CN113362315A - Image quality evaluation method and evaluation model based on multi-algorithm fusion - Google Patents
Image quality evaluation method and evaluation model based on multi-algorithm fusion Download PDFInfo
- Publication number
- CN113362315A CN113362315A CN202110691115.XA CN202110691115A CN113362315A CN 113362315 A CN113362315 A CN 113362315A CN 202110691115 A CN202110691115 A CN 202110691115A CN 113362315 A CN113362315 A CN 113362315A
- Authority
- CN
- China
- Prior art keywords
- distortion
- quality evaluation
- image
- scores
- information
- 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.)
- Granted
Links
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides an image quality evaluation method based on multi-algorithm fusion, which comprises the following steps: extracting the distortion type and the image content information by using a feature extractor, and obtaining a probability vector after transformation as input and information output of other modules; calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image; carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and image content information, and outputting image content adaptive quality evaluation prediction scores at different distortion prior times; and according to the distortion type information, weighting is given to the prediction scores under different distortion priors, and the quality evaluation prediction score with distortion type adaptability is obtained and serves as a final quality evaluation score.
Description
Technical Field
The invention relates to the field of image quality evaluation, in particular to an image quality evaluation method and an evaluation model based on multi-algorithm fusion.
Background
High quality video and pictures can provide accurate and clear visual information to people. However, distortion is inevitably generated during image and video acquisition, compression, transmission and reconstruction. Therefore, in order to guarantee the visual experience of the user, an accurate quality evaluation algorithm needs to be designed to guide and optimize the image processing and video coding system.
The method for measuring user experience is called Quality Assessment (QA), and at present, two methods exist to measure user experience, the first method is subjective Quality Assessment, that is, subjective feeling of a subject watching an image/video is obtained through a subjective experiment, and a specific Quality score is given. However, the subjective experiment needs to consume a large amount of manpower and material resources, and cannot be realized in real-time application. Therefore, in practical applications, the second method, objective quality evaluation, that is, automatically predicting the quality of the current image or video by analyzing the content of the image or video using parameters or a model based on learning, is generally adopted. And objective quality evaluation can be divided into full-reference quality evaluation, half-reference quality evaluation and no-reference quality evaluation according to whether a reference image can be provided or not.
Existing full-reference quality evaluation methods can be roughly divided into five categories, namely pixel error (MSE), structural similarity (SSIM and the like), information theory (VIF), learning (LPIPS and the like) and fusion. Among them, a single full-reference quality evaluation method often cannot perform well on all distortion types and image contents. Therefore, it is necessary to design a quality evaluation method based on fusion, which can adapt to the variation of distortion types and image contents, so that the quality evaluation for various distortion types and image contents becomes reliable.
Disclosure of Invention
In view of the above, the present invention provides an image quality evaluation method and an evaluation model based on multi-algorithm fusion, so as to partially solve at least one of the above technical problems.
In order to achieve the above object, as an aspect of the present invention, there is provided an image quality evaluation method based on multi-algorithm fusion, including:
extracting the distortion type and the image content information by using a feature extractor, and obtaining a probability vector after transformation to be used as the input and the information output of other modules;
calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and image content information, and outputting image content adaptive quality evaluation prediction scores when distortion-free prior exists;
and according to the distortion type information, weighting is given to the prediction scores under different distortion priors, and the quality evaluation prediction score with distortion type adaptability is obtained and serves as a final quality evaluation score.
Wherein the probability vector is:
VD=TD(FD(ID,IR))
VC=TC(FC(IR))
wherein, VD,VCRespectively for the type of distortion and the image content information output, FD,FCRespectively distortion information and content information extraction module, TD,TCAre respectively corresponding transformations, ID,IRRespectively a distorted image and a reference image.
Calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image to further obtain a quality score vector:
VS=[SSIM(ID,IR),PSNR(ID,IR),……]。
the quality fraction vector is combined with the image content probability vector to be fused, and a distortion type prior fraction vector with content characteristics is obtained:
VSC=TSC(VC,VS);
wherein, TSCIs a fusion module.
And giving weights to the prediction scores under different distortion priors according to the distortion type information, and giving weights to the prediction scores by adopting linear weighting.
As another aspect of the present invention, there is provided an image quality evaluation model based on multi-algorithm fusion, the evaluation model including:
the distortion type and image content information extraction module is used for extracting the distortion type and the image content information by using the characteristic extractor, and obtaining a probability vector after transformation as the input and the information output of other modules;
the objective quality evaluation algorithm calculating module is used for calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
the nonlinear transformation fusion module is used for carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and the image content information and outputting image content adaptive quality evaluation prediction scores when distortion prior is not generated;
and the linear weighting module is used for giving weights to the prediction scores under different distortion priors according to the distortion type information to obtain a quality evaluation prediction score with distortion type adaptability as a final quality evaluation score.
Wherein the probability vector is:
VD=TD(FD(ID,IR))
VC=TC(FC(IR))
wherein, VD,VCRespectively for the type of distortion and the image content information output, FD,FCRespectively distortion information and content information extraction module, TD,TCAre respectively corresponding transformations, ID,IRRespectively a distorted image and a reference image.
Calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image to further obtain a quality score vector:
VS=[SSIM(ID,IR),PSNR(ID,IR),……]。
the quality fraction vector is combined with the image content probability vector to be fused, and a distortion type prior fraction vector with content characteristics is obtained:
VSC=TSC(VC,VS);
wherein, TSCIs a fusion module.
And giving weights to the prediction scores under different distortion priors according to the distortion type information, and giving weights to the prediction scores by adopting linear weighting.
Based on the technical scheme, compared with the prior art, the image quality evaluation method and the evaluation model based on multi-algorithm fusion have at least one of the following beneficial effects:
(1) the invention can make adaptive adjustment aiming at different distortion types and image contents, thereby having good subjective consistency for the quality score prediction of various distortion types and image contents.
(2) The invention can provide the analysis result of the image distortion type and the image content, thereby ensuring higher interpretability and reliability of the prediction result.
Drawings
Fig. 1 is a flowchart of a specific implementation of the evaluation method according to an embodiment of the present invention.
Detailed Description
The invention provides an objective quality evaluation method and an evaluation model integrating a plurality of algorithms, which effectively solve the following key technical problems:
1) due to the difference of the acquisition flow, the encoding and decoding modes and the network environment, the types of image distortion are various, the existing fusion algorithm utilizes a weighting mode to perform fusion, and the weight is acquired in a man-made or learning mode. This approach does not allow the weights to be adaptively adjusted depending on the type and content of the image distortion.
2) For the result of image quality evaluation prediction, only prediction scores can be given directly, but the basis information of the prediction result cannot be reflected well, so that the reliability of the prediction result is greatly reduced, and therefore, it is very important to give the analysis of distortion types and image contents while giving the prediction scores.
Specifically, the invention provides a method for guiding multiple full-reference quality evaluations to perform algorithm fusion by using image content and distortion type information, and the method can output related analysis information of the distortion type and the image content while outputting a final quality prediction score, so that a model can provide reliable image quality score prediction under different situations.
The method comprises the steps of firstly calculating the prediction scores of a plurality of objective quality evaluation algorithms, and simultaneously extracting image content information and distortion type information. And fusing the objective quality prediction scores and the image content information together to obtain quality prediction scores under different distortion priors. And carrying out linear weighting on the quality prediction scores under different distortion types in a priori by using the distortion type transformation information to obtain the final quality evaluation score.
In order that the objects, technical solutions and advantages of the present invention will become more apparent, the present invention will be further described in detail with reference to the accompanying drawings in conjunction with the following specific embodiments.
As shown in fig. 1, a specific flowchart of the evaluation method provided by the present invention specifically includes:
extracting the distortion type and the image content information by using a feature extractor, and obtaining a probability vector after transformation to be used as the input and the information output of other modules;
calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and image content information, and outputting image content adaptive quality evaluation prediction scores when distortion-free prior exists;
and according to the distortion type information, weighting is given to the prediction scores under different distortion priors, and the quality evaluation prediction score with distortion type adaptability is obtained and serves as a final quality evaluation score.
The invention also provides an image quality evaluation model based on multi-algorithm fusion, and the evaluation model comprises:
the distortion type and image content information extraction module is used for extracting the distortion type and the image content information by using the characteristic extractor, and obtaining a probability vector after transformation as the input and the information output of other modules;
the objective quality evaluation algorithm calculating module is used for calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
the nonlinear transformation fusion module is used for carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and the image content information and outputting image content adaptive quality evaluation prediction scores when distortion prior is not generated;
and the linear weighting module is used for giving weights to the prediction scores under different distortion priors according to the distortion type information to obtain a quality evaluation prediction score with distortion type adaptability as a final quality evaluation score.
The model provided by the invention can predict the quality scores of given distortion graphs and reference graphs, and can output distortion type information and image content information.
For the input distorted image and the reference image, respectively extracting features by using a feature extractor and transforming to obtain probability vectors:
VD=TD(FD(ID,IR))
VC=TC(FC(IR))
wherein, FD,FCRespectively distortion information and content information extraction module, TD,TCAre respectively corresponding transformations, ID,IRRespectively a distorted image and a reference image. Simultaneously calculating a plurality of objective quality evaluation algorithms to obtain quality score vectors:
VS=[SSIM(ID,IR),PSNR(ID,IR),……]
combining the quality fraction vector with the image content probability vector to fuse to obtain a distortion type prior fraction vector with content characteristics:
VSC=TSC(VC,VS)
wherein T isSCIs a fusion module. Finally using distortion type probability vector pairWeighting the distortion type prior fraction vector to obtain a final fraction:
S=VSC*VD
the final score obtained is used as the objective quality evaluation output of the model, VDAnd VCThe output as distortion type and image content information may reflect probabilities of belonging to different distortion types and image contents.
The method can adjust the adaptability according to different distortion types and image contents, so that the method has good subjective consistency on the quality score prediction of various distortion types and image contents, and as shown in table 1, the higher the relevant index is, the better the quality score prediction is.
Table 1 comparison of the performance of the correlation coefficients of the invention on database 1
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. An image quality evaluation method based on multi-algorithm fusion comprises the following steps:
extracting the distortion type and the image content information by using a feature extractor, and obtaining a probability vector after transformation to be used as the input and the information output of other modules;
calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and image content information, and outputting image content adaptive quality evaluation prediction scores at different distortion prior times;
and according to the distortion type information, weighting is given to the prediction scores under different distortion priors, and the quality evaluation prediction score with distortion type adaptability is obtained and serves as a final quality evaluation score.
2. The evaluation method of claim 1, the probability vector being:
VD=TD(FD(ID,IR))
VC=TC(FC(IR))
wherein, VD,VCRespectively for the type of distortion and the image content information output, FD,FCRespectively distortion information and content information extraction module, TD,TCAre respectively corresponding transformations, ID,IRRespectively a distorted image and a reference image.
3. The evaluation method of claim 1, wherein the computing of the plurality of quality evaluation algorithm scores for the distorted image and the reference image further yields a quality score vector:
VS=[SSIM(ID,IR),PSNR(ID,IR),......]。
4. the evaluation method of claim 1, wherein the quality score vector is fused with an image content probability vector to obtain a distortion type prior score vector with content characteristics:
VSC=TSC(VC,VS);
wherein, TSCIs a fusion module.
5. The method of claim 1, wherein weighting prediction scores at different distortion priors according to distortion type information is performed by linear weighting.
6. An image quality evaluation model based on multi-algorithm fusion, the evaluation model comprising:
the distortion type and image content information extraction module is used for extracting the distortion type and the image content information by using the characteristic extractor, and obtaining a probability vector after transformation as the input and the information output of other modules;
the objective quality evaluation algorithm calculating module is used for calculating a plurality of quality evaluation algorithm scores of the distorted image and the reference image;
the nonlinear transformation fusion module is used for carrying out nonlinear transformation fusion on the obtained multiple objective quality evaluation algorithm scores and the image content information and outputting image content adaptive quality evaluation prediction scores when distortion prior is not generated;
and the linear weighting module is used for giving weights to the prediction scores under different distortion priors according to the distortion type information to obtain a quality evaluation prediction score with distortion type adaptability as a final quality evaluation score.
7. The evaluation model of claim 6, the probability vector being:
VD=TD(FD(ID,IR))
VC=TC(FC(IR))
wherein, VD,VCRespectively for the type of distortion and the image content information output, FD,FCRespectively distortion information and content information extraction module, TD,TCAre respectively corresponding transformations, ID,IRRespectively a distorted image and a reference image.
8. The evaluation model of claim 6, wherein the computing of the plurality of quality evaluation algorithm scores for the distorted image and the reference image further yields a quality score vector:
VS=[SSIM(ID,IR),PSNR(ID,IR),......]。
9. the evaluation model of claim 6, wherein the quality score vector is fused with the image content probability vector to obtain a distortion type prior score vector with content characteristics:
VSC=TSC(VC,VS);
wherein, TSCIs a fusion module.
10. The evaluation model of claim 6, wherein weighting prediction scores at different distortion priors according to distortion type information is weighting using linear weighting.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110691115.XA CN113362315B (en) | 2021-06-22 | 2021-06-22 | Image quality evaluation method and evaluation model based on multi-algorithm fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110691115.XA CN113362315B (en) | 2021-06-22 | 2021-06-22 | Image quality evaluation method and evaluation model based on multi-algorithm fusion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113362315A true CN113362315A (en) | 2021-09-07 |
CN113362315B CN113362315B (en) | 2022-09-30 |
Family
ID=77535659
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110691115.XA Active CN113362315B (en) | 2021-06-22 | 2021-06-22 | Image quality evaluation method and evaluation model based on multi-algorithm fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113362315B (en) |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102708567A (en) * | 2012-05-11 | 2012-10-03 | 宁波大学 | Visual perception-based three-dimensional image quality objective evaluation method |
CN106780432A (en) * | 2016-11-14 | 2017-05-31 | 浙江科技学院 | A kind of objective evaluation method for quality of stereo images based on sparse features similarity |
CN107122787A (en) * | 2017-02-14 | 2017-09-01 | 北京理工大学 | A kind of image scaling quality evaluating method of feature based fusion |
CN107371016A (en) * | 2017-07-25 | 2017-11-21 | 天津大学 | Based on asymmetric distortion without with reference to 3D stereo image quality evaluation methods |
CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
CN109255358A (en) * | 2018-08-06 | 2019-01-22 | 浙江大学 | A kind of 3D rendering quality evaluating method of view-based access control model conspicuousness and depth map |
CN109345552A (en) * | 2018-09-20 | 2019-02-15 | 天津大学 | Stereo image quality evaluation method based on region weight |
CN110310343A (en) * | 2019-05-28 | 2019-10-08 | 西安万像电子科技有限公司 | Image processing method and device |
CN110827241A (en) * | 2019-10-21 | 2020-02-21 | 国家广播电视总局广播电视规划院 | Low-brightness enhanced picture full-reference method based on color distortion and contrast enhancement |
CN112233089A (en) * | 2020-10-14 | 2021-01-15 | 西安交通大学 | No-reference stereo mixed distortion image quality evaluation method |
-
2021
- 2021-06-22 CN CN202110691115.XA patent/CN113362315B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102708567A (en) * | 2012-05-11 | 2012-10-03 | 宁波大学 | Visual perception-based three-dimensional image quality objective evaluation method |
CN106780432A (en) * | 2016-11-14 | 2017-05-31 | 浙江科技学院 | A kind of objective evaluation method for quality of stereo images based on sparse features similarity |
CN107122787A (en) * | 2017-02-14 | 2017-09-01 | 北京理工大学 | A kind of image scaling quality evaluating method of feature based fusion |
CN107371016A (en) * | 2017-07-25 | 2017-11-21 | 天津大学 | Based on asymmetric distortion without with reference to 3D stereo image quality evaluation methods |
CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
CN109255358A (en) * | 2018-08-06 | 2019-01-22 | 浙江大学 | A kind of 3D rendering quality evaluating method of view-based access control model conspicuousness and depth map |
CN109345552A (en) * | 2018-09-20 | 2019-02-15 | 天津大学 | Stereo image quality evaluation method based on region weight |
CN110310343A (en) * | 2019-05-28 | 2019-10-08 | 西安万像电子科技有限公司 | Image processing method and device |
CN110827241A (en) * | 2019-10-21 | 2020-02-21 | 国家广播电视总局广播电视规划院 | Low-brightness enhanced picture full-reference method based on color distortion and contrast enhancement |
CN112233089A (en) * | 2020-10-14 | 2021-01-15 | 西安交通大学 | No-reference stereo mixed distortion image quality evaluation method |
Non-Patent Citations (4)
Title |
---|
LINRU ZHENG等: "No-Reference Quality Assessment for Screen Content Images Based on Hybrid Region Features Fusion", 《 IEEE TRANSACTIONS ON MULTIMEDIA》 * |
RANIA HASSEN等: "Objective Quality Assessment for Multiexposure Multifocus Image Fusion", 《 IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
史丽坤: "光场图像质量评价研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
张勇等: "基于结构相似度与感兴趣区域的图像融合评价方法", 《光子学报》 * |
Also Published As
Publication number | Publication date |
---|---|
CN113362315B (en) | 2022-09-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Shen et al. | Hybrid no-reference natural image quality assessment of noisy, blurry, JPEG2000, and JPEG images | |
Mittal et al. | Blind/referenceless image spatial quality evaluator | |
Ma et al. | Reduced-reference image quality assessment in reorganized DCT domain | |
CN112954312B (en) | Non-reference video quality assessment method integrating space-time characteristics | |
Sun et al. | Deep learning based full-reference and no-reference quality assessment models for compressed ugc videos | |
Wang et al. | Novel spatio-temporal structural information based video quality metric | |
CN112637594B (en) | No-reference 3D point cloud quality assessment method based on bit stream | |
Shao et al. | No-reference view synthesis quality prediction for 3-D videos based on color–depth interactions | |
Attar et al. | Image quality assessment using edge based features | |
CN113192147B (en) | Method, system, storage medium, computer device and application for significance compression | |
Gao et al. | Vdpve: Vqa dataset for perceptual video enhancement | |
CN114125495A (en) | Video quality evaluation model training method, video quality evaluation method and device | |
CN108682005B (en) | Semi-reference 3D synthetic image quality evaluation method based on covariance matrix characteristics | |
Shukla et al. | Opinion unaware image quality assessment via adversarial convolutional variational autoencoder | |
Patil | A survey on image quality assessment techniques, challenges and databases | |
CN113362315B (en) | Image quality evaluation method and evaluation model based on multi-algorithm fusion | |
Ma et al. | Reduced reference video quality assessment based on spatial HVS mutual masking and temporal motion estimation | |
Peng et al. | Multiple levels perceptual noise backed visual information fidelity for picture quality assessment | |
Farah et al. | Full-reference and reduced-reference quality metrics based on SIFT | |
Tiotsop et al. | Full reference video quality measures improvement using neural networks | |
Li et al. | A novel spatial pooling strategy for image quality assessment | |
Boujut et al. | No-reference video quality assessment of H. 264 video streams based on semantic saliency maps | |
Tiotsop et al. | On the link between subjective score prediction and disagreement of video quality metrics | |
Zhang et al. | Quality assessment of multiply and singly distorted stereoscopic images via adaptive construction of cyclopean views | |
KR101574418B1 (en) | Jagged blocking artifacts reduction apparatus and method using hidden markov model and principal component analysis |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |