CN104112274B - Image quality evaluating method based on mixed-scale transformation - Google Patents
Image quality evaluating method based on mixed-scale transformation Download PDFInfo
- Publication number
- CN104112274B CN104112274B CN201410318484.4A CN201410318484A CN104112274B CN 104112274 B CN104112274 B CN 104112274B CN 201410318484 A CN201410318484 A CN 201410318484A CN 104112274 B CN104112274 B CN 104112274B
- Authority
- CN
- China
- Prior art keywords
- sampling
- artwork
- image
- ssim
- distortion map
- 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
Abstract
The present invention discloses a kind of image quality evaluating method based on mixed-scale, and step is：The first step carries out different degrees of down-sampling respectively using the method for down-sampling to artwork x and distortion map y, remembers that the artwork after ith down-sampling and distortion map are respectively xi, yi；Second step, to after ith down-sampling artwork and distortion map pre-process respectively, and calculate its corresponding brightness, contrast, structure components in structural similarity function；Third step, the optimal value (l in conjunction with brightness, contrast, structural similarity under particular dimensionsi,ci,si), obtain final objective quality scores.Experimental result under six image libraries confirms, compares traditional method based on scale, and the present invention can improve the precision of prediction of image quality evaluation.
Description
Technical field
What the present invention designed is a kind of image quality evaluating method, specifically a kind of image matter based on mixed-scale transformation
Measure evaluation method.
Background technology
The information of the mankind 80% is obtained by vision, especially the today's society in information prosperity, DTV, digital shadow
The flourishing method of the various media such as picture, video conference, social media so that increasingly to the high quality demand of image and video
Greatly.Image quality evaluation (IQA) is a classical research direction, it plays important in the numerous areas of Digital Image Processing
Effect, such as compression of images, the development and optimization that store and transmit etc..Generally speaking, image quality evaluation is divided into two classes：It is subjective
Evaluation and objective evaluation.The ITU-R BT.500 subjective assessments introductions proposed according to International Telecommunication Union (ITU) are it is found that pervious
Design passes through the subjective mean opinion score (MOSs) for testing and obtaining observer.Since some of subjective assessment are significant scarce
Fall into (such as inconvenience, it is time-consuming, expensive), objective IQA methods just seem necessary for automatic Prediction picture quality.All existing
In some objective indicators, most notable one group of IQA index is mean square error (MSE) and Y-PSNR (PSNR), because they
It is very convenient and have specific physical significance.
In last decade research, more and more people think that MSE/PSNR does not well sentence picture quality people
Disconnected/MOS phases shut away.Therefore, some advanced indexs as structural similarity index (SSIM), which propose one kind, to replace
Generation and with complementary method to solve the problems, such as IQA, " A universal image quality index "-" Image
quality assessment:These middle introduced of From error visibility to structural similarity "
A little methods belong to single scale type.This image quality evaluating method based on single scale generally comprises two steps：It adopts
With the image transformation of proper proportion coefficient and SSIM indexs.However, such methods do not account for image transform coefficients in SSIM
The Different Effects of three kinds of group component amount (brightness, contrast and structural similarity).Although IQA measurements above-mentioned have fairly good
Performance, but it is not difficult to find that a remarkable affecting genes " scale " is not taken into account.It is more next in order to fill up this blank
More is grown up based on multiple dimensioned IQA methods, for example, multiple dimensioned SSIM (MS-SSIM), fidelity of information criterion
(IFC), visual information fidelity (VIF) and the information content weight (IW) PSNR/SSIM.The precision of prediction of these IQA methods is abundant
Illustrate the validity of multi-scale method.
Recently, some researchs " Perceptual visual that such as W.L is proposed in Trans.Image.Processing
quality metrics:" the Self-adaptive scale transform for IQA that Asurvey ", KeGu are proposed
Metric " (being embodied in ISCAS2013), " GES:a newimage quality assessment metric based on
Energy features in Gabor transform domain " (G.Zhai is published in ISCAS2006), " LGPS:Phase
Basedimage quality assessment metric " (G.Zhai is published in SIPS2007) propose another solution ruler
The method of degree problem, and point out that the IQA methods suitably converted based on single scale can also reach satisfied result.Particularly,
PSNR/SSIM methods based on SAST models can be in the optimal scale parameter adjustment estimated according to picture size and viewing distance
Reference picture and distorted image on realize.This method is very effective, because with the increase of viewing distance, observation angle contracting
Small, observed image detail is with regard to less.
Invention content
For the defects in the prior art, the object of the present invention is to provide a kind of picture qualities based on mixed-scale transformation
Evaluation method can be used for more accurately assessing picture quality.
The present invention is achieved by the following technical solutions, and the present invention includes the following steps：
The first step carries out different degrees of down-sampling, note i-th respectively using the method for down-sampling to artwork x and distortion map y
Artwork and distortion map after secondary down-sampling are respectively xi, yi；
Second step, to after ith down-sampling artwork and distortion map be filtered respectively, and calculate it is down-sampled after
Image corresponding brightness (luminance), contrast (contrast), structure in structural similarity (SSIM) function
(structural) component.
Third step, the optimal value (l in conjunction with brightness, contrast, structural similarity under particular dimensionsi,ci,si), it obtains most
Whole objective quality scores.The correlation of objective quality scores and MOS/DMOS values is stronger, and evaluation method is more accurate.
The principle of the present invention is three group component amount (brightness, contrast, structures in structural similarity function (SSIM)
Similarity) influence to picture quality under the different transformation coefficients of image is different, therefore study respectively brightness, contrast and
Variation coefficient of the structural similarity under different size factors show that luminance component does not become substantially with the variation of size factor
Change, and contrast and structure components are with the conclusion to become smaller afterwards that usually first becomes larger that becomes larger of size factor.Finally, in conjunction with three components
The optimal value obtained under different size factors finally realizes the target for surmounting original structure similarity based method performance.
Compared with prior art, it has the advantages that：
Compared to single two time scales approach, the present invention can provide greater flexibility by the way that the variables such as observation condition are added.
The corresponding transformation coefficient of different component is different in the present invention, six image libraries (LIVE, TID2008, CSIQ, Toyama,
IVC and LIVE Multiply Distortion) under experimental result confirm, compare traditional method based on scale, propose
IQA methods can improve the precision of prediction of image quality evaluation.Calculating speed of the present invention is fast, and complexity is low, and in performance
On be significantly improved than original method.
Description of the drawings
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon：
Fig. 1 is one embodiment of the invention overview flow chart；
Fig. 2 is under different scale transformation coefficient, and brightness (figure a), contrast (figure b), structural similarity (figure c) exist respectively
Change curve on LIVE databases；
Fig. 3 is under different scale transformation coefficient, and brightness (figure a), contrast (figure b), structural similarity (figure c) exist respectively
Change curve on TID2008 databases；
Fig. 4 is under different scale transformation coefficient, and brightness (figure a), contrast (figure b), structural similarity (figure c) exist respectively
Change curve on CSIQ databases；
Fig. 5 is SSIM (figure a), SSIMz (figure b), MS-SSIM (figure c), MIS-SSIM (figure d) on LIVE databases
Scatter plot；
Fig. 6 is SSIM (figure a), SSIMz (figure b), MS-SSIM (figure c), MIS-SSIM (figure d) on TID2008 databases
Scatter plot；
Fig. 7 is SSIM (figure a), SSIMz (figure b), MS-SSIM (figure c), MIS-SSIM (figure d) on CSIQ databases
Scatter plot.
Specific implementation mode
With reference to specific embodiment, the present invention is described in detail.Following embodiment will be helpful to the technology of this field
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill of this field
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made.These belong to the present invention
Protection domain.
Embodiment：
The implementing procedure of the present embodiment is as shown in Figure 1：
The first step carries out 100 various sizes of diminutions respectively to artwork and distortion map, using imresize in matlab
Image is narrowed down to 1,1/2,1/3,1/4 by function successively ... 1/100.
Three third step, fusion components, obtain final image objective quality scores:
It is linearly related (PLCC) that Pearson is respectively adopted, Sperman sequence related coefficients (SROCC) and root-mean-square error
(RMSE) three indexs measure the degree of correlation between objective score and subjective scores, and PLCC, SROCC value are bigger, and RMSE value is got over
Small, method for objectively evaluating is more accurate.
Implementation result
According to above-mentioned steps, tri- image libraries of LIVE, TID2008, CSIQ are tested respectively.All experiments are in PC
It is realized on computer, the major parameter of the PC computers is：Central processing unitCoreTM2Duo CPU E6600@
2.40GHz, memory 3GB.Software platform：MATLAB.
Three common performance indicator Pearson are linearly related (PLCC), Sperman sort related coefficient (SROCC) and
Root-mean-square error (RMSE) is used to assessment in LIVE, TID2008, C-SIQ, IVC, Toyama and LIVE Multiply
The MIS-SSIM algorithms of SSIM in Distortion databases, SSIMz, MS-SSIM and proposition.It is worth emphasizing that experiment is only
5 kinds of type of distortion JPEG2K are used, JPEG, white noise obscures and rapid fading, or uses multiple distortion.PLCC,SROCC
It is described in detail in Table I-III with the experimental result of RMSE.
Table I SSIM, SSIMz, MS-SSIM and MIS-SSIM are in LIVE, TID2008, C-SIQ, IVC, Toyama and LIVE
PLCC values under Multiply Distortion databases (through nonlinear fitting)
Table II SSIM, SSIMz, MS-SSIM and MIS-SSIM in LIVE, TID2008, C-SIQ, IVC, Toyama and
SROCC values under LIVE Multiply Distortion databases (through nonlinear fitting)
Table III SSIM, SSIMz, MS-SSIM and MIS-SSIM in LIVE, TID2008, C-SIQ, IVC, Toyama and
RMSE value under LIVE Multiply Distortion databases (through nonlinear fitting)
Fig. 2 is l, and curvilinear motion of tri- components of c, s on the libraries LIVE, Fig. 3 is l, and tri- components of c, s are in the libraries TID2008
On curvilinear motion, Fig. 4 is l, curvilinear motion of tri- components of c, s on the libraries CSIQ, and Fig. 5 is DMOS values with respect to SSIM,
The scatter plot of SSIMz, MS-SSIM, MIS-SSIM on the libraries LIVE, Fig. 6 are MOS values with respect to SSIM, SSIMz, MS-SSIM,
Scatter plots of the MIS-SSIM on the libraries TID2008, Fig. 7 are DMOS with respect to SSIM, and SSIMz, MS-SSIM, MIS-SSIM is in CSIQ
Scatter plot on library.
Compared with prior art, the present invention has in image evaluation effect and is significantly improved.It is further noted that invention carries
The MIS-SSIM gone out belongs to generalized model, this means that it can be used to correct other IQA indexs, such as some optimizations
SSIM versions.In addition, the present invention can expand the method for proposition, this can only bring small change on SSIM.For
Video quality evaluation VQA, due to its simplification, SSIM has been implanted most of video coding systems and as picture quality
The benchmark of index.Therefore the present invention is based on the VQA methods of MIS-SSIM to have good performance.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited in above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring the substantive content of the present invention.
Claims (1)
1. a kind of image quality evaluating method based on mixed-scale, feature include the following steps：
The first step carries out different degrees of down-sampling respectively using the method for down-sampling to artwork x and distortion map y, remembers under ith
Artwork and distortion map after sampling are respectively xi, yi；The down-sampling is to carry out multiple difference respectively to artwork and distortion map
The diminution of size；
Second step, to after ith down-sampling artwork and distortion map be filtered respectively, and it is similar in structure to calculate it
Spend corresponding brightness, contrast, structure components in function；
Third step, the optimal value (l in conjunction with brightness, contrast, structural similarity under selected scalei,ci,si), it obtains final
The correlation of objective quality scores, objective quality scores and MOS/DMOS values is stronger, and evaluation method is more accurate；
In second step：Calculate separately the value of luminance component l, contrast component c, structure components s under different images transform size：
Wherein C1=(K1L)2,C2=(K2L)2,C3=C2/ 2, with 1.5 sampling values and normalizing to (Σ wi=1) height
This weighting windows is filtered image, w={ wi| i=1,2 ..., N }；Local statistics average value mux, standard deviation sigmaxAnd mutually
Related σxyIt is as follows：
In third step：The structural similarity Function exponential of assessment whole image quality is defined as：
Wherein, x (i) and y (i) is the picture material in i-th of local window, is x and y reference pictures and distorted image, M is image
In local window number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410318484.4A CN104112274B (en) | 2014-07-04 | 2014-07-04 | Image quality evaluating method based on mixed-scale transformation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410318484.4A CN104112274B (en) | 2014-07-04 | 2014-07-04 | Image quality evaluating method based on mixed-scale transformation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104112274A CN104112274A (en) | 2014-10-22 |
| CN104112274B true CN104112274B (en) | 2018-08-03 |
Family
ID=51709053
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410318484.4A Active CN104112274B (en) | 2014-07-04 | 2014-07-04 | Image quality evaluating method based on mixed-scale transformation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104112274B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107316323B (en) * | 2017-06-28 | 2020-09-25 | 北京工业大学 | No-reference image quality evaluation method established based on multi-scale analysis method |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101621709B (en) * | 2009-08-10 | 2011-01-05 | 浙江大学 | Method for evaluating objective quality of full-reference image |
| CN102845071B (en) * | 2010-04-19 | 2015-01-28 | 杜比实验室特许公司 | Quality assessment of high dynamic range, visual dynamic range and wide color gamut image and video |
| CN101853504B (en) * | 2010-05-07 | 2012-04-25 | 厦门大学 | Image quality evaluating method based on visual character and structural similarity (SSIM) |
| CN102421007B (en) * | 2011-11-28 | 2013-09-04 | 浙江大学 | Image quality evaluating method based on multi-scale structure similarity weighted aggregate |
| CN103544708B (en) * | 2013-10-31 | 2017-02-22 | 南京邮电大学 | Image quality objective evaluation method based on MMTD |
-
2014
- 2014-07-04 CN CN201410318484.4A patent/CN104112274B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104112274A (en) | 2014-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104899845B (en) | A kind of more exposure image fusion methods based on the migration of l α β spatial scenes | |
| CN101562675B (en) | No-reference image quality evaluation method based on Contourlet transform | |
| CN105445607B (en) | A kind of electrical equipment fault detection method drawn based on thermoisopleth | |
| CN109472817B (en) | Multi-sequence magnetic resonance image registration method based on loop generation countermeasure network | |
| CN108010024A (en) | It is a kind of blind with reference to tone mapping graph image quality evaluation method | |
| CN104123705B (en) | A kind of super-resolution rebuilding picture quality Contourlet territory evaluation methodology | |
| CN111127360B (en) | Gray image transfer learning method based on automatic encoder | |
| CN109242834A (en) | It is a kind of based on convolutional neural networks without reference stereo image quality evaluation method | |
| CN108830823A (en) | The full-reference image quality evaluating method of frequency-domain analysis is combined based on airspace | |
| CN110443800A (en) | The evaluation method of video image quality | |
| CN110298829A (en) | A kind of lingual diagnosis method, apparatus, system, computer equipment and storage medium | |
| US9087272B2 (en) | Optical match character classification | |
| Morzelona | Human visual system quality assessment in the images using the IQA model integrated with automated machine learning model | |
| Jin et al. | Perceptual Gradient Similarity Deviation for Full Reference Image Quality Assessment. | |
| Feng et al. | Low-light image enhancement algorithm based on an atmospheric physical model | |
| CN106683074B (en) | A kind of distorted image detection method based on haze characteristic | |
| CN104112274B (en) | Image quality evaluating method based on mixed-scale transformation | |
| Fu et al. | Screen content image quality assessment using Euclidean distance | |
| CN102609891B (en) | Texture-characteristic-based method for passively and blindly obtaining evidence of digital image | |
| Yang et al. | EHNQ: Subjective and objective quality evaluation of enhanced night-time images | |
| CN107578406A (en) | Based on grid with Wei pool statistical property without with reference to stereo image quality evaluation method | |
| CN106056575B (en) | A kind of image matching method based on like physical property proposed algorithm | |
| CN110415816B (en) | Skin disease clinical image multi-classification method based on transfer learning | |
| CN108022241B (en) | Consistency enhancement quality evaluation method for underwater image set | |
| CN112132774A (en) | Quality evaluation method of tone mapping image |
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 |