensive marking, is comprised the following steps that：
1st step, the weight ratio by counting sparse coefficient corresponding to the sub- dictionary of feature in subgraph, define a certain color
Passage subgraph scoresFor
N is the segment number that subgraph to be evaluated decomposites in formula；WithThe respectively sub- word of c Color Channels feature
Allusion quotation and the sub- dictionary of noise represent the sparse coefficient of k-th of segment, | | | |1For a norm, wkFor weight corresponding to k-th of segment
Coefficient.wkIt is defined as wk=σk+ M, wherein M are a constant, can be adjusted according to concrete application scene, be set as 20 in this experiment
Preferably test and appraisal effect can be obtained；σkIt is the standard deviation of k-th of segment.
2nd step, define the weighted scoring that final image quality score is three Color Channel scorings：
Wherein,WithThe respectively scoring of the subgraph of three Color Channels of red, green and blue.Each subchannel
Coefficient corresponding to scoring is converted into Y in YIQ spaces with reference to rgb space in NTSC (American National Television System Committee) standard and led to
The conversion formula coefficient in road (lightness passage)
(Y=0.2989R+0.5870G+0.1140B).
Work as RdWhen smaller, the part proportion that the sub- dictionary of feature represents in representative image is larger, and what the sub- dictionary of noise represented
Part proportion is smaller, it is taken as that picture quality is better；On the contrary, work as RdWhen larger, then it is assumed that image is relatively tight by noise pollution
Weight, picture quality are poorer.Weight wkIt is set as the standard deviation sigma of a certain segmentkWith a constant M's and, reflect making an uproar in the segment
Sound intensity.When a certain segment noise is larger, a norm of the sub- dictionary coefficient of noise corresponding to the segmentWith weight wkAll
It is larger, proportion of the segment noise coefficient in integrally scoring is further enhancing after multiplication；When a certain segment noise is smaller,
One norm of the sub- dictionary coefficient of noise corresponding to the segmentWith weight wkIt is all smaller, the figure is further reduced after multiplication
Proportion of the block noise coefficient in integrally scoring.
In this part, by FQ-DSR methods proposed by the present invention compared with other technologies carry out Experimental comparison.It is real
Other the blind image quality evaluating methods being related in testing have a BIQI, LBIQ, DIIVINE, BLIINDS-II, BRISQUE and
CORNIA.In addition, also FQ-DSR methods are contrasted with the reference picture evaluation method of some main flows in experiment, the side being related to
Method includes SSIM, PSNR, IFC and VIF.
Above several method is tested on the image quality evaluation database LIVE IQA of main flow respectively in experiment.
LIVE IQA databases include 5 kinds of image degradation types：JPEG2000, JPEG, white Gaussian noise and quickly decline at Gaussian Blur
Fall.Only consider Gaussian noise in this experiment, Gaussian noise corresponding part includes 29 undistorted reference pictures and 145 in database
Open the image of different degrees of noise pollution.Each image has corresponding human eye subjective scoring (DMOS) in database.In experiment,
Calculating marking is carried out to image with method mentioned above first, the scoring calculated and DMOS scorings are sought into the degree of correlation to comment afterwards
The quality of valency algorithm.It is Spearman coefficient of rank correlations (SROCC) to seek the criterion used in the degree of correlation, and the coefficient is closer to 1
Represent algorithm appraisal result and human eye scoring is closer.
It is trained because the blind image quality evaluation algorithm of some main flows needs to use artificial scorable marker image, therefore
In corresponding document, every group of test image is divided into two parts by author.Wherein 80% part is used to train, afterwards again another
Tested on 20% image.Method proposed by the present invention is trained without handmarking's image, only need to be random in several width
Natural image on training dictionary.But in line with the principle of justice contrast, in every group of test, we also only choose 20%
Image (different degrees of image polluted by noise corresponding to 6 width reference pictures and 30 width) is tested, and ignores other 80% images.
Experiment includes 1000 test sets, and each test set randomly selects 6 width reference pictures, 30 amplitude and noise sound pollution figure corresponding with its
Picture.Experiment is all given a mark with algorithm to every group of test set and seeks SROCC coefficients, 1000 obtained after repeating 1000 times
The median of SROCC scorings is as final evaluation result.Experimental result is as shown in table 1.
Table 1
As can be seen from the table, FQ-DSR scores proposed by the present invention have not exceeded only the blind image quality evaluation side of main flow
Method, it have also exceeded the reference image quality appraisement method of main flow.Test result indicates that method proposed by the present invention is made an uproar for Gauss
The image of sound pollution, the quality evaluation score being sufficiently close to human eye evaluation can be obtained in the case of without reference to image.
Claims (6)
1. a kind of blind image quality evaluating method based on distinctiveness rarefaction representation, it is characterised in that comprise the following steps：
(1) training obtains the sub- dictionary of feature and the sub- word of noise from the natural image and noise image sample of noise-less pollution respectively
Allusion quotation, and two sub- dictionaries are merged into distinctiveness dictionary；Including：
(1.1) segment is extracted from the natural image of noise-less pollution as features training collection, segment is extracted from noise image
As noise training set；
(1.2) from two training sets of step (1.1), training obtains the sub- dictionary D of feature+With the sub- dictionary D of noise-, construction difference
Property dictionary DdIt is characterized sub- dictionary D+With the sub- dictionary D of noise-Set；
(2) image to be evaluated is represented with the distinctiveness dictionary, obtains sparse coefficient corresponding to two sub- dictionaries；Including：
(2.1) subgraph by picture breakdown to be evaluated for three Color Channels of red, green and blue；
(2.2) to the subgraph of each Color Channel, by subgraph split into the segment size identical segment in training set,
With distinctiveness dictionary DdSubgraph is represented, obtains sparse coefficient corresponding to each subgraph, the sparse coefficient includes feature
Dictionary D+Represent the sparse coefficient and the sub- dictionary D of noise corresponding to image-Represent the sparse coefficient corresponding to image；
(3) final picture quality scoring is obtained by counting the ratio of two parts sparse coefficient；Including：
(3.1) sparse coefficient corresponding to the sub- dictionary of noise and the sub- dictionary pair of feature in the subgraph by counting each Color Channel
The weight ratio for the sparse coefficient answered, obtain the scoring of corresponding subgraph；
(3.2) weighted sum scored according to the subgraph of three Color Channels, coefficient reference corresponding to each subgraph scoring
Rgb space is converted into the conversion formula coefficient of Y passages in YIQ spaces, obtains final picture quality scoring Rd, RdIt is smaller, then recognize
It is better for picture quality.
2. the blind image quality evaluating method according to claim 1 based on distinctiveness rarefaction representation, it is characterised in that institute
State in step (2.2) with distinctiveness dictionary DdSubgraph is represented, obtains the specific method of sparse coefficient corresponding to each subgraph
For：Equation below is solved with orthogonal matching pursuit algorithm, obtains sparse coefficient α：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mi>&alpha;</mi>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>R</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>D</mi>
<mi>d</mi>
</msub>
<msub>
<mi>&alpha;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>&alpha;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>0</mn>
</msub>
<mo>&le;</mo>
<mi>L</mi>
<mo>&ForAll;</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, R is the operator that segment is extracted from subgraph y to be evaluated, and subscript i, j are position of the segment upper left corner in subgraph
Coordinate is put, | | αij||0It is zero norm, limits αijThe number of middle nonzero element is no more than L；The sparse coefficient tried to achieve is by two parts group
Into α=[α+, α-], wherein α+It is characterized sub- dictionary D+Represent the sparse coefficient corresponding to image, α-For the sub- dictionary D of noise-Represent
Sparse coefficient corresponding to image.
3. the blind image quality evaluating method according to claim 1 based on distinctiveness rarefaction representation, it is characterised in that institute
State the scoring of the subgraph of each Color Channel in step (3.1)Calculation formula be：
<mrow>
<msubsup>
<mi>R</mi>
<mi>d</mi>
<mi>c</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<msubsup>
<mi>&alpha;</mi>
<mi>k</mi>
<mrow>
<mi>c</mi>
<mo>-</mo>
</mrow>
</msubsup>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>1</mn>
</msub>
<mo>&times;</mo>
<msub>
<mi>w</mi>
<mi>k</mi>
</msub>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<msubsup>
<mi>&alpha;</mi>
<mi>k</mi>
<mrow>
<mi>c</mi>
<mo>+</mo>
</mrow>
</msubsup>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
</mrow>
Wherein N is the segment number that subgraph decomposites,WithThe respectively sub- dictionary of c Color Channels feature and the sub- word of noise
Allusion quotation represents the sparse coefficient of k-th of segment, | | | |1For a norm, wkFor weight coefficient corresponding to k-th of segment.
4. the blind image quality evaluating method according to claim 3 based on distinctiveness rarefaction representation, it is characterised in that wk
It is defined as wk=σk+ M, wherein M are a constant, σkIt is the standard deviation of k-th of segment.
5. the blind image quality evaluating method according to claim 1 based on distinctiveness rarefaction representation, it is characterised in that institute
It is simulation Gaussian noise image to state the noise image in step (1).
6. the blind image quality evaluating method according to claim 1 based on distinctiveness rarefaction representation, it is characterised in that institute
Step (1) is stated to train to obtain the sub- dictionary of feature and the sub- dictionary of noise using Fisher distinctiveness dictionary learning methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510381379.XA CN105005990B (en) | 2015-07-02 | 2015-07-02 | A kind of blind image quality evaluating method based on distinctiveness rarefaction representation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510381379.XA CN105005990B (en) | 2015-07-02 | 2015-07-02 | A kind of blind image quality evaluating method based on distinctiveness rarefaction representation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105005990A CN105005990A (en) | 2015-10-28 |
| CN105005990B true CN105005990B (en) | 2017-11-28 |
Family
ID=54378647
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510381379.XA Active CN105005990B (en) | 2015-07-02 | 2015-07-02 | A kind of blind image quality evaluating method based on distinctiveness rarefaction representation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105005990B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997585A (en) * | 2016-01-22 | 2017-08-01 | 同方威视技术股份有限公司 | Imaging system and image quality evaluating method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102945552A (en) * | 2012-10-22 | 2013-02-27 | 西安电子科技大学 | No-reference image quality evaluation method based on sparse representation in natural scene statistics |
| CN103473745A (en) * | 2013-09-16 | 2013-12-25 | 东南大学 | Low-dosage CT image processing method based on distinctive dictionaries |
| CN104376565A (en) * | 2014-11-26 | 2015-02-25 | 西安电子科技大学 | Non-reference image quality evaluation method based on discrete cosine transform and sparse representation |
-
2015
- 2015-07-02 CN CN201510381379.XA patent/CN105005990B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102945552A (en) * | 2012-10-22 | 2013-02-27 | 西安电子科技大学 | No-reference image quality evaluation method based on sparse representation in natural scene statistics |
| CN103473745A (en) * | 2013-09-16 | 2013-12-25 | 东南大学 | Low-dosage CT image processing method based on distinctive dictionaries |
| CN104376565A (en) * | 2014-11-26 | 2015-02-25 | 西安电子科技大学 | Non-reference image quality evaluation method based on discrete cosine transform and sparse representation |
Non-Patent Citations (2)
| Title |
|---|
| Sparse Representation-based Image Quality Assessment;Tanaya Guha 等;《Signal Processing:Image Communication》;20141130;第29卷(第10期);1138-1148 * |
| 无参考图像质量评价综述;王志明;《自动化学报》;20150615;第41卷(第6期);1062-1079 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105005990A (en) | 2015-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103996192B (en) | Non-reference image quality evaluation method based on high-quality natural image statistical magnitude model | |
| CN105447884B (en) | A kind of method for objectively evaluating image quality based on manifold characteristic similarity | |
| CN106778676B (en) | Attention assessment method based on face recognition and image processing | |
| CN104134204B (en) | Image definition evaluation method and image definition evaluation device based on sparse representation | |
| CN104751456B (en) | Blind image quality evaluating method based on conditional histograms code book | |
| CN108052980B (en) | Image-based air quality grade detection method | |
| CN103325122B (en) | Based on the pedestrian retrieval method of Bidirectional sort | |
| CN101976444B (en) | Pixel type based objective assessment method of image quality by utilizing structural similarity | |
| CN108289222A (en) | A kind of non-reference picture quality appraisement method mapping dictionary learning based on structural similarity | |
| CN106447646A (en) | Quality blind evaluation method for unmanned aerial vehicle image | |
| CN104021545A (en) | Full-reference color image quality evaluation method based on visual saliency | |
| CN104361574A (en) | No-reference color image quality assessment method on basis of sparse representation | |
| CN103745466A (en) | Image quality evaluation method based on independent component analysis | |
| CN107066972A (en) | Natural scene Method for text detection based on multichannel extremal region | |
| CN103745231B (en) | Teleutospore image identification method for Tillctia Controversa Kahn (TCK) and allied variety TCT (Tilletia caries (DC.) Tul.) of TCK | |
| CN106023214B (en) | Image quality evaluating method and system based on central fovea view gradient-structure similitude | |
| CN105894507B (en) | Image quality evaluating method based on amount of image information natural scene statistical nature | |
| CN109886945A (en) | Based on contrast enhancing without reference contrast distorted image quality evaluating method | |
| CN106169174A (en) | A kind of image magnification method | |
| CN109741285A (en) | A kind of construction method and system of underwater picture data set | |
| CN107767367A (en) | It is a kind of for HDR figures without reference mass method for objectively evaluating | |
| CN105005990B (en) | A kind of blind image quality evaluating method based on distinctiveness rarefaction representation | |
| CN107018410A (en) | A kind of non-reference picture quality appraisement method based on pre- attention mechanism and spatial dependence | |
| CN108010023B (en) | High dynamic range image quality evaluation method based on tensor domain curvature analysis | |
| CN104835172A (en) | No-reference image quality evaluation method based on phase consistency and frequency domain entropy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210528 Address after: 555 Yeshan Road, Yuyao City, Ningbo City, Zhejiang Province Patentee after: XINGAOYI MEDICAL EQUIPMENT Co.,Ltd. Address before: 210096 No. four archway, 2, Jiangsu, Nanjing Patentee before: SOUTHEAST University |