CN107659806A - The appraisal procedure and device of video quality - Google Patents
The appraisal procedure and device of video quality Download PDFInfo
- Publication number
- CN107659806A CN107659806A CN201710725869.6A CN201710725869A CN107659806A CN 107659806 A CN107659806 A CN 107659806A CN 201710725869 A CN201710725869 A CN 201710725869A CN 107659806 A CN107659806 A CN 107659806A
- Authority
- CN
- China
- Prior art keywords
- quality
- reference picture
- mass
- distorted image
- evaluating
- 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
- 238000000034 method Methods 0.000 title claims abstract description 249
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims description 29
- 238000010801 machine learning Methods 0.000 claims description 18
- 238000013441 quality evaluation Methods 0.000 claims description 14
- 239000012141 concentrate Substances 0.000 claims description 4
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000001303 quality assessment method Methods 0.000 abstract description 4
- 238000011156 evaluation Methods 0.000 description 15
- 238000004891 communication Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 6
- 230000000007 visual effect Effects 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 5
- 230000000903 blocking effect Effects 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000003646 Spearman's rank correlation coefficient Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004549 pulsed laser deposition Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 208000013021 vision distortion Diseases 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N17/00—Diagnosis, testing or measuring for television systems or their details
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
- Image Analysis (AREA)
Abstract
The application provides a kind of appraisal procedure and device of video quality, and this method includes:Obtain training dataset, the training data, which is concentrated, includes at least one source images and at least one distorted image corresponding with each source images;Calculate the first mass of each source images;Calculate second mass of at least one distorted image relative to the source images;According to first mass and second mass, the quality of image is determined.The appraisal procedure and device of the video quality that the application provides can improve the accuracy of video quality assessment.
Description
Technical field
The invention relates to image and video quality evaluation field, more particularly to a kind of appraisal procedure of video quality
And device.
Background technology
With the development of the communication technology and multimedia technology, video is widely developed and applied, its application scenarios
More complicate, therefore, how the quality of video is accurately assessed, the problem of being one extremely important.
Fig. 1 is the schematic flow sheet assessed in the prior art video quality, as shown in figure 1, dividing video into
After multiple image, after benchmark image is encoded by encoder obtain assess object, and assume assess object include blocking effect and
Two kinds of type of distortion are obscured, benchmark image and assessment object are had into reference picture method for evaluating quality (method 1 by multiple respectively
With method 2) assessed, wherein, method 1 and method 2 for example can be visual information fidelity (Visual Information
Fidelity;) and loss in detail index (Detail Loss Metric VIF;DLM) etc., there is reference picture quality by multiple
After appraisal procedure carries out quality evaluation, then multiple results of acquisition are weighted processing by machine learning, to obtain the figure
The quality score of picture.
However, of the prior art have reference picture method for evaluating quality, due to being assuming that the benchmark image chosen is
Carried out on the premise of perfect quality, so as to cause the assessment result of video quality inaccurate.
The content of the invention
The embodiment of the present application provides a kind of appraisal procedure and device of video quality, to solve the assessment result of video quality
Inaccurate technical problem.
The application first aspect provides a kind of appraisal procedure of video quality, including:
Obtain training dataset, the training data, which is concentrated, includes at least one source images and corresponding with each source images
At least one distorted image;
Calculate the first mass of each source images;
Calculate second mass of at least one distorted image relative to the source images;
According to first mass and second mass, the quality of image is determined.
In this programme, source images are high quality graphic, and server can pass through non-reference picture method for evaluating quality meter
The first mass of each source images is calculated, and by there is reference picture method for evaluating quality to calculate each distortion corresponding to each source images
Second mass of image, finally combine the first mass and the second mass calculated, it may be determined that go out the image that needs are assessed
Quality.
In such scheme, include at least one source images and at least one distortion corresponding with each source images by obtaining
The training dataset of image, and the first mass of each source images, and at least one distorted image are calculated relative to source images
Second mass, further according to the first mass and the second mass, determine the quality of image.Due to the first matter by calculating source images
Amount, and the quality for the image for needing to assess according to the first mass calculated and the determination of the second mass, so as to avoid existing skill
Needed in art assuming that the benchmark image chosen be it is perfect under the premise of carry out the phenomenon of quality evaluation, it is possible thereby to improve video
The accuracy of quality evaluation.
In one implementation, it is described to calculate second matter of at least one distorted image relative to the source images
Amount, including:
Determine the type of distortion and distortion level of each distorted image;
According to the type of distortion and the distortion level, each distorted image is grouped, obtains at least one distortion
Image sets, wherein, the type of distortion and distortion level all same of the distorted image in same distorted image group;
According to multiple first assessment results and the second assessment result, have from m in reference picture method for evaluating quality and determine
N therein have reference picture method for evaluating quality;First assessment result has reference picture matter to be respectively adopted the m
Appraisal procedure is measured, the result obtained after assessing the distorted image in each distorted image group, described second assesses knot
Fruit is the result obtained after user's subjectivity is assessed each distorted image, and m and n are positive integer, and m is more than or equal to n;
There is reference picture method for evaluating quality according to the n, calculate second mass.
In one implementation, it is described according to multiple first assessment results and the second assessment result, there is reference chart from m
As determining n method therein in method for evaluating quality, including:
There is reference picture method for evaluating quality by m, the distorted image in each distorted image group is carried out respectively
Assess, obtain multiple first assessment results;
In each distorted image group, according to multiple first assessment results, there is reference picture quality to comment from the m
Estimate and determine that t therein have reference picture method for evaluating quality in method;The t have reference picture method for evaluating quality be by
Have according to m consistent between first assessment result and second assessment result corresponding to reference picture method for evaluating quality
The order of property degree from high to low, there is reference picture method for evaluating quality corresponding to the preceding t degree of consistency selected;
Being had according to the t corresponding to all distorted image groups in reference picture method for evaluating quality each has reference
The order of the frequency that image quality measure method occurs from high to low, n have reference picture method for evaluating quality before determining;Its
In, t is the positive integer less than or equal to m, and more than or equal to n.
In this programme, it is abnormal that type of distortion can for example include fuzzy, blocking effect, color exception, flower screen and contrast
Deng, distortion level such as can include slight, moderate and severe, it can be indicated with discrete range format.In addition,
Existing popular have reference picture method for evaluating quality to include Y-PSNR (Peak Signal to Noise
Ratio;PSNR), structural similarity (Structural SIMilarity;SSIM), Multi-scale model similitude (Multi-
Scale SSIM;MS-SSIM), visual information fidelity (Visual Information Fidelity;VIF), loss in detail
Index (Detail Loss Metric;) and characteristic similarity (Feature SIMilarity DLM;FSIM) etc..
In one implementation, it is described to have reference picture method for evaluating quality according to the n, calculate second matter
Amount, including:
Respectively by the n quality for thering is reference picture method for evaluating quality to calculate the distorted image, the multiple 3rd is obtained
Quality;
The multiple 3rd mass is weighted processing by machine learning, obtains second mass.
In this programme, server will have by n respectively after determining that n have reference picture method for evaluating quality
Reference picture method for evaluating quality calculates the quality of each distorted image, obtains multiple 3rd mass, and pass through machine learning side
Multiple 3rd mass are weighted processing by method, so as to obtain the second mass.Wherein, machine learning method can include support to
Amount returns (Support Vector Machine;) and support vector regression (Support Vector Regression SVM;
SVR) etc..
It is subjective with user using the result that a variety of machine learning methods obtain on being weighted processing by machine learning
The result of evaluation is contrasted, and optimal performance appraisal procedure is obtained according to Performance Evaluation index (PCC, SROCC, RMSE), so as to
The accuracy of the second Mass Calculation can be improved.Wherein, Performance Evaluation index includes Pearson came linearly dependent coefficient (Pearson
Linear Correlation Efficient;PLCC), Spearman rank correlation coefficient (Spearman's Rank
Correlation Coefficient;) and root-mean-square error (Root Mean Square Error SROCC;RMSE) etc..
In one implementation, first mass for calculating each source images, including:
According to the pixel domain of each source images, first mass is calculated.
In one implementation, it is described according to first mass and second mass, the quality of image is determined, is wrapped
Include:
First mass is subjected to standardization processing, obtains quality coefficient, the value of the quality coefficient is more than or equal to
0.0, and less than or equal to 1.0;
The quality coefficient is multiplied by second mass, and the result of acquisition is defined as to the quality of described image.
The application second aspect provides a kind of apparatus for evaluating of video quality, including:
Acquiring unit, for obtaining training dataset, the training data concentrate include at least one source images and with it is each
At least one distorted image corresponding to the source images;
Computing unit, for calculating the first mass of each source images;
The computing unit, it is additionally operable to calculate second matter of at least one distorted image relative to the source images
Amount;
Determining unit, for according to first mass and second mass, determining the quality of image.
In one implementation, the computing unit, including:
Determination subelement, for determining the type of distortion and distortion level of each distorted image;
Subelement is grouped, for according to the type of distortion and the distortion level, being grouped, obtaining to each distorted image
At least one distorted image group is obtained, wherein, the type of distortion and distortion level of the distorted image in same distorted image group
All same;
The determination subelement, it is additionally operable to according to multiple first assessment results and the second assessment result, has reference chart from m
As determining that n therein have reference picture method for evaluating quality in method for evaluating quality;First assessment result is to adopt respectively
There is reference picture method for evaluating quality with the m, obtained after assessing the distorted image in each distorted image group
Result, second assessment result is the result that obtains after user's subjectivity is assessed each distorted image, and m and n are
Positive integer, and m is more than or equal to n;
Computation subunit, for having reference picture method for evaluating quality according to the n, calculate second mass.
In one implementation, the determination subelement, is specifically used for:
There is reference picture method for evaluating quality by m, the distorted image in each distorted image group is carried out respectively
Assess, obtain multiple first assessment results;
In each distorted image group, according to multiple first assessment results, there is reference picture quality to comment from the m
Estimate and determine that t therein have reference picture method for evaluating quality in method;The t have reference picture method for evaluating quality be by
Have according to m consistent between first assessment result and second assessment result corresponding to reference picture method for evaluating quality
The order of property degree from high to low, there is reference picture method for evaluating quality corresponding to the preceding t similarity selected;
Being had according to the t corresponding to all distorted image groups in reference picture method for evaluating quality each has reference
The order of the frequency that image quality measure method occurs from high to low, n have reference picture method for evaluating quality before determining;Its
In, t is the positive integer less than or equal to m, and more than or equal to n.
In one implementation, the computation subunit, is specifically used for:
Respectively by the n quality for thering is reference picture method for evaluating quality to calculate the distorted image, the multiple 3rd is obtained
Quality;
The multiple 3rd mass is weighted processing by machine learning, obtains second mass.
In one implementation, the computing unit, is specifically used for:
According to the pixel domain of each source images, first mass is calculated.
In one implementation, the determining unit is specifically used for:
First mass is subjected to standardization processing, obtains quality coefficient, the value of the quality coefficient is more than or equal to
0.0, and less than or equal to 1.0;
The quality coefficient is multiplied by second mass, and the result of acquisition is defined as to the quality of described image.
The application third aspect provides a kind of apparatus for evaluating of video quality, and the device includes processor and memory, deposited
Reservoir is used for storage program, and processor calls the program of memory storage, to perform the method for the application first aspect offer.
The application fourth aspect provides a kind of server, including for perform above first aspect method it is at least one
Treatment element (or chip).
The aspect of the application the 5th provides a kind of appraisal procedure of video quality, and the program is used to hold when being executed by processor
The method of row above first aspect.
A kind of program product of the aspect offer of the application the 6th, such as computer-readable recording medium, including the 5th aspect
Program.
The application provide video quality appraisal procedure and device, by obtain include at least one source images and with it is each
The training dataset of at least one distorted image corresponding to source images, and the first mass of each source images is calculated, and at least one
Individual distorted image further according to the first mass and the second mass, determines the quality of image relative to the second mass of source images.Due to
By calculating the first mass of source images, and the image for being determined according to the first mass calculated and the second mass to need to assess
Quality, so as to avoid need in the prior art assuming that the benchmark image chosen be it is perfect under the premise of carry out quality evaluation
Phenomenon, it is possible thereby to improve the accuracy of video quality assessment.
Brief description of the drawings
Fig. 1 is the schematic flow sheet assessed in the prior art video quality;
Fig. 2 is the usage scenario schematic diagram of the appraisal procedure for the video quality that the embodiment of the present application provides;
Fig. 3 is the schematic flow sheet of the appraisal procedure embodiment one for the video quality that the embodiment of the present application provides;
Fig. 4 is the schematic flow sheet for calculating the second mass;
Fig. 5 is the structural representation of the apparatus for evaluating embodiment one for the video quality that the embodiment of the present application provides;
Fig. 6 is the structural representation of the apparatus for evaluating embodiment two for the video quality that the embodiment of the present application provides;
Fig. 7 A are a kind of possible structural representation of the application server;
Fig. 7 B are the alternatively possible structural representation of the application server.
Embodiment
The appraisal procedure for the video quality that the embodiment of the present application provides, go for the scene that volume/transcoded quality is assessed
In, Fig. 2 is the usage scenario schematic diagram of the appraisal procedure for the video quality that the embodiment of the present application provides, as shown in Fig. 2 video takes
The flow of business is as follows:(1) video source contents, i.e. film source are initially injected;(2) film source is switched to by various code rate by volume/transcoder
Version, various code rate version correspond to different quality, in this process, in order to ensure the quality of volume/transcoder output stream
Meet predeterminated target requirement, it is therefore desirable to coding source quality is assessed, and then cataloged procedure is carried out according to assessment result and determined
Plan, it will reset coding parameter when such as coding output quality is not inconsistent with aimed quality setting and implement new cataloged procedure, so,
It can be very good the experience of guarantee user;(3) packed, encapsulated and sent out for meeting the encoded output stream of predeterminated target requirement
Stream;(4) video flowing carries out network transmission;(5) user watches Video service by terminal device.A kind of therefore it provides accuracy
The appraisal procedure of higher video quality, for ensureing that the experience of user is very important.
The appraisal procedure for the video quality that the application provides, it is intended to which the appraisal procedure for solving video quality in the prior art is commented
The technical problem for the result inaccuracy estimated.
How to be solved to the technical scheme of the application and the technical scheme of the application with specifically embodiment below above-mentioned
Technical problem is described in detail.These specific embodiments can be combined with each other below, for same or analogous concept
Or process may repeat no more in certain embodiments.Below in conjunction with accompanying drawing, embodiments herein is described.
Fig. 3 is the schematic flow sheet of the appraisal procedure embodiment one for the video quality that the embodiment of the present application provides.The application
The appraisal procedure for the video quality that embodiment provides can be performed by the device for the appraisal procedure for arbitrarily performing video quality, should
Device can be realized by software and/or hardware.In the present embodiment, the device can integrate in the server.As shown in figure 3,
The method of the present embodiment can include:
Step 301, obtain training dataset, the training data concentrate include at least one source images and with each source images pair
At least one distorted image answered.
In the present embodiment, training data, which is concentrated, includes at least one source images and corresponding with each source images at least one
Distorted image, wherein, source images are at least two field picture in video to be assessed, and the source images are high quality and undistorted
Image.For each source images, after passing it through encoder coding, it will at least one distorted image corresponding to obtaining,
From each source images corresponding at least one distorted image usually require the mistake that is related to different type of distortion and different levels
True degree.Wherein, type of distortion is such as can include fuzzy, blocking effect, color exception, flower shields and contrast is abnormal, distortion
Degree is such as can include slight, moderate and severe, and it can be indicated with discrete range format, such as the mistake of image
When really spending between 0-30%, distortion level is slight, and when the distortion factor is between 31%-70%, distortion level is moderate, distortion
When degree is between 71%-100%, distortion level is severe etc., the distortion level of the above by way of example only, in concrete implementation
During, the grade of distortion level and the specific division to each grade can be configured according to actual conditions or experience,
Grade for distortion level and the specific dividing mode to each grade, this is not restricted for the present embodiment.
Step 302, the first mass for calculating each source images.
In the present embodiment, server will calculate the first mass of each source images after training dataset is got,
That is the proper mass of source images.In a kind of possible embodiment, server will pass through non-reference picture method for evaluating quality
Calculate the first mass of each source images.The first mass is calculated herein according to the pixel domain of each source images.Specifically, no reference
Image quality measure method when assessing image/video quality, does not contrast benchmark, directly according to the feature of object to be assessed
Or quality score is calculated in parameter.The quality of pixel domain is assesses the essential proper mass of object, and the quality of compression domain is then
It is mass change caused by the compression process assessed, premised on vision distortion problem is not present in the input of encoder..
Step 303, calculate second mass of at least one distorted image relative to source images.
In the present embodiment, server will calculate at least one corresponding to each source images after training dataset is got
Second mass of distorted image, i.e. distorted image relative to source images quality.In one embodiment, server will pass through
There is the second mass that reference picture method for evaluating quality calculates at least one distorted image, further, server passes through mixing
Type has reference picture method for evaluating quality to calculate the second mass.
Fig. 4 is the schematic flow sheet for calculating the second mass, as shown in figure 4, the calculation specifically includes:
Step 3031, the type of distortion and distortion level for determining each distorted image.
Specifically, distorted image corresponding to each source images that training data is concentrated, may all correspond to a variety of distortion classes
Type, each type of distortion may correspond to a variety of distortion levels, and therefore, for each distorted image, server will determine to lose
The type of distortion and distortion level of true image, wherein, type of distortion can include fuzzy, blocking effect, color exception, flower shields and right
More abnormal etc. than degree, distortion level can determine according to the distortion factor of image.Such as:Distorted image corresponding to source images 1 is image
11st, image 12, image 13, image 14 and image 15, the type of distortion of image 11 is fuzzy, and distortion level is slight, image 12
Type of distortion be fuzzy, distortion level is severe, and the type of distortion of image 13 is blocking effect, and distortion level is moderate, image
14 type of distortion is that color is abnormal, and distortion level is moderate, and the type of distortion of image 15 is abnormal for color, and distortion level is light
Degree.
Need to illustrate, calculated in step 302 and type of distortion and distortion are determined in first mass and step 3031
It is decoupling between the process of degree, wherein, the first mass is calculated in step 302 can be not dependent on distortion in step 3031
Detection or rely on formula, but have differences therebetween:Independent of the mode of distortion detection, it is to video quality
Assessment result accuracy higher than the mode relied on, and be applied to weak distortion scene, the nothing point use of participating in evaluation and electing herein for source is non-
Dependence formula.
Step 3032, according to type of distortion and distortion level, each distorted image is grouped, obtains at least one distortion
Image sets, wherein, the type of distortion and distortion level section all same of the distorted image in same distorted image group.
Specifically, server is after the type of distortion of each distorted image and distortion level is determined, will be according to determining
Type of distortion and distortion level distorted image is grouped, can be by type of distortion and mistake during concrete implementation
The distorted image of true degree all same is divided into one group.Table 1 shows the result being grouped to each distorted image, as shown in table 1,
Type of distortion can use a, b, c ... to represent, distortion level can be represented with 1,2,3 ... M.
Table 1
Step 3033, according to multiple first assessment results and the second assessment result, have reference picture quality evaluation side from m
N are determined in method reference picture method for evaluating quality;First assessment result has reference picture quality to comment to be respectively adopted m
Estimate method, the result obtained after assessing the distorted image in each distorted image group, second assessment result is to use householder
The result obtained after assessing each distorted image is seen, m and n are positive integer, and m is more than or equal to n.
Specifically, the mode for quality evaluation being carried out to image/video generally includes two classes, respectively subjective evaluation method and
Objective evaluation method, wherein, subjective evaluation method be user in specific controlled environment (including viewing distance, viewing duration,
Illumination, the selection of test object, personnel's selection etc.) test video is beaten by subjective feeling according to the opinion scale of regulation
Point, all evaluation score values are weighted and averagely obtain Mean Opinion Score (Mean Opinion Score;MOS) value, i.e., it is subjective
Scoring.Subjective method can reflect the quality of image/video exactly, be real Consumer's Experience.But this method implement it is cumbersome,
It poor real and can not automate, be unfavorable for the system integration and realization, it is impossible to be used in actual business.Objective method is a kind of accurate
It is true, being easily achieved, the appraisal procedure used can be automated, its accuracy passes through the uniformity journey with subjective method result
Degree is weighed.Its realization is the feature or parameter according to evaluation object, passes through and calculates acquisition quality score.Objective method is according to ginseng
Examine source information (assessing benchmark) use be divided into have with reference to and without refer to two classes.There is reference (reference) method:Assess
During the quality of image/video object, there are the benchmark of contrast, i.e. reference source, by the feature for differentiating object and reference source to be assessed
Or difference or the change of parameter, obtain assessment result.
Above-mentioned subjective evaluation method can be utilized in the present embodiment and utilizes objective evaluation method, has reference chart from m
As determining that n have reference picture method for evaluating quality in method for evaluating quality, wherein, the first assessment result is individual for m is respectively adopted
There is reference picture method for evaluating quality, the result obtained after assessing the distorted image in each distorted image group, that is, utilize
The result that above-mentioned objective evaluation method obtains, the second assessment result are to be obtained after user's subjectivity is assessed each distorted image
As a result, i.e., the result obtained using above-mentioned subjective evaluation method.
Below, it will be described in detail and how to determine the n processes for having reference picture method for evaluating quality.
Server has reference picture method for evaluating quality by m first, respectively to the distortion map in each distorted image group
As being assessed, multiple first assessment results are obtained;In each distorted image group, according to multiple first assessment results, from m
Have and determine that t have reference picture method for evaluating quality in reference picture method for evaluating quality, wherein, t have reference picture quality
Appraisal procedure is to have according to m corresponding to reference picture method for evaluating quality between the first assessment result and the second assessment result
The order of the degree of consistency from high to low, there is reference picture method for evaluating quality corresponding to the preceding t degree of consistency selected;
Being had according to t corresponding to all distorted image groups each has reference picture method for evaluating quality in reference picture method for evaluating quality
The order of the frequency of appearance from high to low, n has reference picture method for evaluating quality before determining, wherein, t be less than or equal to
M, and the positive integer more than or equal to n.
Specifically, it is existing popular to there is reference picture method for evaluating quality to include Y-PSNR (Peak Signal
to Noise Ratio;PSNR), structural similarity (Structural SIMilarity;SSIM), Multi-scale model similitude
(Multi-Scale SSIM;MS-SSIM), visual information fidelity (Visual Information Fidelity;VIF it is), thin
Section loses index (Detail Loss Metric;) and characteristic similarity (Feature SIMilarity DLM;FSIM) etc., when
So, in addition to above-mentioned appraisal procedure, reference picture method for evaluating quality includes other method, for specifically having ginseng
Image quality measure method is examined, can be selected according to actual conditions, the present embodiment is not restricted to this.
Server by m it is different have reference picture method for evaluating quality, respectively to the mistake in each distorted image group
True image is assessed, such as:The distorted image group that type of distortion is a and distortion level is 1 include image 11, image 12,
Image 13 and image 14, then image 11, image 12, image 13 and image 14 can be assessed respectively by PSNR methods,
The first assessment result is obtained, image 11, image 12, image 13 and image 14 are assessed respectively by SSIM methods, is obtained
First assessment result, image 11, image 12, image 13 and image 14 are assessed respectively by MS-SSIM methods, obtain the
One assessment result, image 11, image 12, image 13 and image 14 are assessed respectively by VIF methods, obtain first and assess
As a result etc., similarly, each distorted image in other distorted image groups is also similarly evaluated, it is possible thereby to obtain multiple
One assessment result.
For each distorted image group, after multiple first assessment results are obtained, can according to the first assessment result and
The order of the degree of consistency from high to low between second assessment result, there is reference corresponding to the t degree of consistency before selecting
Image quality measure method.As an example it is assumed that m is 4, t 3, the distorted image group that type of distortion is a and distortion level is 1
Include image 11, image 12, image 13 and image 14, what user's subjectivity obtained after assessing each distorted image second comments
The score value for estimating result is 80,90,85 and 82, and server is by PSNR methods respectively to image 11, image 12, image 13 and image
14 are assessed, and the score value of the first assessment result of acquisition is 80,88,83 and 82, the first assessment result and the second assessment result
Between the degree of consistency be 95%, server is entered to image 11, image 12, image 13 and image 14 respectively by SSIM methods
Row is assessed, and the score value for obtaining the first assessment result is 75,93,82 and 80, between the first assessment result and the second assessment result
The degree of consistency is 85%, and image 11, image 12, image 13 and image 14 are assessed respectively by MS-SSIM methods, obtained
The score value for obtaining the first assessment result is 78,90,82 and 83, the degree of consistency between the first assessment result and the second assessment result
For 91%, image 11, image 12, image 13 and image 14 are assessed respectively by VIF methods, obtain the first assessment result
Score value be 81,92,83 and 82, the degree of consistency between the first assessment result and the second assessment result is 93%, then according to
The order of the degree of consistency from high to low between first assessment result and the second assessment result, preceding 3 uniformity selected
It is respectively PSNR methods, VIF methods and MS-SSIM methods to have reference picture method for evaluating quality corresponding to degree, other distortions
T is determined in image sets that the mode of reference picture method for evaluating quality is similar to the above, and here is omitted.So, may be used
To determine there is reference picture method for evaluating quality corresponding to the preceding t degree of consistency in all distorted image groups.Shown in table 2
The t appraisal procedure selected in all distorted image groups, wherein, the appraisal procedure selected in each distorted image group can
It with identical, can also completely differ, can also be that part is identical, partly differ.
Table 2
Have in all distorted image groups are determined corresponding to the preceding t degree of consistency reference picture method for evaluating quality it
Afterwards, will count in all distorted image groups, each frequency for thering is reference picture method for evaluating quality to occur selected, and according to
The order of the frequency from high to low, n have reference picture method for evaluating quality before determining.Such as:Assuming that n is 3, if distortion class
Selected in the distorted image group that type is a and distortion level is 1 have reference picture method for evaluating quality include PSNR methods,
VIF methods and MS-SSIM methods, if that is selected in the distorted image group that type of distortion is a and distortion level is 2 has reference chart
Picture method for evaluating quality includes PSNR methods, FSIM methods and DLM methods, if the distortion that type of distortion is a and distortion level is 3
That is selected in image sets has reference picture method for evaluating quality to include SSIM methods, VIF methods and DLM methods, if distortion class
Selected in the distorted image group that type is b and distortion level is 1 have reference picture method for evaluating quality include PSNR methods,
VIF methods and DLM methods, if that is selected in the distorted image group that type of distortion is b and distortion level is 2 has reference picture matter
Amount appraisal procedure includes PSNR methods, MS-SSIM methods and FSIM methods, if the distortion that type of distortion is b and distortion level is 3
That is selected in image sets has reference picture method for evaluating quality to include SSIM methods, VIF methods and PSNR methods, then all
That is selected in distorted image group has the order of the frequency of reference picture method for evaluating quality appearance to be from high to low followed successively by:PSNR
Method (5 times), VIF methods (4 times), DLM methods (3 times), MS-SSIM methods (2 times), FSIM methods (2 times) and SSIM methods
(2 times), so, you can determining first 3, to have reference picture method for evaluating quality be respectively PSNR methods, VIF methods and DLM side
Method.
Step 3034, there are reference picture method for evaluating quality, the second mass of calculating according to n.
Specifically, server will have reference by n respectively after determining that n have reference picture method for evaluating quality
Image quality measure method calculates the quality of each distorted image, obtains multiple 3rd mass, and will by machine learning method
Multiple 3rd mass are weighted processing by machine learning, so as to obtain the second mass.Wherein, machine learning method can wrap
Include support vector regression (Support Vector Machine;) and support vector regression (Support Vector SVM
Regression;SVR) etc., it is, of course, also possible to including other machine learning methods, as long as multiple 3rd mass can be carried out
Weighted average processing, for the concrete form of machine learning method, this is not restricted for the present embodiment.The n such as determined
It is individual to there is reference picture method for evaluating quality collection to be combined into R={ method 1, method 2 ..., method n }, use each method in set R
After assessing distorted image, the 3rd mass of acquisition is Q={ Q1, Q2 ..., Qn }, by SVM or SVR to multiple three
Quality is weighted processing, obtains the second mass.
For example, each distorted image is assessed by PSNR methods, VIF methods and DLM methods respectively, obtained more
Individual 3rd mass, multiple 3rd mass of acquisition are weighted processing by SVM methods, so as to obtain the second mass.
It should be noted that the method for a variety of machine learning can be used to be weighted processing, and by each machine learning
The result that method obtains and the result of user's subjective assessment are contrasted, and are obtained according to Performance Evaluation index (PCC, SROCC, RMSE)
To optimal performance appraisal procedure, so as to improve the accuracy of the second Mass Calculation.Wherein, Performance Evaluation index includes Pierre
Inferior linearly dependent coefficient (Pearson Linear Correlation Efficient;PLCC), Spearman rank correlation system
Number (Spearman's Rank Correlation Coefficient;) and root-mean-square error (Root Mean SROCC
Square Error;RMSE) etc..
In the present embodiment, data set can be grouped according to the type of distortion and distortion level of distorted image, from existing m
It is individual to have in reference picture method for evaluating quality, select and assessed accurately for the distorted image of different type of distortion and distortion level
N higher method of property, and processing is weighted by machine learning, the quality of distorted image is obtained, so as to improve figure
As the accuracy of quality evaluation.
Additionally, it is appreciated that the step 302 of the above and the execution sequence of step 303 are only a kind of signals.Step 302 with
The differentiation for the sequencing that step 302 is not carried out, step 302 can be first carried out, then perform step 303;It can also first carry out
Step 303, then step 302 is performed;The two steps can also be performed simultaneously, and the embodiment of the present application is not specially limited to this.
Step 304, according to the first mass and the second mass, determine the quality of image.
In the present embodiment, server is calculating the first mass of source images and corresponding with source images at least one
After second mass of distorted image, according to the first mass and the second mass, it can will determine to need the matter of image assessed
Amount.
Furthermore it is possible to by the way that the first mass is carried out into standardization processing, quality coefficient is obtained, and the quality coefficient is multiplied by
Second mass, the result of acquisition is defined as to the quality for the image that needs are assessed, wherein, the value of the quality coefficient is more than or waited
In 0.0, and less than or equal to 1.0.
Further, server carries out quality evaluation in every two field picture in video according to the method in above steps
And then the time response of the quality evaluation result of every two field picture and video is combined, you can obtain the assessment of video quality
As a result.Wherein, time response is frame level kinetic characteristic, such as take the luminance difference of the pixel of consecutive frame etc..
The embodiment of the present application provide video quality appraisal procedure, by obtain include at least one source images and with it is each
The training dataset of at least one distorted image corresponding to source images, and the first mass of each source images is calculated, and at least one
Individual distorted image further according to the first mass and the second mass, determines the quality of image relative to the second mass of source images.Due to
By calculating the first mass of source images, and the image for being determined according to the first mass calculated and the second mass to need to assess
Quality, so as to avoid need in the prior art assuming that the benchmark image chosen be it is perfect under the premise of carry out quality evaluation
Phenomenon, it is possible thereby to improve the accuracy of video quality assessment.
Fig. 5 is the structural representation of the apparatus for evaluating embodiment one for the video quality that the embodiment of the present application provides, referring to figure
5, the apparatus for evaluating includes:Acquiring unit 11, computing unit 12 and determining unit 13, wherein:
Acquiring unit 11 is used to obtain training dataset, the training data concentrate include at least one source images and with it is each
At least one distorted image corresponding to the source images;
Computing unit 12 is used for the first mass for calculating each source images;
The computing unit 12 is additionally operable to calculate second matter of at least one distorted image relative to the source images
Amount;
Determining unit 13 is used for the quality for according to first mass and second mass, determining image.
Said apparatus can be used for performing the method that above-mentioned corresponding method embodiment provides, specific implementation and technique effect
It is similar, repeat no more here.
Fig. 6 is the structural representation of the apparatus for evaluating embodiment two for the video quality that the embodiment of the present application provides, referring to figure
6, on the basis of embodiment illustrated in fig. 5, the computing unit 12, including:
Determination subelement 121 is used for the type of distortion and distortion level for determining each distorted image;
Subelement 122 is grouped to be used to, according to the type of distortion and the distortion level, be grouped each distorted image,
At least one distorted image group is obtained, wherein, the type of distortion and distortion journey of the distorted image in same distorted image group
Spend all same;
The determination subelement 121 is additionally operable to according to multiple first assessment results and the second assessment result, has reference from m
N therein are determined in image quality measure method reference picture method for evaluating quality;First assessment result is difference
There is reference picture method for evaluating quality using the m, obtained after assessing the distorted image in each distorted image group
The result obtained, second assessment result are the result obtained after user's subjectivity is assessed each distorted image, m and n
For positive integer, and m is more than or equal to n;
Computation subunit 123 is used for having reference picture method for evaluating quality according to the n, calculates second mass.
Alternatively, the determination subelement 121 is specifically used for:
There is reference picture method for evaluating quality by m, the distorted image in each distorted image group is carried out respectively
Assess, obtain multiple first assessment results;
In each distorted image group, according to multiple first assessment results, there is reference picture quality to comment from the m
Estimate and determine that t therein have reference picture method for evaluating quality in method;The t have reference picture method for evaluating quality be by
Have according to m consistent between first assessment result and second assessment result corresponding to reference picture method for evaluating quality
The order of property degree from high to low, there is reference picture method for evaluating quality corresponding to the preceding t similarity selected;
Being had according to the t corresponding to all distorted image groups in reference picture method for evaluating quality each has reference
The order of the frequency that image quality measure method occurs from high to low, n have reference picture method for evaluating quality before determining;Its
In, t is the positive integer less than or equal to m, and more than or equal to n.
Alternatively, the computation subunit 123 is specifically used for:
Respectively by the n quality for thering is reference picture method for evaluating quality to calculate the distorted image, the multiple 3rd is obtained
Quality;
The multiple 3rd mass is weighted processing by machine learning, obtains second mass.
Alternatively, the computing unit 123 is specifically used for:
According to the pixel domain of each source images, first mass is calculated.
Alternatively, the determining unit 13 is specifically used for:
First mass is subjected to standardization processing, obtains quality coefficient, the value of the quality coefficient is more than or equal to
0.0, and less than or equal to 1.0;
The quality coefficient is multiplied by second mass, and the result of acquisition is defined as to the quality of described image.
Said apparatus can be used for performing the method that above-mentioned corresponding method embodiment provides, specific implementation and technique effect
It is similar, repeat no more here.
It should be noted that it should be understood that the division of the unit of the apparatus for evaluating of above video quality is only one kind patrols
The division of function is collected, can be completely or partially integrated on a physical entity when actually realizing, can also be physically separate.And
These units can be realized all in the form of software is called by treatment element;All it can also realize in the form of hardware;
It can be realized in the form of unit is called by software by treatment element, unit is realized by the form of hardware.
For example, acquiring unit can be the treatment element individually set up, it can also be integrated in some chip of server and realize, this
Outside, it can also be stored in the form of program in the memory of server, be called and held by some treatment element of server
The function of the row acquiring unit.The realization of other units is similar therewith.In addition these units can completely or partially be integrated in one
Rise, can also independently realize.Computing unit described here can be a kind of integrated circuit, have the disposal ability of signal.
In implementation process, each step of the above method or more unit can pass through the integration logic of the hardware in processor elements
The instruction of circuit or software form is completed.
The above unit can be arranged to implement one or more integrated circuits of above method, such as:One
Or multiple specific integrated circuits (Application Specific Integrated Circuit, ASIC), or, one or more
Individual microprocessor (digital signal processor, DSP), or, one or more field programmable gate array
(Field Programmable Gate Array, FPGA) etc..For another example, some unit dispatches journey by treatment element more than
When the form of sequence is realized, the computing unit can be general processor, such as central processing unit (Central Processing
Unit, CPU) or it is other can be with the processor of caller.For another example, these units can integrate, with on-chip system
The form of (system-on-a-chip, SOC) is realized.
Fig. 7 A are a kind of possible structural representation of the application server.Referring to shown in Fig. 7 A, the server 700 wraps
Include:Processing unit 702 and communication unit 703.The action that processing unit 702 is used for server 700 is controlled management, example
Such as, processing unit 702 be used for support server 700 perform aforementioned video quality appraisal procedure embodiment in each step and/
Or other processes for technology described herein.Communication unit 703 is used to support server 700 and other network entities
Communication, such as the communication between terminal device.Server 700 can also include memory cell 701, for storage server
700 program code and data.
Wherein, processing unit 702 can be processor or controller, such as can be central processing unit (Central
Processing Unit, CPU), general processor, digital signal processor (Digital Signal Processor, DSP),
Application specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array
It is (Field Programmable Gate Array, FPGA) or other PLDs, transistor logic, hard
Part part or its any combination.What it can realize or perform with reference to described by present disclosure various exemplary patrols
Collect square frame, module and circuit.The processor can also be the combination for realizing computing function, such as include one or more micro- places
Manage device combination, combination of DSP and microprocessor etc..Communication unit 703 can be communication interface, transceiver, transmission circuit etc.,
Wherein, communication interface is to be referred to as, and can include one or more interfaces.Memory cell 701 can be memory.
When processing unit 702 is processor, communication unit 703 is communication interface, when memory cell 701 is memory, this
The involved server of application can be the server shown in Fig. 7 B.
Referring to shown in Fig. 7 B, the server 710 includes:Processor 712, communication interface 713, memory 711.Optionally,
Server 710 can also include bus 714.Wherein, communication interface 713, processor 712 and memory 711 can be by total
Line 714 is connected with each other;Bus 714 can be Peripheral Component Interconnect standard (Peripheral Component
Interconnect;PCI) bus or EISA (Extended Industry Standard
Architecture;EISA) bus etc..The bus 714 can be divided into address bus, data/address bus, controlling bus etc..For just
Only represented in expression, Fig. 7 B with a thick line, it is not intended that an only bus or a type of bus.
Wherein, memory 711 is used to store the instruction that can be performed by processor 712, and processor 712 is used to call memory
The instruction stored in 711, to perform each step in the appraisal procedure of aforementioned video quality.
The application also provides a kind of assessment system of video quality, including the as above video quality described in any embodiment
Apparatus for evaluating.
The application also provides a kind of storage medium, including:Readable storage medium storing program for executing and computer program, the computer program
The appraisal procedure of the video quality provided for realizing foregoing any embodiment.
The application also provides a kind of program product, and the program product includes computer program (i.e. execute instruction), the calculating
Machine program storage is in readable storage medium storing program for executing.At least one processor of server can read the calculating from readable storage medium storing program for executing
Machine program, at least one computing device computer program cause the video that the foregoing various embodiments of server implementation provide
The appraisal procedure of quality.
The embodiment of the present application additionally provides a kind of apparatus for evaluating of video quality, including at least one memory element and at least
One treatment element, at least one memory element are used for storage program, when the program is performed so that the video quality
Apparatus for evaluating perform the operation of the server in any of the above-described embodiment.
Realizing all or part of step of above-mentioned each method embodiment can be completed by the related hardware of programmed instruction.
Foregoing program can be stored in a readable access to memory.Upon execution, execution includes above-mentioned each method embodiment to the program
The step of;And foregoing memory (storage medium) includes:Read-only storage (read-only memory, ROM), RAM, quick flashing
Memory, hard disk, solid state hard disc, tape (magnetic tape), floppy disk (floppy disk), CD (optical disc)
And its any combination.
Claims (12)
- A kind of 1. appraisal procedure of video quality, it is characterised in that including:Obtain training dataset, the training data, which is concentrated, includes at least one source images and corresponding extremely with each source images A few distorted image;Calculate the first mass of each source images;Calculate second mass of at least one distorted image relative to the source images;According to first mass and second mass, the quality of image is determined.
- 2. according to the method for claim 1, it is characterised in that described to calculate at least one distorted image relative to institute The second mass of source images is stated, including:Determine the type of distortion and distortion level of each distorted image;According to the type of distortion and the distortion level, each distorted image is grouped, obtains at least one distorted image Group, wherein, the type of distortion and distortion level all same of the distorted image in same distorted image group;According to multiple first assessment results and the second assessment result, have from m in reference picture method for evaluating quality and determine wherein N have reference picture method for evaluating quality;First assessment result has reference picture quality to comment to be respectively adopted the m Estimate method, the result obtained after assessing the distorted image in each distorted image group, second assessment result is The result that user's subjectivity obtains after assessing each distorted image, m and n are positive integer, and m is more than or equal to n;There is reference picture method for evaluating quality according to the n, calculate second mass.
- 3. according to the method for claim 2, it is characterised in that described assessed according to multiple first assessment results and second is tied Fruit, have from m and determine that n therein have reference picture method for evaluating quality in reference picture method for evaluating quality, including:There is reference picture method for evaluating quality by m, the distorted image in each distorted image group assessed respectively, Obtain multiple first assessment results;In each distorted image group, according to multiple first assessment results, there is reference picture quality evaluation side from the m T therein are determined in method reference picture method for evaluating quality;It is according to m that the t, which have reference picture method for evaluating quality, The individual uniformity having corresponding to reference picture method for evaluating quality between first assessment result and second assessment result The order of degree from high to low, there is reference picture method for evaluating quality corresponding to the preceding t degree of consistency selected;Being had according to the t corresponding to all distorted image groups in reference picture method for evaluating quality each has reference picture The order of the frequency that method for evaluating quality occurs from high to low, n have reference picture method for evaluating quality before determining;Wherein, t For the positive integer less than or equal to m, and more than or equal to n.
- 4. according to the method in claim 2 or 3, it is characterised in that described to have reference picture quality evaluation according to the n Method, second mass is calculated, including:Respectively by the n quality for thering is reference picture method for evaluating quality to calculate the distorted image, multiple 3rd mass are obtained;The multiple 3rd mass is weighted processing by machine learning, obtains second mass.
- 5. according to the method described in claim any one of 1-4, it is characterised in that first matter for calculating each source images Amount, including:According to the pixel domain information of each source images, first mass is calculated.
- 6. according to the method described in claim any one of 1-5, it is characterised in that described according to first mass and described Two mass, the quality of image is determined, including:First mass is subjected to standardization processing, obtains quality coefficient, the value of the quality coefficient is more than or equal to 0.0, and Less than or equal to 1.0;The quality coefficient is multiplied by second mass, and the result of acquisition is defined as to the quality of described image.
- A kind of 7. apparatus for evaluating of video quality, it is characterised in that including:Acquiring unit, for obtaining training dataset, the training data concentrate include at least one source images and with it is each described At least one distorted image corresponding to source images;Computing unit, for calculating the first mass of each source images;The computing unit, it is additionally operable to calculate second mass of at least one distorted image relative to the source images;Determining unit, for according to first mass and second mass, determining the quality of image.
- 8. device according to claim 7, it is characterised in that the computing unit, including:Determination subelement, for determining the type of distortion and distortion level of each distorted image;Subelement is grouped, for according to the type of distortion and the distortion level, being grouped, being obtained extremely to each distorted image A few distorted image group, wherein, the type of distortion and distortion level of the distorted image in same distorted image group are homogeneous Together;The determination subelement, it is additionally operable to according to multiple first assessment results and the second assessment result, has reference picture matter from m Determine that n therein have reference picture method for evaluating quality in amount appraisal procedure;First assessment result is is respectively adopted Stating m has reference picture method for evaluating quality, the knot obtained after assessing the distorted image in each distorted image group Fruit, second assessment result are the result obtained after user's subjectivity is assessed each distorted image, and m and n are just whole Number, and m is more than or equal to n;Computation subunit, for having reference picture method for evaluating quality according to the n, calculate second mass.
- 9. device according to claim 8, it is characterised in that the determination subelement, be specifically used for:There is reference picture method for evaluating quality by m, the distorted image in each distorted image group assessed respectively, Obtain multiple first assessment results;In each distorted image group, according to multiple first assessment results, there is reference picture quality evaluation side from the m T therein are determined in method reference picture method for evaluating quality;It is according to m that the t, which have reference picture method for evaluating quality, The individual uniformity having corresponding to reference picture method for evaluating quality between first assessment result and second assessment result The order of degree from high to low, there is reference picture method for evaluating quality corresponding to the preceding t similarity selected;Being had according to the t corresponding to all distorted image groups in reference picture method for evaluating quality each has reference picture The order of the frequency that method for evaluating quality occurs from high to low, n have reference picture method for evaluating quality before determining;Wherein, t For the positive integer less than or equal to m, and more than or equal to n.
- 10. device according to claim 8 or claim 9, it is characterised in that the computation subunit, be specifically used for:Respectively by the n quality for thering is reference picture method for evaluating quality to calculate the distorted image, multiple 3rd mass are obtained;The multiple 3rd mass is weighted processing by machine learning, obtains second mass.
- 11. according to the device described in claim any one of 7-10, it is characterised in that the computing unit, be specifically used for:According to the pixel domain information of each source images, first mass is calculated.
- 12. according to the device described in claim any one of 7-11, it is characterised in that the determining unit is specifically used for:First mass is subjected to standardization processing, obtains quality coefficient, the value of the quality coefficient is more than or equal to 0.0, and Less than or equal to 1.0;The quality coefficient is multiplied by second mass, and the result of acquisition is defined as to the quality of described image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710725869.6A CN107659806B (en) | 2017-08-22 | 2017-08-22 | The appraisal procedure and device of video quality |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710725869.6A CN107659806B (en) | 2017-08-22 | 2017-08-22 | The appraisal procedure and device of video quality |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107659806A true CN107659806A (en) | 2018-02-02 |
CN107659806B CN107659806B (en) | 2019-08-16 |
Family
ID=61128606
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710725869.6A Active CN107659806B (en) | 2017-08-22 | 2017-08-22 | The appraisal procedure and device of video quality |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107659806B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110290381A (en) * | 2019-08-01 | 2019-09-27 | 字节跳动(香港)有限公司 | Video quality evaluation method, device, electronic equipment and computer storage medium |
CN111062914A (en) * | 2019-11-28 | 2020-04-24 | 重庆中星微人工智能芯片技术有限公司 | Method, apparatus, electronic device and computer readable medium for acquiring facial image |
CN111163338A (en) * | 2019-12-27 | 2020-05-15 | 广州市百果园网络科技有限公司 | Video definition evaluation model training method, video recommendation method and related device |
CN111935545A (en) * | 2020-08-03 | 2020-11-13 | 腾讯音乐娱乐科技(深圳)有限公司 | Method, device and equipment for transcoding video data and storage medium |
CN112616051A (en) * | 2020-12-23 | 2021-04-06 | 北京金山云网络技术有限公司 | Video quality comparison method and device, storage medium and electronic equipment |
CN113658093A (en) * | 2021-06-30 | 2021-11-16 | 浙江大华技术股份有限公司 | Image quality evaluation method, device, equipment and computer storage medium |
CN113748683A (en) * | 2019-05-12 | 2021-12-03 | 脸谱公司 | System and method for preserving in-band metadata in compressed video files |
CN114627085A (en) * | 2022-03-17 | 2022-06-14 | 腾讯科技(深圳)有限公司 | Target image identification method and device, storage medium and electronic equipment |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1727375A1 (en) * | 2005-05-27 | 2006-11-29 | Psytechnics Limited | Assessment of perceived quality of a packetized video stream |
CN101635846A (en) * | 2008-07-21 | 2010-01-27 | 华为技术有限公司 | Method, system and device for evaluating video quality |
CN102202227A (en) * | 2011-06-21 | 2011-09-28 | 珠海世纪鼎利通信科技股份有限公司 | No-reference objective video quality assessment method |
CN102204265A (en) * | 2010-12-31 | 2011-09-28 | 华为技术有限公司 | Method, terminal, server, and system for evaluating video quality |
JP2011205253A (en) * | 2010-03-24 | 2011-10-13 | Nippon Telegr & Teleph Corp <Ntt> | Device, method, and program for transcoded video quality objective evaluation |
CN103338379A (en) * | 2013-06-05 | 2013-10-02 | 宁波大学 | Stereoscopic video objective quality evaluation method based on machine learning |
CN103379360A (en) * | 2012-04-23 | 2013-10-30 | 华为技术有限公司 | Assessment method and device for video quality |
CN106412571A (en) * | 2016-10-12 | 2017-02-15 | 天津大学 | Video quality evaluation method based on gradient similarity standard deviation |
-
2017
- 2017-08-22 CN CN201710725869.6A patent/CN107659806B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1727375A1 (en) * | 2005-05-27 | 2006-11-29 | Psytechnics Limited | Assessment of perceived quality of a packetized video stream |
CN101635846A (en) * | 2008-07-21 | 2010-01-27 | 华为技术有限公司 | Method, system and device for evaluating video quality |
JP2011205253A (en) * | 2010-03-24 | 2011-10-13 | Nippon Telegr & Teleph Corp <Ntt> | Device, method, and program for transcoded video quality objective evaluation |
CN102204265A (en) * | 2010-12-31 | 2011-09-28 | 华为技术有限公司 | Method, terminal, server, and system for evaluating video quality |
CN102202227A (en) * | 2011-06-21 | 2011-09-28 | 珠海世纪鼎利通信科技股份有限公司 | No-reference objective video quality assessment method |
CN103379360A (en) * | 2012-04-23 | 2013-10-30 | 华为技术有限公司 | Assessment method and device for video quality |
CN103338379A (en) * | 2013-06-05 | 2013-10-02 | 宁波大学 | Stereoscopic video objective quality evaluation method based on machine learning |
CN106412571A (en) * | 2016-10-12 | 2017-02-15 | 天津大学 | Video quality evaluation method based on gradient similarity standard deviation |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113748683A (en) * | 2019-05-12 | 2021-12-03 | 脸谱公司 | System and method for preserving in-band metadata in compressed video files |
CN110290381A (en) * | 2019-08-01 | 2019-09-27 | 字节跳动(香港)有限公司 | Video quality evaluation method, device, electronic equipment and computer storage medium |
CN111062914A (en) * | 2019-11-28 | 2020-04-24 | 重庆中星微人工智能芯片技术有限公司 | Method, apparatus, electronic device and computer readable medium for acquiring facial image |
CN111062914B (en) * | 2019-11-28 | 2023-05-30 | 重庆中星微人工智能芯片技术有限公司 | Method, apparatus, electronic device and computer readable medium for acquiring facial image |
CN111163338A (en) * | 2019-12-27 | 2020-05-15 | 广州市百果园网络科技有限公司 | Video definition evaluation model training method, video recommendation method and related device |
CN111935545A (en) * | 2020-08-03 | 2020-11-13 | 腾讯音乐娱乐科技(深圳)有限公司 | Method, device and equipment for transcoding video data and storage medium |
CN112616051A (en) * | 2020-12-23 | 2021-04-06 | 北京金山云网络技术有限公司 | Video quality comparison method and device, storage medium and electronic equipment |
CN112616051B (en) * | 2020-12-23 | 2023-05-05 | 北京金山云网络技术有限公司 | Video quality comparison method and device, storage medium and electronic equipment |
CN113658093A (en) * | 2021-06-30 | 2021-11-16 | 浙江大华技术股份有限公司 | Image quality evaluation method, device, equipment and computer storage medium |
CN114627085A (en) * | 2022-03-17 | 2022-06-14 | 腾讯科技(深圳)有限公司 | Target image identification method and device, storage medium and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN107659806B (en) | 2019-08-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107659806B (en) | The appraisal procedure and device of video quality | |
US20070103551A1 (en) | Method and system for measuring video quality | |
CN108537289B (en) | Training method and device of data recognition model and storage medium | |
EP1924101B1 (en) | Video communication quality estimation device, method, and program | |
CN100389437C (en) | Method and apparatus for modeling film grain patterns in the frequency domain | |
US7764844B2 (en) | Determining sharpness predictors for a digital image | |
Hall et al. | Hepatic steatosis estimated microscopically versus digital image analysis | |
CN102917183A (en) | Imaging device, image processing method and program | |
JP7046073B2 (en) | Real-time data processing method and equipment | |
CN103067671B (en) | A kind of method and device showing image | |
CN105100784A (en) | Smoothness test method and device | |
Redi et al. | Comparing subjective image quality measurement methods for the creation of public databases | |
CN109783486A (en) | Data cleaning method, device and server | |
CN110348519A (en) | Financial product cheats recognition methods and the device of clique | |
CN106096047A (en) | User partition preference computational methods based on Information Entropy and system | |
CN109146271A (en) | Performance management method, apparatus, terminal and computer readable storage medium | |
CN101616315A (en) | A kind of method for evaluating video quality, device and system | |
Rousselot et al. | Quality metric aggregation for HDR/WCG images | |
CN104103065A (en) | No-reference fuzzy image quality evaluation method based on singular value decomposition | |
CN114064445A (en) | Test method, device, equipment and computer readable storage medium | |
Barkowsky et al. | Subjective experiment dataset for joint development of hybrid video quality measurement algorithms | |
Choudhury et al. | HDR image quality assessment using machine-learning based combination of quality metrics | |
US11255705B2 (en) | Method for setting quality thresholds of products for testing purposes and device employing method | |
Barkowsky et al. | Perceptually motivated spatial and temporal integration of pixel based video quality measures | |
CN101895787A (en) | Method and system for subjectively evaluating video coding performance |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220208 Address after: 550025 Huawei cloud data center, jiaoxinggong Road, Qianzhong Avenue, Gui'an New District, Guiyang City, Guizhou Province Patentee after: Huawei Cloud Computing Technologies Co.,Ltd. Address before: 518129 Bantian HUAWEI headquarters office building, Longgang District, Guangdong, Shenzhen Patentee before: HUAWEI TECHNOLOGIES Co.,Ltd. |
|
TR01 | Transfer of patent right |