CN109714591A - Based on the picture quality subjective evaluation method and system to assessment label - Google Patents
Based on the picture quality subjective evaluation method and system to assessment label Download PDFInfo
- Publication number
- CN109714591A CN109714591A CN201910020772.4A CN201910020772A CN109714591A CN 109714591 A CN109714591 A CN 109714591A CN 201910020772 A CN201910020772 A CN 201910020772A CN 109714591 A CN109714591 A CN 109714591A
- Authority
- CN
- China
- Prior art keywords
- image
- assessment
- label
- quality
- subjective
- 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
Abstract
The present invention is a kind of picture quality subjective evaluation method based on to assessment label, belongs to image quality evaluation and art of image analysis.Comprise determining that evaluation personnel, standards of grading and viewing condition, prescreening are preferentially tested and assessed sequence；For selected preferential cycle tests, by the measures of prescreening image data composition image pair, the relative mass that structure observation person opens image to image pair two judges.The invention also discloses evaluation systems, comprising: user management module, image Pre-screening module, sequence playing module, data processing module.The present invention is by constituting preferred image quality test sequence, by the measures of image data composition image pair, picture quality subjective evaluation method use scope is expanded, suitable for evaluating similar underwater picture etc. there are the image of COMPLEX MIXED distortion, improves the stability and reliability of evaluation result.
Description
Technical field
The invention belongs to image quality evaluations and art of image analysis, especially for the figure being distorted with COMPLEX MIXED
Picture is related to based on the picture quality subjective evaluation method to assessment label, and the invention further relates to evaluated using preceding method
Evaluation system.
Background technique
In many applications that the mankind are visual information ultimate consumers, how digital picture quality is accurately and effectively assessed
It is particularly important.Vision is a very important foundation in scientific research of seas.Various underwater monitoring platforms for many years, ROV/
AUV, submarine observation system and Shui Xia observation tower, fishing boat collected number with the underwater picture of billion grade, video and higher-dimension ultraphotic
Spectrogram picture.In water, water body optical attenuator, scattering and light source illuminate so that the underwater picture of shooting there are low contrasts, fuzzy, non-
The various complicated and mixing such as Uniform Illumination, speck, color projection and various noises are degenerated, and ocean scene investigation and oceanographic observation are non-
It is often complicated and expensive, therefore underwater picture quality evaluating method can be automatic screening high quality graphic, be underwater image restoration
Or enhancing processing, it improves Underwater Imaging and transmission system design provides objective standard, Marine Sciences artificial intelligence analysis, ocean are appointed
The decision system and prediction etc. of being engaged in provide important value.Image quality evaluation includes subjective assessment and objectively evaluates.In image matter
It measures in evaluation study, since human eye is the final receptor of image, subjective assessment is considered as most directly, most accurately characterizing
The method of visual perception.The subjective quality of image is to measure the performance for objectively evaluating algorithm, image classification screening and further divide
The basis of analysis.The complicated image subjective quality assessment method degenerated is mixed suitable for existing currently without a kind of.
Currently used subjective picture quality evaluation method is mainly for the natural image quality evaluation data having built up
Designed by library.The subjective scoring method of image quality evaluation can be divided mainly into three classes: first is that single advocate approach, second is that double excitation method,
The third is Paired Comparisons.
Single motivational techniques evaluation and test needs to play image to be evaluated over the display, the ginseng without knowing original image
Examine information.Compare typically single excitation continuous mass evaluation method in single motivational techniques.Single excitation continuous mass evaluation method
Material to be evaluated is played by random order, observer beats the image seen using Pyatyi continuous mass marking system
Point.
Generally there are two types of embodiments for double excitation method.One is double excitation distortion measure method (DSIS), this method exists
Reference material and corresponding test material are playd in order when test, and appraiser's marking is prompted when playing test material,
Reference material and test material recycling can also be played one time, require appraiser to score when the second wheel plays,
The marking mode made frequently with 5 points.Another kind is double excitation continuous mass two time scales approach (DSCQS), and this method is also successively broadcast
Which is two materials put test material and reference material, but appraiser can't clearly be told to be played in an experiment
Reference material, which is test material.Therefore, it is necessary to all score two broadcasting materials.This kind of method is generally to test
Material and reference material loop play twice, require appraiser to give a mark when playing second and recycling.It is lost by calculating
Point poor quality to measure test material between vacuum material and reference material.
Pairs of comparative approach by being compared marking to two distortion levels from same reference picture, comment by realization
Estimate the performance difference between different product or algorithms of different.Scoring is seven class by this kind of method: excessively poor, poor, somewhat poor,
The two is identical, somewhat good, fine, very good.Paired Comparisons is mainly used for evaluating and testing different systems, algorithms of different and difference
Influence of the processing parameter to same content material.
Light transmits in water, and the absorption and scattering that water body internal optics attribute (IOP) determines affect entire Underwater Imaging
Effect.Forward scattering makes point light source be diffused as blur circle, fuzzy so as to cause image；Back scattering makes the contrast of image
It reduces, generates misty fuzzy superposition on the image.It absorbs and scattering is not only to be generated by water body itself, further include that dissolution is organic
Object and floating granules influence, and planktonic organism, the concentration of colored dissolved organic matter and total suspended matter matter and target range are at being also shadow
The principal element of color image quality under Xiangshui County.In addition, the absorption of the color projection of submarine target and water body to Different lightwave length
It is related with decaying.With the increase of depth under water, color successively disappears according to wavelength, and blue passes under water since wavelength is most short
Broadcast apart from longest.Although visual range can be increased by increasing artificial light, non-uniform lighting situation often will lead to,
Speck is generated in the picture, and it is very dark around speck.And artificial light source makes scattering caused by suspended material more serious.
The caused spray, whirlpool, silt and the influence of various marine organisms also result in the irregular fuzzy of image when motion work.
In addition to this, imaging system, light source color temperature will all have an impact the quality of underwater color image.Therefore, the underwater figure of shooting
As having the following problems mostly: limited visual range, low contrast, non-uniform lighting, fuzzy, hot spot, color projection and respectively
The noise of kind complicated factor.
Underwater picture is different from natural image, can be used as the original clear underwater picture of reference due to that can not obtain,
So traditional double excitation method is not suitable for underwater picture, single advocate approach scores often not for the underwater picture of similar mass
The nuance that subjective picture quality impression can be represented, for enhance or restore obtain during serious degraded image a bit
The raising of point mass is difficult to judge that a kind of method is more preferable than another result, and for real-time and automatic processing, this is
It is most important.Paired Comparisons is mainly used for evaluating and testing different systems, algorithms of different and different disposal parameter to same content
The influence of material.
Summary of the invention
It is a kind of based on to assessment label the technical problem to be solved by the present invention is in view of the deficiencies of the prior art, propose
Picture quality subjective evaluation method, this method have the more superior performance for differentiating subtle difference in quality, and wide adaptability can be used
In the image of evaluation COMPLEX MIXED distortion.
Another technical problem to be solved by this invention realizes the above-mentioned image based on to assessment label there is provided a kind of
The system of quality subjective evaluation method.
The technical problem to be solved by the present invention is to what is realized by technical solution below.The present invention is that one kind is based on
It, will be to mapping using the prescreening method measured based on color image quality to the picture quality subjective evaluation method of assessment label
As the measures of data composition image pair, relative mass is marked in structure observation person, is generated to assessment label, to label
It is handled to obtain subjective picture quality；Its step are as follows:
Step 1, determines evaluation personnel, standards of grading and observation condition, and prescreening is preferentially tested and assessed sequence；
(1) evaluation personnel should have (correction is extremely) normal visual acuity and normal color vision；It can not be and be engaged in figure
The expert of shape iconology；Specific number is determined according to preferential cycle tests set sizes, generally requires the survey of every observer
(including check and demonstrate) is commented the time to be no more than 30 minutes.
(2) standards of grading are determined
Two picture qualities height of testing image centering is marked in evaluation personnel, provides marking to assessment for image pair
Label.For image to (I1,I2), such as I1It is better than I in subjective quality2, then corresponding to assessment label l1,2It is set as+1, subscript
1,2 is expressed as image 1, image 2, while l2,1=-1；If I2Subjective quality is better than I1, assignment l1,2=-1, l2,1=+
1, in addition, can not be marked to assessment label, image is to (I at this time when picture quality is not easily distinguishable1,I2) to assessment label
It is recorded as l1,2And l2,1It is 0.
(3) viewing condition is determined
The subjective assessment environment of arrangement, to obtain most believable data.The test environment of subjective experiment: apart from screen 0.55
- 0.65 meter of rice；Maximum viewing angle < 30 °；Testing image cannot be blocked on display screen.
(4) prescreening based on color image quality measurement
Assuming that total N width image in testing image set, then produce N (N-1)/2 possible image pair.With each image
3 to 5 seconds are needed for meter to marking, the test phase of half an hour removes inspection, training, demonstration etc. before test, a test rank
Section about 300 groups of images pair of general observable, therefore determine to select 44850 groups of images pair, i.e. 300 images in advance, it generates preferential
Test image set.
To avoid the testing image Mass Distribution in preferential cycle tests set from excessively concentrating, guarantee that testing image quality exists
Uniform sampling in existing range.The present invention can choose color image luminance contrast, tone variance, saturation degree mean value these three
Selection criteria of the index as preferential cycle tests.For the N width image in testing image set, three kinds of every image are calculated
Index histogram (divides ten minizones), randomly selects 10 images on different sections.
Step 2: it for selected preferential cycle tests, is tested using subjective assessment system, evaluation personnel is to be measured
Image pair two picture qualities height be marked, provide image pair to assessment label.
Step 3: the personal information and image for saving evaluation personnel are to assessment label evaluation result data, according to assessment
Label is ranked up all images.And label score and hundred-mark system score are calculated, calculation method is as follows:
(1) image tag score is calculated
For image i, by by image i and other images j all images pair obtained to assessment label li,j,i≠
J accumulation calculating goes out image i label score Si:
According to all images pair to assessment label, the N respective label scores of image, the mark of every image can be generated
Label score all falls within the section [- N+1, N-1].Assuming that present image set covers all possible picture quality ranges of test,
From preferably to worst.Total label and N-1 correspond to best quality value, the corresponding worst mass value of-N+1.
(2) it calculates image and corresponds to hundred-mark system score
The hundred-mark system quality score S of image i is calculated according to Linear Mappingip:
Finally, the subjective scores of all picture qualities in test image data acquisition system are obtained, the higher table of subjective quality score
Show that the picture quality is better.
A kind of picture quality subjective evaluation method based on to assessment label of the present invention, further preferred skill
Art scheme is: in step 2, using the test carried out to assessment label image quality subjective evaluation system as unit of image pair,
It is carried out by the sequence generated at random；After each pair of one group of image of evaluation personnel is to quality status stamp is made, automatically switch next group of image
To continuing to test；The specific method is as follows:
(1) typing evaluation personnel information；
(2) enter demonstration and introduce interface；
(3) it is loaded into cycle tests；
(4) image to by etc. be simultaneously displayed on screen in a manner of sizes；
(5) the quality height that evaluation personnel carries out image pair marks, if can not judge the relative mass of two images
Quality should then select judge button and carry out next group of test；
(6) all images are completed to give a mark to evaluation.
A kind of picture quality subjective evaluation method based on to assessment label of the present invention, further preferred skill
Art scheme is: the present invention is based on pairs of specified in the picture quality subjective evaluation method and ITU-R BT.500 to assessment label
Comparison method is different, and difference is: the picture quality subjective assessment mode based on to assessment label allows testing image pair
With different picture materials, it is not necessary to distinguish type of distortion and level, and can have different sizes.
A kind of picture quality subjective evaluation method based on to assessment label of the present invention, further preferred skill
Art scheme is: the present invention is based on to assessment label picture quality subjective assessment mode and ITU-R BT.500 specified in pairs
Comparison method is different, and difference is: the picture quality subjective assessment mode based on to assessment label is using three tier structure quality
Label label, does not require to mark to the similar image of quality, and calculates image matter using all pairs of assessment labels obtained
The method measured point.
The invention also discloses a kind of systems based on the picture quality subjective assessment to assessment label, its main feature is that, it should
System may be implemented image collection prescreening, user information registration, image by human-computer interaction and remember to play mode selection, scoring
Record and operation, the system such as to calculate to assessment label include: user management module, image Pre-screening module, sequence playing module,
Data processing module；Wherein:
User management module, including user information addition and user information delete two submodules, for testing to user
The management such as addition, deletion of information；
The image quality evaluations such as image Pre-screening module, including luminance contrast, tone variance, saturation degree mean value calculate son
Module and a decimation blocks are respectively used to realize the image quality evaluation value for calculating every image in testing image set, and
Sample mode is set using decimation blocks, generates preferential cycle tests；
Sequence playing module is mainly used for the broadcasting of image pair when subjective assessment, different play mode may be selected, and defaults
For shuffle；
Data management module, including evaluation marking, data save and label calculates three submodules, for realizing scoring number
According to record, preservation and result statistical analysis.
The method of the present invention is suitable for the image of the presence mixing distortion under various complex environments, such as underwater picture.It is also suitable
In natural image, with the performance for differentiating subtle difference in quality more superior than other methods.It is proposed by the invention to assessment
The unlimited imaged content of label image quality subjective evaluation method is used between different content, using three tier structure to assessment label
Scoring, and the similar image of quality is not required to mark.By calculate image collection in all images to assessment label pair
Picture quality is ranked up, and applicability is wider.
The present invention is used to assessment label image quality subjective evaluation method, so that not needing during subjective assessment original
Image makes reference, and for carrying out relative mass judgement to the different content image that is distorted there are COMPLEX MIXED, obtains between image
Opposite subjective picture quality.
The method of the present invention is not limited by picture material, and the resolution that can avoid image degenerated form and degree is difficult and subjective
To the sensibility of picture material, viewing condition, resolution energy of the obtained picture quality score to subtle difference in quality in evaluation
Power is more superior, and being more suitable for evaluation, there are the images that COMPLEX MIXED is distorted.
The method of the present invention ensure that scoring tool between observer's group without being grouped according to the type of degeneration to test image
There is consistency.Using in the present invention to assessment label image quality subjective evaluation method, it can be achieved that picture quality subjective assessment
Database extends immediately, without continuous iteration new version with extending database.As the continuous expansion of image data base can be gradually
Realize image fault type and horizontal being uniformly distributed in extensive range in database, avoid deterioration level be unevenly distributed and
The too narrow problem of range.Tissue of the present invention is simple, requires observer low, need to only make opposite judgement without considering marking, accidentally
Difference is small, and practical application is high.
Picture quality subjective evaluation method of the present invention based on to assessment label, the think of that incremental learning is sorted
Wesy has the non-reference picture quality subjective evaluation Data-Statistics distribution that can not separate mixing distortion in foundation.It is marked using to assessment
Label subjective evaluation method is ranked up marking to picture quality, solves in the imaging circumstances such as underwater picture without reference to image
Problem.Compared with traditional subjective quality assessment method, subtle difference in quality in terms of be substantially better than
Other subjective picture quality evaluation methods.Meet human vision using establishing to have to assessment label image quality subjective evaluation method
The sets of image data of perceived quality score, the available image are located at the relative scores in present image library, realize to figure
Image quality amount is evaluated without reference.
Compared with prior art, beneficial effects of the present invention are summarized as follows:
(1) present invention is in addition to it can be used for the applicable all scenes of traditional subjective evaluation method.
More it is directed to the image for the presence mixing distortion under underwater picture and other various complex environments.
(2) present invention does not need original image and makes reference.
(3) present invention compared with existing Primary Subjective image quality evaluating method, can more accurate resolution have
The nuance of the underwater picture of similar mass.
(4) present invention can avoid right in the resolution of the degenerated forms such as underwater picture and degree difficulty and subjective assessment well
Picture material, the sensibility for watching condition；
(5) present invention ensure that observation without being grouped according to the type of degeneration to test image in assessment process
It scores between member's group with uniformity；
(6) present invention observer in assessment process, which can choose image similar in perceptual quality, does not give a mark, to observation
Member requires low, and error is small.
Detailed description of the invention
The flow chart of Fig. 1 subjective evaluation method；
Fig. 2 is using luminance contrast as the histogram (dividing ten minizones) of index；
Fig. 3 is using tone variance as the histogram (dividing ten minizones) of index；
Fig. 4 is using saturation degree mean value as the histogram (dividing ten minizones) of index；
Fig. 5 is using UCIQE as the histogram (dividing ten minizones) of index；
Fig. 6 is the score of two kinds of method for subjective testing；
Fig. 7-10 is corresponding to the image of the identical score of single to assessment stamp methods score；
Figure 11 is 24 color colour atlas of tank experiments shooting；
Figure 12-14 is the 24 color colour atla pictures obtained captured by tank experiments；
Figure 15 is the distribution scatter plot of all image scores in group one；
Figure 16 is the distribution scatter plot of all image scores in group two；
Figure 17 is the distribution scatter plot of all image scores in group three；
Figure 18 is the distribution scatter plot of all image scores in group four；
Figure 19 is that two width in two images of group peel off point image.
Specific embodiment
Technical solution of the present invention is described further below, is made it is further understood that this hair
It is bright, without constituting the limitation to right of the present invention.
Embodiment 1, a kind of picture quality subjective evaluation method based on to assessment label, which, which uses, is based on cromogram
The prescreening method of image quality metric is by the measures of testing image data composition image pair, and structure observation person is to relative mass
Judged, generates to assessment label, label is handled to obtain subjective picture quality；Its step are as follows:
Step 1, determines evaluation personnel, standards of grading and observation condition, and prescreening is preferentially tested and assessed sequence.
Assuming that total N underwater pictures in given testing image set, then produce N (N-1)/2 possible image pair.
Assessment process follows the requirement in the subjective testing recommendation of International Telecommunication Union's publication, to avoid one test of observer's fatigue
Stage design is half an hour.3 to 5 seconds are needed in terms of to marking by each image, before the test phase of half an hour removes test
Check, train, demonstrate etc., about 300 groups of images pair of a general observable of test phase can select 44850 groups of images pair in advance,
That is 300 images, are preferentially tested.
In image Pre-screening module, CIELAB spatial brightness contrast, tone variance, saturation degree mean value are as pre-selection
Standard, calculates the index of correlation of all images in image collection, and the histogram of each index and comprehensive UCIQE (are divided into 10 areas
Between) as shown in Figure 2-5.Randomly choose about 10, each section image.
45 evaluation personnels are chosen, it is most of to be all from Huaihai Institute of Technology Institute Of Electrical Engineering, for image procossing and figure
As quality evaluation has certain understanding.In subjective assessment system log-in interface, the name of evaluation personnel, gender, age, profession are carried on the back
Scape etc. is counted, to be used for subsequent research.Above step is carried out in user management module.
Step 2: for selected preferential cycle tests, using the picture quality subjective assessment system to assessment label into
Row test.
300 underwater picture quality are marked, image group { I is constructed1,...,In, N=300.Every piece image with
Image in addition to oneself establishes image pair, generates to image collection P, the size of P is N (N-1)/2=44850 group image pair.
(1) introduction of observer's read interface program influences the principle of image quality factors, the evaluation criterion of comparative evaluation, comments
Valence process time etc..Software interface will show that several groups of typical differences are degenerated, the image of different level is to set, and demonstration was evaluated
Journey.Before formally starting assessment, 5 or so " analog demenstration " is played, first to stablize the scoring of observer.It is given in this several groups of demonstrations
The calculating that test result is not involved in assessment label out.
(2) in the laboratory environment of subjective testing standard for meeting International Telecommunication Union's publication, sequence playing module will
Two images of image pair are simultaneously displayed on display.Executable pretreatment unified image size.It is broadcast using random sequence
Image pair is put, every group of image is to requiring observer to carry out superiority and inferiority judgement in not more than 3s, and the scoring time of every observer is not
More than half an hour.
Two picture qualities height of testing image centering is marked in evaluation personnel, provides marking to assessment for image pair
Label.For image to (I1,I2), if subjective image I1Quality is better than I2, then to assessment label l1,2It is set as+1, subscript 1,2
It is expressed as image 1, image 2, while l2,1=-1；If subjective image I2Quality is better than I1, assignment l1,2=-1, l2,1=
+ 1, in addition, can not be marked to assessment label, image is to (I at this time when picture quality is not easily distinguishable1,I2) to assessment label
It is recorded as l1,2And l2,1It is 0.
Step 3: the personal information and image for saving evaluation personnel are to assessment label result data, according to assessment label
All images are ranked up.And label score and hundred-mark system score are calculated, calculation method is as follows:
For image i, by by image i and other images j all images pair obtained to assessment label li,j,i≠
J accumulation calculating goes out image i label score Si:
According to all images pair to assessment label, the respective label score S of 300 images can be generatedi, every image
Label score all fall within [- 299,299] section, present image set covers all possible picture quality ranges of test,
From preferably to worst.Total label and 299 corresponding best quality values, -299 corresponding worst mass values.
The hundred-mark system quality score S of image i is calculated according to Linear Mappingip:
Finally, the subjective scores of all picture qualities in underwater picture data acquisition system are obtained, the higher table of subjective quality score
Show that the picture quality is better.
Embodiment 2, to assessment label image quality subjective evaluation method and single stimulating image quality subjective evaluation method
Comparative experiments.
Since underwater picture does not have original image, double excitation method is obviously not suitable for, and Paired Comparisons be to from
Two distorting objects of same raw material are compared marking, are not inconsistent with the purpose of the present invention.Comparison other is determined as by we
Single excitation consecutive image quality evaluating method (SSCQE).The selection of laboratory environment built with observer is kept and to assessment
The setting of label image quality subjective evaluation method is consistent, with single excitation subjective picture quality evaluation assessment to 300 underwater figures
As carrying out 5 grades of system scoring tests.The step of training before starting and " analog demenstration " are introduced all referring to the present invention is tested, with
Ensure the interference compared there is no other factors two methods.300 underwater pictures are divided into 6 groups, and every group 50 is opened, in batches
It is tested, to prevent observer tired.Image, which plays, uses random sequence.It is required that observer is in the not more than time of 3s or so
It gives a mark to each image according to 1-5 points of systems.To the average image quality subjectivities of assessment stamp methods and single stimulating method at
The comparison of achievement (Mean Opinion Sore, MOS) value is as shown in Fig. 6.
By the visible picture quality subjective assessment to assessment label image quality subjectivity evaluation and test method and single motivational techniques of attached drawing 6
Value has intimate identical change curve, and the degree of correlation has reached 0.95, still, such as four images of attached drawing 7-10, pierced by single
Swash assessment, the quality score of four images is 4 points, although we still can be seen that the quality difference between them.To survey
The score that assessment of bids label picture quality subjective evaluation method obtains embodies the ability that these nuances are distinguished in scoring.
Embodiment 5 is discussed by analyzing the quality of 24 color colour atla (21.59 × 27.94cm) images shot in pond
The accuracy of picture quality score is obtained to assessment label image quality subjective evaluation method.
Pond is 2.53 meters long where image taking, 1.02 meters wide, 1.03 meters high, and photographic subjects are 24 color color of Ai Seli standard
Card, as shown in Fig. 11.These images are shot with OTI-UWC-325/P/E color camera.In 94.5cm transparency water body and
Underwater picture (960 × 576) are obtained under the conditions of natural illumination, captured photo is increased and degree of degeneration increasing with camera distance
The underwater picture added.
Imatest is a image evaluating software being widely used, and contains SFR, Colorcheck, Stepchart
Etc. modules.To 24 color colour atla image (see Figure 12, Figure 13, Figure 14) quality under three captured different distances in the present embodiment
Subjective testing and software test score data are as shown in table 1.With the increasing of distance under identical shooting condition difference shooting distance
Long picture quality constantly declines, more to label score/hundred-mark system score for obtaining of assessment label image quality subjective evaluation method
This point accurately showed, and there have the two apparent images of width mass difference to obtain in the test result of single to be identical
Score illustrates to stimulate 5 grades of systems marking that can more reflect picture quality than single assessment label image quality subjective evaluation method
Slight change.
The CIE1976L*a*b* colour space (the CIE LAB colour space) is to be recommended by international lighting association (CIE) for 1976
Uniform colour space.The space is three-dimensional cartesian coordinate system system.It is the chromatic measuring system being widely used at present.It is sat with lightness L* and coloration
A*, b* are marked to indicate position of the color in the colour space.In table 1, Meancamera chroma (saturation) is camera
The average chrominance of color divided by ideal Colorchecker color average chrominance, as a percentage.Generally arrived 100%
Between 120%.
Meancamera chroma (saturation)=100% × mean ((a*2+b*2)1/2)/mean((a*ideal2
+b*ideal2)1/2) (4)
This numerical value is meant that the difference of the performance color and colour atla standard color of 24 kinds of colors on test target image,
Image appearance color and original color difference are bigger, this numerical value is bigger.ΔC*abuncorr,ΔC*abchroma corr
With Δ E*ab be in device-independent CIElab color space color error measure, by measure the Europe between them it is several in
Moral distance illustrates the difference between perceived color.Δ C*abuncorr and Δ C*abchroma corr only calculate color,
Wherein:
Δ C*abuncorr=(Δ a*)2+(Δb*)2)1/2 (5)
Chroma corr refers to before being compared, and the mean chroma of camera is adjusted to identical as reference value,
It indicates the accuracy of color if mean chroma is identical as reference value.
Δ E*ab=((Δ L*)2+(Δa*)2+(Δb*)2)1/2 (6)
By will be compared to the subjective scores of assessment label image quality subjective evaluation method with software output data,
It can be seen that the real quality of test image and the MOS value to assessment label image quality subjective evaluation are in a linear relationship.This card
The accuracy of proposed invention method is illustrated.
The MOS and software of 1 24 color colour atla image of table export score
。
Embodiment 3 is upchecked to assessment the obtained degeneration image sequence of label image quality subjective evaluation method mode
Quality score accuracy.
Experimental image is the degeneration image sequence shot according to same region with angle difference turbidity, and similar image is in this reality
It applies in example and shares four groups according to different contents of shooting one in 300 images, every group of score is plotted in Figure 15-18 respectively.
Picture number is bigger, indicates that the turbidity of water when image taking is lower.By being obtained to assessment label image quality subjective evaluation method
PLCC, SROCC and KROCC between the MOS obtained and picture number are listed in table 2.Due to being automatic sampling selection, Mei Gexu
Amount of images in column is different.
Correlation between 2 MOS value of table and picture number
The result shows that is proposed linearly closes the result of assessment label image quality subjective evaluation and the turbidity of water
System, the real quality for accurately reflecting image are horizontal.There are apparent exceptional value in group shown in Figure 16 and 17, two are organized
KROCC value is minimum, and Figure 19 is highlighted two outlier images in group two.As shown in figure 19, between two adjacent images
Difference be non-norm paste.And the image selected in group two is more concentrated in Mass Distribution.Group four is such as schemed closest to linearly
Shown in 18.Infer: (1) image of group four is more richer than the image color of group two, the picture quality that (2) randomly select in group four
Distribution gradient becomes apparent, and there is no the consecutive images with similar image quality.
Claims (6)
1. a kind of picture quality subjective evaluation method based on to assessment label, it is characterised in that: the invention is used based on colour
The prescreening method of image quality evaluation is by the measures of testing image data composition image pair, and structure observation person is to opposite matter
Amount is marked, and generates to assessment label, is handled label to obtain subjective picture quality；The specific steps of which are as follows:
Step 1, determines evaluation personnel, standards of grading and observation condition, and prescreening is preferentially tested and assessed sequence；
(1) evaluation personnel should have normal or correction to normal visual acuity and normal color vision；Should not Seeking Truth be engaged in
The expert of graph image；Specific number is determined according to preferential cycle tests set sizes, generally requires every observer's
The time test and assess including checking and demonstrating no more than 30 minutes；
(2) standards of grading are determined
Two picture qualities height of testing image centering is marked in evaluation personnel, provides the image of two images composition
Pair to assessment label；For image to (I1, I2), work as I1It is better than I in subjective quality2When, then it is corresponding to assessment label
l1,2It is set as+1, subscript 1,2 is expressed as image 1, image 2, while l2,1=-1；Work as I2Subjective quality is better than I1When, it assigns
Value l1,2=-1, l2,1=+1；In addition, when picture quality is not easily distinguishable, can not mark to assessment label, at this time image pair
(I1, I2) to assessment label record be l1,2And l2,1It is equal to 0；
(3) viewing condition is determined
The subjective assessment environment of arrangement, to obtain most believable data；The test environment of subjective experiment: apart from 0.55 meter of screen-
0.65 meter；Maximum viewing angle < 30 °；Testing image is not blocked on display screen；
(4) prescreening based on color image quality measurement
Assuming that total N width image in testing image set, then produce N (N-1)/2 possible image pair；With each image air exercise
Divide and need 3 to 5 seconds for meter, the test phase of half an hour removes inspection, training, demonstration before testing, and a test phase is general
About 300 groups of images pair of observable, therefore determine to select 44850 groups of images pair, i.e. 300 images in advance, generate preferential test chart
Image set closes；
Step 2: it for selected preferential cycle tests, is tested using to assessment label subjective assessment system, evaluation personnel
Two picture qualities of testing image centering height is marked, provide image pair to assessment label；
Step 3: the personal information and image for saving evaluation personnel are to assessment label label result data, according to assessment label
All images are ranked up；And label score and hundred-mark system score are calculated, calculation method is as follows:
(1) image tag score is calculated
For image i, by by image i and other images j all images pair obtained to assessment label li,j, i ≠ j is tired
Add and calculates image i label score Si:
According to all images pair to assessment label, the N respective label scores of image are generated, the label score of every image is equal
Fall in the section [- N+1, N-1]；Assuming that present image set covers the possible picture quality ranges of all tests, from preferably to
It is worst；Total label and N-1 correspond to best quality value, the corresponding worst mass value of-N+1；
(2) it calculates image and corresponds to hundred-mark system score
The hundred-mark system quality score S of image i is calculated according to Linear Mappingip:
Finally, obtaining the subjective quality score of all images in test image data acquisition system, the higher expression of subjective quality score should
Picture quality is better.
2. a kind of picture quality subjective evaluation method based on to assessment label according to claim 1, it is characterised in that:
In step 2, using the test carried out to assessment label image quality subjective evaluation system as unit of image pair, by random raw
At sequence carry out；After each pair of one group of image of evaluation personnel is to quality status stamp is made, automatically switch next group of image to continuing to survey
Examination；The specific method is as follows:
(1) typing evaluation personnel information；
(2) enter demonstration and introduce interface；
(3) it is loaded into cycle tests；
(4) image to by etc. be simultaneously displayed on screen in a manner of sizes；
(5) the quality height that evaluation personnel carries out image pair marks, if can not judge the quality of the relative mass of two images,
It should then select judge button and carry out next group of test；
(6) all images are completed to give a mark to evaluation.
3. a kind of picture quality subjective evaluation method based on to assessment label according to claim 1 or 2, feature exist
In: allow testing image to different picture materials based on the picture quality subjective assessment mode to assessment label, it is not necessary to
Type of distortion and level are distinguished, and can have different sizes.
4. a kind of picture quality subjective evaluation method based on to assessment label according to claim 1 or 2, feature exist
In: in step 1, color image luminance contrast, tone variance, these three indexs of saturation degree mean value are chosen as preferential test
The selection criteria of sequence；For the N width image in testing image set, three kinds of index histograms of every image are calculated, three kinds
Index histogram divides ten minizones, randomly selects 10 images on different sections.
5. a kind of picture quality subjective evaluation method based on to assessment label according to claim 1 or 2, feature exist
In: it is evaluated based on the picture quality subjective assessment mode to assessment label using three tier structure quality tab, figure similar to quality
Picture does not require to evaluate, and the method for calculating picture quality score using all pairs obtained assessment labels.
6. a kind of system based on the picture quality subjective assessment to assessment label, it is characterised in that: system passes through human-computer interaction
Realize that image collection prescreening, user information registration, image selects play mode, scoring records and calculates assessment label
Operation；System includes: user management module, image Pre-screening module, sequence playing module and data processing module；Wherein:
User management module: two submodules are deleted including user information addition and user information, for testing to user information
The management such as addition, deletion；
Image Pre-screening module: the image quality evaluation including luminance contrast, tone variance, saturation degree mean value calculates submodule
Block and a decimation blocks are respectively used to realize the image quality evaluation value for calculating every image in testing image set, and
Sample mode is set using decimation blocks, generates preferential cycle tests；
Sequence playing module: being mainly used for the broadcasting of image pair when subjective assessment, different play mode may be selected, be defaulted as with
Machine sowing is put；
Data management module: it is saved including evaluation marking, data and label calculates three submodules, for realizing score data
Record, preservation and result statistical analysis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910020772.4A CN109714591B (en) | 2019-01-09 | 2019-01-09 | Image quality subjective evaluation method and system based on evaluation label |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910020772.4A CN109714591B (en) | 2019-01-09 | 2019-01-09 | Image quality subjective evaluation method and system based on evaluation label |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109714591A true CN109714591A (en) | 2019-05-03 |
| CN109714591B CN109714591B (en) | 2020-08-25 |
Family
ID=66260981
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910020772.4A Active CN109714591B (en) | 2019-01-09 | 2019-01-09 | Image quality subjective evaluation method and system based on evaluation label |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109714591B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN111179238A (en) * | 2019-12-24 | 2020-05-19 | 东华大学 | Subset confidence ratio dynamic selection method for subset-oriented guidance consistency enhancement evaluation |
| CN111556192A (en) * | 2020-03-24 | 2020-08-18 | 上海交通大学 | Mobile phone photo dragging comparison sorting method and system |
| CN112950567A (en) * | 2021-02-25 | 2021-06-11 | 北京金山云网络技术有限公司 | Quality evaluation method, quality evaluation device, electronic device, and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080029371A (en) * | 2006-09-29 | 2008-04-03 | 광운대학교 산학협력단 | Method of image quality evaluation, and system thereof |
| CN103458267A (en) * | 2013-09-04 | 2013-12-18 | 中国传媒大学 | Video picture quality subjective evaluation method and system |
| CN103632369A (en) * | 2013-12-05 | 2014-03-12 | 淮海工学院 | Method for universally detecting quality of non-reference underwater images on basis of combination of block average definition |
-
2019
- 2019-01-09 CN CN201910020772.4A patent/CN109714591B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080029371A (en) * | 2006-09-29 | 2008-04-03 | 광운대학교 산학협력단 | Method of image quality evaluation, and system thereof |
| CN103458267A (en) * | 2013-09-04 | 2013-12-18 | 中国传媒大学 | Video picture quality subjective evaluation method and system |
| CN103632369A (en) * | 2013-12-05 | 2014-03-12 | 淮海工学院 | Method for universally detecting quality of non-reference underwater images on basis of combination of block average definition |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN111179238A (en) * | 2019-12-24 | 2020-05-19 | 东华大学 | Subset confidence ratio dynamic selection method for subset-oriented guidance consistency enhancement evaluation |
| CN111179238B (en) * | 2019-12-24 | 2022-12-20 | 东华大学 | Subset confidence ratio dynamic selection method for underwater image set-oriented guidance consistency enhancement evaluation |
| CN111556192A (en) * | 2020-03-24 | 2020-08-18 | 上海交通大学 | Mobile phone photo dragging comparison sorting method and system |
| CN111556192B (en) * | 2020-03-24 | 2021-05-14 | 上海交通大学 | Mobile phone photo dragging comparison sorting method and system |
| CN112950567A (en) * | 2021-02-25 | 2021-06-11 | 北京金山云网络技术有限公司 | Quality evaluation method, quality evaluation device, electronic device, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109714591B (en) | 2020-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109714591A (en) | Based on the picture quality subjective evaluation method and system to assessment label | |
| CN105741328B (en) | The shooting image quality evaluating method of view-based access control model perception | |
| AU2009301638B2 (en) | Modified apparatus and method for assessment, evaluation and grading of gemstones | |
| Peletier et al. | The distribution of dust in Sb's and Sc's-K-band infrared imaging of a diameter limited sample of 37 galaxies. | |
| Yee et al. | A perceptual metric for production testing | |
| CN101207832A (en) | Method for checking digital camera color reduction | |
| CN109741285A (en) | A kind of construction method and system of underwater picture data set | |
| JP2004073802A (en) | Multipurpose measuring unit for skin | |
| Yang et al. | Pair comparison based progressive subjective quality ranking for underwater images | |
| Meyer et al. | Electronic image analysis of crop residue cover on soil | |
| Foster et al. | Colour constancy failures expected in colourful environments | |
| US7024037B2 (en) | Cross-polarized imaging method for measuring skin ashing | |
| US11137355B2 (en) | Process and system for diamond clarity measurement | |
| CN113092489A (en) | System and method for detecting appearance defects of battery | |
| CN108235831B (en) | The control method and control system and electronic equipment of lamps and lanterns | |
| KR101901812B1 (en) | A method for discriminating new rice and staled rice using electronic eye | |
| RU2008141365A (en) | METHOD AND DEVICE FOR MEASURING SURFACE QUALITY OF SUBSTRATE | |
| Milovanovic et al. | Computer Vision System: A better tool for assessing pork and beef colour than a standard colourimeter | |
| Storlie et al. | Growth analysis of whole plants using video imagery | |
| Nyman et al. | From image fidelity to subjective quality: a hybrid qualitative/quantitative methodology for measuring subjective image quality for different image contents | |
| Yun et al. | Fashion image digital color analysis method | |
| Vlasuyk et al. | Synthesis of the Perceptionally Linear Color Space Using Machine Learning Methods | |
| Montrucchio et al. | Thresholds of vision of the human visual system: Visual adaptation for monocular and binocular vision | |
| CN108596901A (en) | A kind of implementation method of intelligence test and appraisal real shooting photo imaging image quality | |
| JP2000227399A (en) | Preliminary screening device of blood test body |
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 |