CN108446292A - Subjective quality assessment method based on more distortion screenshotss images - Google Patents
Subjective quality assessment method based on more distortion screenshotss images Download PDFInfo
- Publication number
- CN108446292A CN108446292A CN201810043265.8A CN201810043265A CN108446292A CN 108446292 A CN108446292 A CN 108446292A CN 201810043265 A CN201810043265 A CN 201810043265A CN 108446292 A CN108446292 A CN 108446292A
- Authority
- CN
- China
- Prior art keywords
- distortion
- image
- subjective
- screenshotss
- distorted
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to one kind based on more distortion screenshotss image subjective picture quality evaluation methods, and steps are as follows：Original image obtains；Establish more distortion screenshotss image data bases；The main method of subjective assessment：Select twenty-twenty subject to carry out subjective quality assessment to losing screenshotss image, each subject scores one by one to each distorted image for more being distorted screenshotss image lane database more；Data processing and analysis：MOS values calculate, and remove singular value using Grubbs methods of inspection, MOS is recalculated after rejecting outlier.
Description
Technical field
The invention belongs to image processing field, specifically the subjective assessment system of screenshotss picture quality, more mistakes are involved setting up
True screenshotss image data base and to database carry out subjective picture quality evaluation method.
Background technology
User shares Web content between different digital equipment, to being based on the Internet, applications technology such as game, long-range control
System and the vision content quality of webpage etc. propose requirements at the higher level.Vision content in these applications is usually gone out with screenshotss image format
Existing, screenshotss image contains the media formats such as image and text, therefore contains more information than natural image.Although for tradition
The algorithm for having database and proposing function admirable is established in the single distortion and more distortions of natural image, and is established and be directed to
Single distortion screenshotss image data base proposes respective algorithms, but there are no the correlative studys to more being distorted screenshotss image.Mirror
Being widely present and drastically influence visual experience of the people to screenshotss image in mostly distortion screenshotss images, in order to promote snapshot
As the development of related application, the research for screen picture quality evaluation just becomes very urgent.Although screenshotss image is different from
Traditional natural image, but the algorithm values that subjective quality assessment is carried out for traditional natural image must refer to.The present invention uses for reference
The thought of subjective assessment is carried out to being distorted traditional natural image more, while considering the evaluation method of single distortion screenshotss image,
By analyzing performance of more distortions on screenshotss image, it is proposed that a kind of to be commented based on the subjective picture qualities for being distorted screenshotss images more
Valence method.
Invention content
It is an object of the invention to be directed to more distortion screenshotss picture quality evaluation problems, more distortion screenshotss image datas are established
Library simultaneously proposes the method for carrying out subjective picture quality evaluation to image.Technical solution is as follows：
One kind is based on more distortion screenshotss image subjective picture quality evaluation methods, and steps are as follows：
The first step：Original image obtains
Second step：Establish more distortion screenshotss image data bases
Third walks：The main method of subjective assessment
Select twenty-twenty subject to carry out subjective quality assessment to losing screenshotss image, each subject to being distorted screenshotss more more
Each distorted image in image data base scores one by one.
4th step：Data processing and analysis
(1) MOS values calculate：
Wherein N is subject quantity, mijIndicate subjective feeling score of i-th of subject to jth width distorted image
(2) Grubbs methods of inspection are used to remove singular value, if meeting the formula of removal singular value：
GN＞ G1- α(N) σ then judges mijFor singular value, G1- α(N) it is critical value, detects horizontal α=0.05, statistic GN
For：
GN=| mij-MOSj|
Standard deviation sigma is：
(3) MOS is recalculated after rejecting outlierjValue, standard deviation sigma and statistic GN, then judge whether outlier,
Until formula until removing singular value is invalid, gained MOS values are final subjective scores value.
In step 2, original image is added respectively tri- kinds of type of distortion of JPEG, Gblur and WN of different specified distortion levels；
Gblur and WN the mixing distortion of different specified distortion levels are added to original image respectively and the Gblur of different specified distortion levels is added
It mixes and is distorted with JPEG；JPEG, Gblur and the WN for every width original image being added respectively different specified distortion levels mix distortion, structure
Build more distortion screenshotss image data bases.
Description of the drawings：
Each figure of Fig. 1 is original image；
Fig. 2 is distorted the MOS values of screenshotss image more；
Fig. 3 MOS value scatter plots；
Specific implementation mode
The present invention proposes a kind of subjective quality assessment method based on more distortion screenshotss images.To make the technical side of the present invention
Case is clearer, is further described through below to the specific embodiment of the invention.
The first step：Establish more distortion screenshotss image data bases
Original image is obtained by intercepting size for 1280 × 720 high definition screen pictures, and 15 width screen content images include 7
The screen picture of width game screen image and 8 web texts, as shown in Figure 1.
In view of JPEG reaches 10 to compression of images:The image fault that still can only feel slightly when 1 compression factor,
Image would generally be carried out to JPEG compression processing to reduce transmission cost etc., therefore JPEG coding techniques is added in database image
To study influence of the JPEG compression in more distortions to screen content picture quality.And image is in acquisition, transmission and processing
Can usually introduce in the process in Gblur and WN, such as acquisition image process would generally introduce WN, and screenshotss image is being shot with camera
When camera shake, focus inaccurate and photographic subjects and shake etc. and can introduce Gblur.Therefore database considers that Gblur and WN is added
Screen picture is studied in distortion.
Tri- kinds of type of distortion of JPEG, Gblur and WN are added in 15 width original pattern content images.Wherein JPEG compression pair
The qualitative factor grade of original image compression is 30,20,10, and quality factor value is lower, and compression is more serious；Gblur qualitative factors
Value is respectively：0.5,1.0,1.5, value is bigger, obscures more serious；WN quality factor values are respectively：0.002,0.008,0.032,
Value is bigger, and noise is more serious.Database distorted image composition is as follows：Three specified distortion levels are added to 15 width original images respectively
Three kinds of type of distortion；Gblur and WN the mixing distortion of different specified distortion levels are added to original image respectively and are added different
Gblur and JPEG the mixing distortion of specified distortion level；Every width original image is added respectively JPEG, Gblur of different specified distortion levels
It mixes and is distorted with WN.Therefore the MDSCID databases that this experiment is established include 15 width original images and the more distorted images of 810 width.
Second step:Subjective picture quality is evaluated
This experiment collects image subjective quality scores using E-prime Experiment of Psychology softwares, is led using E-prime
The method for seeing evaluation image quality has single stimulation and double stimulus methods, this experiment is using double stimulating methods.With other subjective experiments one
Sample, this experiment need to consider environmental factor：It is carried out in the environment of no noise, to prevent interference to be tested；This experiment needs
It is carried out under normal lighting conditions, can understand viewing image to be tested；The distance of subject viewing image is 3 times of screen height.
E-prime softwares are in 3.40GHz Inter Core i7 processors, 12GB memories and Nvidia GeForce GTX560 video cards
Computer on run, display is 21.7 inches of LED liquid-crystal displays, using keying formula sound reflect collection data, picture quality
Fraction range is from 0~9, and score is higher, and expression subject is better to the subjective feeling of image.Distorted image goes out during scoring image
It is now random.
Before experiment starts, subjective picture quality evaluation rubric is introduced to each subject, entire experiment be divided into practice with just
Formula tests two parts, and two examples can be randomly selected from database in practice part gives subject practice subjective scoring, if subject
Also it is unfamiliar with evaluation rubric, can selects to return to the repetition practice of practice part, be carried out after subject is familiar with evaluation rubric formal
Experiment.To avoid subject fatigue, experiment every time only carries out 30 minutes.This experiment invites 25 subjects to more distorted screen content graphs
As carrying out subjective quality assessment.After subjective experiment, the subjective scores value of all subjects is further analysed processing
Third walks：Data processing and analysis
After the subjective scores value for collecting all subjects, the MOS values of each width distorted image are calculated：
Wherein N is subject quantity, mijIndicate subjective feeling score of i-th of subject to jth width distorted image.
Due to the influence of environment or other factors such as mood, subject may train off far from MOS values to piece image
Score value, this one kind value are singular value, and it should not includes singular value to calculate MOS values, and singular value is removed using Grubbs methods of inspection.
If meeting following formula：
GN＞ G1- α(N)·σ(3)
Then judge mijFor singular value, G1- α(N) it is critical value, detects horizontal α=0.05, statistic GNFor：
GN=| mij-MOSj|(4)
Standard deviation sigma is：
MOS is recalculated after rejecting outlierjValue, standard deviation sigma and statistic GN, recycle (3) formula judge whether from
Group's value, until (2) are invalid.Gained MOS values are final subjective scores value.
It is the MOS Distribution value situations of more distortion screenshotss images as shown in Figure 2, abscissa is MOS values, and ordinate is MOS values
Corresponding distorted image number, it can be seen that MOS distorted image numbers between 2 to 5 are relatively more, in followed by 5 to 8.5, Fig. 3
The MOS value scatter plots for depicting 5 all type of distortion of width image, using red, tri- kinds of color table distortion levels of blue, green are passed
Increase, using nine kinds of symbols (square, star, diamond, circle, mark, plus, triangle (down), triangle
(up), pentagram) indicate the combination that distortion level is different between different type of distortion.Letter contains and means in abscissa:”G”:
Gblur；”J”:JPEG；”W”:WN；”G+J”:Gblur+JPEG；”G+W”:Gblur+WN；”G+J+W”:Gblur+JPEG+WN.
As can be seen from the figure the MOS values of single distorted image are substantially 5 or more, but two or more noises mix as：
Gblur+JPEG；Gblur+WN；When Gblur+JPEG+WN, MOS is on the high side less than 5.Illustrate meeting when low distortion mixes
Subject is set to generate the serious impression of image fault.Influence of the mixing distortion to screen content image is very big known to Fig. 2 and Fig. 3.
Claims (2)
1. one kind is based on more distortion screenshotss image subjective picture quality evaluation methods, steps are as follows：
The first step：Original image obtains
Second step：Establish more distortion screenshotss image data bases
Third walks：The main method of subjective assessment
Select twenty-twenty subject to carry out subjective quality assessment to losing screenshotss image, each subject to being distorted screenshotss image more more
Each distorted image of lane database scores one by one.
4th step：Data processing and analysis
(1) MOS values calculate：
Wherein N is subject quantity, mijIndicate subjective feeling score of i-th of subject to jth width distorted image
(2) Grubbs methods of inspection are used to remove singular value, if meeting the formula of removal singular value：
GN＞ G1- α(N) σ then judges mijFor singular value, G1- α(N) it is critical value, detects horizontal α=0.05, statistic GNFor：
GN=| mij-MOSj| standard deviation sigma is：
(3) MOS is recalculated after rejecting outlierjValue, standard deviation sigma and statistic GN, then judge whether outlier, until
Remove singular value formula it is invalid until, gained MOS values be final subjective scores value.
2. according to the method described in claim 1, it is characterized in that, in step 2, different distortions are added to original image respectively
Tri- kinds of type of distortion of JPEG, Gblur and WN of grade；The Gblur and WN for original image being added respectively different specified distortion levels are mixed
It closes distortion and Gblur and JPEG the mixing distortion of different specified distortion levels is added；Different mistakes are added to every width original image respectively
JPEG, Gblur and WN mixing distortion of true grade, build and are distorted screenshotss image data base more.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810043265.8A CN108446292A (en) | 2018-01-17 | 2018-01-17 | Subjective quality assessment method based on more distortion screenshotss images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810043265.8A CN108446292A (en) | 2018-01-17 | 2018-01-17 | Subjective quality assessment method based on more distortion screenshotss images |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108446292A true CN108446292A (en) | 2018-08-24 |
Family
ID=63191040
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810043265.8A Pending CN108446292A (en) | 2018-01-17 | 2018-01-17 | Subjective quality assessment method based on more distortion screenshotss images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108446292A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255765A (en) * | 2018-09-12 | 2019-01-22 | 朱光兴 | A kind of remote sensing images color correcting method based on image type analysis |
| CN110473182A (en) * | 2019-10-08 | 2019-11-19 | 中国人民解放军61646部队 | A kind of subjective assessment processing method, device, electronic equipment and medium towards the full link simulation image of visible light |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7489807B2 (en) * | 2003-08-07 | 2009-02-10 | Kyungtae Hwang | Statistical quality assessment of fingerprints |
| US8195672B2 (en) * | 2009-01-14 | 2012-06-05 | Xerox Corporation | Searching a repository of documents using a source image as a query |
| CN106548472A (en) * | 2016-11-03 | 2017-03-29 | 天津大学 | Non-reference picture quality appraisement method based on Walsh Hadamard transform |
-
2018
- 2018-01-17 CN CN201810043265.8A patent/CN108446292A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7489807B2 (en) * | 2003-08-07 | 2009-02-10 | Kyungtae Hwang | Statistical quality assessment of fingerprints |
| US8195672B2 (en) * | 2009-01-14 | 2012-06-05 | Xerox Corporation | Searching a repository of documents using a source image as a query |
| CN106548472A (en) * | 2016-11-03 | 2017-03-29 | 天津大学 | Non-reference picture quality appraisement method based on Walsh Hadamard transform |
Non-Patent Citations (1)
| Title |
|---|
| DINESH JAYARAMAN等: "Objective quality assessment of multiply distorted images", 《2012 CONFERENCE RECORD OF THE FORTY SIXTH ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS AND COMPUTERS (ASILOMAR)》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255765A (en) * | 2018-09-12 | 2019-01-22 | 朱光兴 | A kind of remote sensing images color correcting method based on image type analysis |
| CN110473182A (en) * | 2019-10-08 | 2019-11-19 | 中国人民解放军61646部队 | A kind of subjective assessment processing method, device, electronic equipment and medium towards the full link simulation image of visible light |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bainbridge et al. | Drawings of real-world scenes during free recall reveal detailed object and spatial information in memory | |
| Ni et al. | ESIM: Edge similarity for screen content image quality assessment | |
| Schetinger et al. | Humans are easily fooled by digital images | |
| Mantiuk et al. | Comparison of four subjective methods for image quality assessment | |
| Farid et al. | Perceptual discrimination of computer generated and photographic faces | |
| CN105975980A (en) | Method of monitoring image mark quality and apparatus thereof | |
| CN107396095B (en) | A kind of no reference three-dimensional image quality evaluation method | |
| CN102881010A (en) | Method for evaluating perception sharpness of fused image based on human visual characteristics | |
| CN103745430B (en) | Rapid beautifying method of digital image | |
| CN106033535A (en) | Electronic paper marking method | |
| CN104346809A (en) | Image quality evaluation method for image quality dataset adopting high dynamic range | |
| Thompson et al. | Simulating visibility under reduced acuity and contrast sensitivity | |
| CN108446292A (en) | Subjective quality assessment method based on more distortion screenshotss images | |
| Nehmé et al. | Comparison of subjective methods, with and without explicit reference, for quality assessment of 3D graphics | |
| CN110346704A (en) | Determination method, apparatus, equipment and the storage medium of test file in board test | |
| Wang et al. | How real is reality? A perceptually motivated system for quantifying visual realism in digital images | |
| Xia et al. | Separate and simultaneous adjustment of light qualities in a real scene | |
| CN112270261A (en) | Segmentation method and device for question stem and answer mixture and storage medium | |
| Fedorovskaya et al. | Image harmony for consumer images | |
| Harrison et al. | The effect of colour and size on attentional bias to alcoho-related pictures | |
| Rizzi et al. | Visual glare limits of HDR displays in medical imaging | |
| Josephs et al. | Artifact magnification on deepfake videos increases human detection and subjective confidence | |
| CN111083468A (en) | Short video quality evaluation method and system based on image gradient | |
| Bernardo et al. | A study on the user perception to color variations | |
| Shen et al. | Robust stereoscopic crosstalk prediction |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180824 |