CN105100789A - Method for evaluating video quality - Google Patents
Method for evaluating video quality Download PDFInfo
- Publication number
- CN105100789A CN105100789A CN201510435863.6A CN201510435863A CN105100789A CN 105100789 A CN105100789 A CN 105100789A CN 201510435863 A CN201510435863 A CN 201510435863A CN 105100789 A CN105100789 A CN 105100789A
- Authority
- CN
- China
- Prior art keywords
- video
- video quality
- build
- evaluation
- result
- 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 invention relates to a method for evaluating video quality. The method includes a first step of subjectively evaluating the video quality in nine scoring level, calculating average subjective scores of each scene among a plurality of video scenes, and finishing subjective evaluation of video quality to obtain a subjectively evaluating result; a second step of carrying out objective evaluating pretreatment on video quality, including peak value signal-to-noise ratio processing, edge structure similarity processing, sharpness processing and spatial-time domain value processing; a third step of objectively evaluating the video quality to obtain an objective evaluating result; and a fourth step of comprehensively evaluating the video quality. In the fourth step, subjective evaluating result and the objective evaluating result are combined to obtain a comprehensive result through comprehensive calculation. The accuracy of the video evaluation result is effectively improved, and high-efficient evaluation is realized.
Description
Technical field
The present invention is the method for evaluating video quality being applied to mobile terminal, video mainly for the panel computer in Android system is analyzed, by video quality evaluation method, subjective assessment and objective evaluation are carried out to video, video quality is efficiently assessed, thus can control and improve the quality of digital video, better for mobile terminal business men provides guidance foundation.
Background technology
Nowadays at video from simulated to digitized in development, various difference can be produced in the process of video sampling, compression, transmission and reconstruction, cause video distortion in various degree.Evaluation majority at present for video quality is that Based PC launches, and it is relatively less to the research field based on mobile terminal, how efficiently to assess based on mobile terminal video quality, it is the emphasis of people's future studies, not only can provide guidance foundation to mobile terminal business men, different video fields can also be served.
Video quality subjective assessment aspect, because test process affects by various factors, method of testing does not have versatility, and therefore a set of general standard test system becomes a research direction of subjective assessment.
Video quality objective assessment aspect, sets up evaluation model by finding suitable evaluation index, but human eye is a complicated system, should fully further investigate in conjunction with human visual system in the process of carrying out objective evaluation.For the research of the subjective and objective overall merit of video quality, not also very ripe at present, especially less in the video evaluation system research of mobile terminal.
Summary of the invention
A set of video quality overall evaluation system that the present invention invents to solve the aforementioned problems in the prior just, first the video quality evaluation theory that Based PC is studied is transferred to the mobile terminals such as panel computer, design the appraisement system system of the panel computer based on Android system.Construct a set of video quality overall evaluation system.Evaluation system forms primarily of two parts, the subjective assessment of video quality and the objective evaluation of video quality.
Objective evaluation, have selected four objective evaluation indexs: Y-PSNR, marginal texture similitude, definition, Space-time domain model, both the feature of human visual system had been considered, reacted again sdi video information and temporal information, the selection of multi objective solves inaccuracy and the limitation of single index.The present invention calculates the weight of each index by adopting entropy power enabling legislation, and introduces effect function and be normalized data, is sorted, complete the evaluation of video quality by compute euclidian distances to video quality.With subjective results, there is good consistency after overall merit.
Subjective assessment, that adopt is the absolute scale point system ACR-HRR of the removal implicit reference defined in ITU-TP.910, test environment is arranged in strict accordance with the carrying out provided in standard, what cycle tests was selected is the video that VQEG organizes 5 scenes provided, each scene comprises 1 original video and 16 distortion videos, organize several observer to carry out subjective assessment, and evaluation result has been processed, obtained mean subjective suggestion and divide.Evaluation result shows, and the trend trend of each scene graph is consistent, ensure that the accuracy of experimental result.
In order to realize foregoing invention object, by the following technical solutions: a kind of method for evaluating video quality, it comprises:
Step 1, video quality subjective assessment process, it adopts nine grades of grading systems processed to evaluate, and the mean subjective suggestion calculating each scene in multiple video scene is divided, and completes video quality subjective assessment, obtains subjective evaluation result;
Step 2, video quality objective assessment preliminary treatment, it comprises Y-PSNR process, the process of marginal texture similitude, definition process, the process of Space-time domain value;
Described Y-PSNR process, it passes through formula:
calculate Y-PSNR PSNR, wherein, wherein x
iwith
represent the pixel value in original video and distortion frame of video, M, N are the length of video and wide;
The process of described marginal texture similitude, it is based on luminance video, contrast and structural similarity three calculation of parameter marginal texture similitude ESSIM;
Described definition process, it passes through formula:
calculate definition, wherein, m, n are that video is long and wide, and df is the amplitude of variation of luminance video, and dx is the distance increment between video frame pixel;
The process of described Space-time domain value, it is based on the motility index calculate Space-time domain index IMB in the ambiguity in spatial domain, blocking effect and time domain;
Step 3, video quality objective assessment process, it specifically comprises:
Step a1, Evaluations matrix build, and it is to the original data processing comprising Y-PSNR, marginal texture similitude, definition, Space-time domain value, to build Evaluations matrix;
Step a2, normalized matrix build, and it uses utility function to carry out standardization conversion, to build normalized matrix to the Evaluations matrix comprising initial data;
Step a3, weight calculation, it carries out weight calculation to each standardized index in normalized matrix, completes the calculating of evaluation index entropy and the calculating of evaluation index entropy power;
Step a4, objective evaluation data processing, its by initial data Evaluations matrix build, normalized matrix build and weight calculation result process so that obtain objective evaluation result;
Step 4, the process of video quality overall merit, it, in conjunction with subjective evaluation result and objective evaluation result, obtains synthesis result by COMPREHENSIVE CALCULATING.
Preferably, video quality overall merit treatment step comprises further:
Step b1, reception subjective evaluation result and objective evaluation result are as the second initial data;
Step b2, the second initial data Evaluations matrix build, and it is to the second original data processing comprising subjective evaluation result and objective evaluation result, to build the second Evaluations matrix;
Step b3, the second normalized matrix build, and it uses the second utility function to carry out standardization conversion to the second Evaluations matrix comprising the second initial data, to build the second normalized matrix;
Step b4, the second weight calculation, it carries out weight calculation to each standardized index in the second normalized matrix, completes the calculating of the second evaluation index entropy and the calculating of the second evaluation index entropy power;
Step b5, the process of video quality overall merit, its by second initial data Evaluations matrix build, second normalized matrix build and the second weight calculation result process so that obtain synthesis result.
Method for evaluating video quality provided by the invention, objective evaluation has selected four index algorithms, Y-PSNR, marginal texture similitude, definition, Space-time domain algorithm; Subjective objective evaluation is all by Java language programming realization, sets up overall evaluation system.After realization, evaluating system is divided into two-layer, the raw data matrix of experimental data is converted, uses utility function to matrix standardization, calculate entropy and the entropy power of each index, objective results is obtained finally by Euclidean distance, mean subjective suggestion further combined with scene is divided and objective evaluation result, is finally obtained the synthesis result of all video sequences by computational methods, completes the comprehensive evaluation result of video quality, it ensures the accuracy of video evaluation result further, achieves efficient assessment.
Accompanying drawing explanation
Fig. 1 is video quality overall merit flow chart of the present invention.
Fig. 2 is video quality overall merit data handling system figure of the present invention.
Fig. 3 is marginal texture similitude ESSIM Method And Principle block diagram.
Fig. 4 is definition algorithm schematic diagram.
Fig. 5 is adjacent horizontal pixels block schematic diagram.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in further details.
Specific implementation process can be: first build based on Android development environment, video player is made based on Androidyuv by calling SDL third party library, carry out the design and implimentation of evaluating system, the interface of design video quality subjective assessment, comprise logging in system by user and database real-time update, scene selects autonomous interface random, video playback interface, the restriction of scoring function.And objective evaluation only allows keeper to log in, therefore be provided with keeper's authentication, need authentic administrator information, with in database during information matches, enter objective evaluation system, otherwise cannot enter.Keeper can check the data of subjective assessment, carries out objective evaluation and checks objective computation result.
Concrete use procedure can be: the panel computer based on Android carries out carrying out overall merit to many group videos, first from Android panel computer, opens application.Carry out subjective assessment test, require that at least 15 amateur testers participate in, pass through register account number, this user logs in and enters into subjective assessment interface, different video according to many group scenes carries out subjective scoring, only terminate in video playback, when there is gray scale picture, just can give a mark, require that scoring is between 0 ~ 9, after scoring terminates, next video of manual selection is play, repeat said process, until all video sequences all evaluate end, can log off, observer provides all scorings, scoring will be updated in database automatically, thus complete the subjective assessment of video quality.Carry out objective evaluation only open to keeper, therefore be provided with keeper's authentication, need authentic administrator information, with in database during information matches, enter objective evaluation system.Keeper can check the data of subjective assessment, also can check objective computation result simultaneously.Y-PSNR in objective results, marginal texture similitude, definition three indexs, be all that result of calculation is larger, the mass discrepancy representing distortion video and original video is less, and the quality of video is better, belongs to profit evaluation model index.Space-time domain value is then contrary, and its value is less, and the mass discrepancy representing distortion video and original video is less, and the quality of video is better, belongs to cost type index.
See Fig. 1, the detailed process of video quality overall merit test can be: video to be tested is loaded in list of videos, issue this program in the panel computer of Android system, open software and enter main interface, enter login interface, keeper is divided into log in, tester logs in, user logs in and is tester's login, tester logs in can enter into subjective assessment system, at least 15 amateur observers test, nine grades of grading systems processed are adopted to carry out evaluation marking to all videos, integrate the total data result of subjective assessment, its data need to be logged in by keeper just can show, keeper logs in the login interface being the worker carrying out video evaluation analysis, keeper logs in and enters objective evaluation system, video is carried out to the objective analysis of automation, according to marginal texture similitude, Y-PSNR, definition, four aspect indexs of Space-time domain algorithm carry out algorithm scoring, by the objective analysis of system, obtain the objective experimental data of video sequence, obtain the data result of four indexs, according to certain algorithm and the relative weighting according to four indexs, the objective results of all video sequences of final calculating, comprehensive evaluation result, mean subjective suggestion in conjunction with scene is divided and objective evaluation result, the synthesis result of all video sequences is finally obtained by computational methods.
Wherein, method for evaluating video quality can be specially:
Step 1, video quality subjective assessment process, it adopts nine grades of grading systems processed to evaluate, and the mean subjective suggestion calculating each scene in multiple video scene is divided, and completes video quality subjective assessment, obtains subjective evaluation result;
Step 2, video quality objective assessment preliminary treatment, it comprises Y-PSNR process, the process of marginal texture similitude, definition process, the process of Space-time domain value;
Described Y-PSNR process, it passes through formula:
calculate Y-PSNR PSNR, wherein, wherein x
iwith
represent the pixel value in original video and distortion frame of video, M, N are the length of video and wide;
The process of described marginal texture similitude, it is based on luminance video, contrast and structural similarity three calculation of parameter marginal texture similitude ESSIM;
Described definition process, it passes through formula:
calculate definition, wherein, m, n are that video is long and wide, and df is the amplitude of variation of luminance video, and dx is the distance increment between video frame pixel;
The process of described Space-time domain value, it is based on the motility index calculate Space-time domain index IMB in the ambiguity in spatial domain, blocking effect and time domain;
Step 3, video quality objective assessment process, it specifically comprises:
Step a1, Evaluations matrix build, and it is to the original data processing comprising Y-PSNR, marginal texture similitude, definition, Space-time domain value, to build Evaluations matrix;
Step a2, normalized matrix build, and it uses utility function to carry out standardization conversion, to build normalized matrix to the Evaluations matrix comprising initial data;
Step a3, weight calculation, it carries out weight calculation to each standardized index in normalized matrix, completes the calculating of evaluation index entropy and the calculating of evaluation index entropy power;
Step a4, objective evaluation data processing, its by initial data Evaluations matrix build, normalized matrix build and weight calculation result process so that obtain objective evaluation result;
Step 4, the process of video quality overall merit, it, in conjunction with subjective evaluation result and objective evaluation result, obtains synthesis result by COMPREHENSIVE CALCULATING.
Preferably, video quality overall merit treatment step comprises further:
Step b1, reception subjective evaluation result and objective evaluation result are as the second initial data;
Step b2, the second initial data Evaluations matrix build, and it is to the second original data processing comprising subjective evaluation result and objective evaluation result, to build the second Evaluations matrix;
Step b3, the second normalized matrix build, and it uses the second utility function to carry out standardization conversion to the second Evaluations matrix comprising the second initial data, to build the second normalized matrix;
Step b4, the second weight calculation, it carries out weight calculation to each standardized index in the second normalized matrix, completes the calculating of the second evaluation index entropy and the calculating of the second evaluation index entropy power;
Step b5, the process of video quality overall merit, its by second initial data Evaluations matrix build, second normalized matrix build and the second weight calculation result process so that obtain synthesis result.
See Fig. 2, video quality evaluation is made the following instructions further:
1, concrete data handling procedure, it comprises objective evaluation data processing and overall merit data processing, and wherein, objective evaluation data processing comprises:
1.1, according to the result of objective evaluation four indexs, for the objective experimental data of video sequence build Evaluations matrix.This layer mainly to the calculating of video quality four index weights, and calculates the objective results of each sequence.The initial data Evaluations matrix R=(r of 4 evaluation indexes of each scene n different video
ij)
n × 4for:
1.2, r in matrix R
ijrepresent a jth evaluation index of i-th video sequence, evaluation index is divided three classes: cost type, profit evaluation model, interval type, there is different dimensions for dissimilar index, need evaluation index standardization before evaluation, make it can drop on a certain dimensionless interval.This problem uses utility function formula to carry out standardization to initial data, the mean value of a note jth evaluation index
such as formula:
For cost type index middle transition variable such as formula
for profit evaluation model index middle transition variable such as formula
transition matrix A is obtained as follows after initial data desired value is processed:
Below by the utility function standardization of transition matrix data acquisition, such as formula:
After initial data is processed, finally obtain initial data normalized matrix B:
Obviously, b
ij=f (a
ij) function is a S type curve, a
ijthat react is initial data r
ijdeviation average
degree, for profit evaluation model index: when
time, utility function b after standardization
ij>=0; For the indicator of costs: when
time, utility function b after standardization
ij≤ 0.Initial data r
ijlarger, utility function b
ijlarger, as initial data r
ijwhen increasing to a certain degree, utility function b
ijwill close to " saturated ".
1.3, to after the standardization of evaluation index initial data, need to carry out weight calculation to each standardized index, entropy power enabling legislation agriculture products weight, is mainly divided into two processes: the calculating of evaluation index entropy, the calculating of evaluation index entropy power.Evaluate in the problem of sequence n objective evaluation index at m, the entropy H of a jth evaluation index
jdefine such as formula:
Wherein:
and suppose: work as f
ijwhen=0, make f
ijlnf
ij=0.The entropy power definition ω of a definition jth index
jfor:
Wherein, 0≤ω
j≤ 1,
the entropy of evaluation index is larger, and its entropy power will be less, illustrates that this index is more inessential.Quote Zadeh
[75]definition is by L
λthe overall merit of (i) objective indicator the most, such as formula:
Under normal circumstances, L during λ=1
1be called Hamming distances, the summation of the deviation of more emphasis; L during λ=2
2be called Euclidean distance, more emphasis be the situation that individual deviation is larger.In this problem, adopt L
2result as overall performane calculates, such as formula
euclidean distance is less, and video quality is better.
Overall merit data processing: the calculating of this layer data is the same with thinking when calculating objective evaluation data processing, and computing formula is also with reference to objective evaluation data processing.
1.4, after calculating second layer objective evaluation index result, with subjective evaluation result as ground floor, form raw data matrix R as follows, calculate video quality evaluation result.
1.5, utility function formula is used to carry out standardization to initial data, the mean value of a note jth evaluation index
such as formula:
for cost type index middle transition variable such as formula:
for profit evaluation model index middle transition variable such as formula
transition matrix A is obtained as follows after initial data desired value is processed:
Below by the utility function standardization of transition matrix data acquisition, such as formula:
After initial data is processed, finally obtain initial data normalized matrix B:
1.6, evaluate in the problem of sequence n objective evaluation index at m, the entropy H of a jth evaluation index
jdefine such as formula:
Wherein:
and suppose: work as f
ijwhen=0, make f
ijlnf
ij=0.The entropy power definition ω of a definition jth index
jfor:
Wherein, 0≤ω
j≤ 1,
the entropy of evaluation index is larger, and its entropy power will be less, illustrates that this index is more inessential.Quote Zadeh
[75]definition is by L
λthe overall merit of (i) objective indicator the most, such as formula:
Adopt L
2result as overall performane calculates, such as formula
euclidean distance is less, and video quality is better.Quality-ordered is carried out to the video sequence of each scene, completes the evaluation of video quality.
2, four parameter indexs of objective evaluation are:
(1) Y-PSNR PSNR
This evaluation index belongs to full reference model, and mean square error MSE and Y-PSNR PSNR, most widely used general, physical significance is simply clear.
MSE formula is as follows:
formula (2-1)
Wherein, f
ij, f
ij' representing the corresponding frame of original reference video sequence and distortion video sequence respectively, M, N represent the height and width of frame of video.
The calculating of PSNR is carried out based on independent pixel, and namely utilize the mean square error of reference video and each frame pixel value of distortion video to evaluate, formula is as follows
[71]:
formula (2-2)
Wherein x
iwith
represent the pixel value in original video and distortion frame of video.M, N are the length of video and wide.PSNR value is larger, and the difference of distortion video and original video is less.Under 8b sampling condition, signal peak amplitude is 255.This problem calculates the PSNR value of each pixel of every two field picture, and the mean value of all pixels is the PSNR value of frame, and the mean value of all frames is the PSNR value of distortion video sequence.
(2) marginal texture similitude ESSIM
Conventional structural similarity algorithm is based on brightness, contrast and structural similarity three calculation of parameter, for original video sequence X and distortion video sequence Y, brightness comparison function l (x, y) such as formula (2-3), contrast comparison function c (x, y) such as formula (2-4), structural similarity comparison function s (x, y) such as formula (2-5), COMPREHENSIVE CALCULATING result is such as formula (2-6).
Brightness ratio is comparatively:
formula (2-3)
Contrast compares:
formula (2-4)
Structural similarity compares:
formula (2-5)
SSIM=[(l (x, y)]
α[(c (x, y)]
β[s (x, y)]
γformula (2-6)
Wherein:
formula (2-7)
formula (2-8)
formula (2-9)
formula (2-10)
formula (2-11)
Wherein, μ
xand μ
ythe mean flow rate of original video sequence and distortion video sequence, δ
xand δ
ythe standard deviation of original video sequence and distortion video sequence, δ
xyit is the covariance of original video sequence and distortion video sequence.Traditional structural similarity calculates from statistically calculating based on image, and the correlation between pixel is poor, can not the subjective feeling of good appraiser.Consider the visual masking effect of human eye, the edge feature of image and SSIM are combined herein, form new evaluation index, based on the structural similarity index ESSIM (Edge-SSIM) of marginal information, principle as shown in Figure 3.Carry out Sobel rim detection to video image, the marginal information of compute gradient image is as the 4th parameter of structural similarity.Conventional detective operators
[71]detect, such as formula (2-12) and formula (2-13) from horizontal direction 0 ° and vertical direction 90 ° of both directions.In order to ensure all edges to be detected, the present invention adopts this 4 angle detecting operators, adds 45 ° of directions and 135 ° of directions, such as formula (2-14) and formula (2-15).Carry out 4 edge extractings to brightness, colourity according to gradient operator respectively, the G of definition synthesis is the edge intensity value computing in 4 directions, such as formula (2-16).
formula (2-12)
formula (2-13)
formula (2-14)
formula (2-15)
formula (2-16)
Calculate the edge intensity value computing G of gained, as the marginal information of image, the edge energy e (x, y) of edge calculation image, such as formula (2-17):
Edge energy:
formula (2-17)
Wherein, θ
xyrepresent the gray scale covariance of the edge image of original video sequence and distortion video sequence, θ
xand θ
yrepresent the gray variance of the edge image of original video sequence and distortion video sequence.
The synthesis result of innovatory algorithm is such as formula (2-18):
ESSIM (x, y)=[(l (x, y)]
α[(c (x, y)]
β[s (x, y)]
γ[e (x, y)]
κformula (2-18)
C
1, c
2, c
3, c
4be in order to avoid denominator is 0 constant established, all get 0 herein; α, beta, gamma, κ is weighed value adjusting, all gets 1.
So marginal texture Similarity measures formula of the present invention is such as formula (2-19):
formula (2-19)
The present invention calculates the ESSIM value of brightness Y, calculates the ESSIM value of each pixel of every two field picture, and the mean value of all pixels is the ESSIM value of frame, and the mean value of all frames is the ESSIM value of distortion video sequence.ESSIM value is larger, and the difference of distortion video and original video is less.
(3) definition PSV
This index just quantitatively calculates video definition, only carries out statistical computation to the gray scale spread condition around each pixel in distortion video sequence image frame, belongs to no reference model.Therefore this algorithmic formula is as follows:
formula (2-20)
Wherein, m, n are that video is long and wide, and df is the amplitude of variation of luminance video, and dx is the distance increment between video frame pixel.8 neighborhood points are got to each pixel in frame of video, carries out luminance difference calculating successively, and to 8 weighted differences summations.The size calculating weighting according to the distance of distance objective pixel is got, and distant weights are less.
As shown in Figure 4, in 8 neighborhood points, the pixel in 45 ° (3,6) and 135 ° of (1,8) directions is distant, and luminance difference need be divided by
the pixel close together in 0 ° (4,5) and 90 ° of (2,7) directions, weights are taken as 1.The luminance difference weighted sum of 8 neighborhood points is such as formula (2-21):
Formula (2-21)
This algorithm be to each pixel around the statistics of gray scale diffusion, the present invention calculates the PSV value of brightness Y, calculate the PSV value of each pixel of every two field picture, the mean value of all pixels is the PSV value of frame, and the mean value of all frames is the PSV value of distortion video sequence.PSV value is larger, and video diffusion Shaoxing opera is strong, and video image is also more clear, and the difference of distortion video and original video is less.
(4) Space-time domain index IMB
Video image in the transmission, except the spatial domain mass losses such as such as mosaic, fuzzy, noise, also have that picture jumps, the loss of time domain quality such as stagnate, the spatial information (si) and the time-domain information that consider video are the trend studying video quality evaluation, this problem emphasis considers the motility index in ambiguity, blocking effect and the time domain in spatial domain, as the 4th evaluation index.
Ambiguity, be the subjective details impression of human eye to video setup, video image ambiguity B (k) calculates such as formula (2-22), and through type (2-23) calculates the values of ambiguity of whole video, wherein N
kfor the pixel count of video image k, F is the frame number of video, f (x, y) for the gray value of pixel (x, y), S be maximum gray scale, values of ambiguity B is larger, and video is fuzzyyer.
formula (2-22)
formula (2-23)
Blocking effect, the i.e. non-continuous event of the block boundary of decoded picture.Current most compression algorithm is all adopt DCT algorithm spatial information (si) is transformed into frequency domain after image is divided into the block of pixels of 8 × 8
[73].Therefore, blocking effect detects an important spatial domain index of video image quality.First frame of video is divided into the block of pixels of 8 × 8, two adjacent horizontal pixel blocks as shown in Figure 5.
The luminance difference of adjacent two the block of pixels boundaries of calculated level, such as formula (2-24):
formula (2-24)
D
1(m, n)=a (m, n+1)-a (m, n) formula (2-25)
formula (2-26)
Wherein, the brightness value that a (m, n) is boundary pixel point, d
1(m, n) is the absolute luminance differences at adjacent two block boundary places, d
2(m, n) is adjacent two pieces of luminance difference averages near border.Use the same method and calculate the luminance difference D of two vertically adjacent block of pixels boundaries
h.Human visual system has shielding effect and non-linear to blocking effect, introduces law in weber-Fick
[74], the block of pixels effect function formula of frame of video can be obtained such as formula (2-27) and formula (2-28):
formula (2-27)
formula (2-28)
Definition, D
vor D
hwhen being 0, I
vand I
hbe 0.Wherein, k=1, L
0, α is constant, and value is 150 and 2, and threshold value T=0.02L, L are the average brightness value of frame of video background, I
vfor the blocking effect value of horizontal adjacent pixels block, I
hfor the blocking effect value of vertical adjacent pixels block.The blocking effect of frame of video is exactly try to achieve cumulative for the blocking effect of all horizontal directions and vertical direction, such as formula (2-29), frame of video blocking effect is averaging to the blocking effect value obtaining whole video, such as formula (2-30).
formula (2-29)
formula (2-30)
Wherein, N
pfor the pixel count of frame of video, M × N is the number of 8 × 8 block of pixels in frame of video, and F is the frame number of video.
Select motility index MA
[75](MotionActivity) describe the operation information of video, calculate the motility index of mean absolute difference MAD (MeanAbsoluteDifference) as video of adjacent two frame brightness, such as formula (2-31):
formula (2-31)
Wherein, N
kfor the pixel count of frame of video, L
k(x, y) is the gray value of pixel (x, y) in kth frame, frame of video MA value is averaging to the motility index result obtaining whole video, such as formula (2-32).
formula (2-32)
Comprehensive above three indexs, obtain the 4th index of video quality objective assessment, research shows, in the details of video image spatial domain, the impact of blocking effect on video image quality is larger
[76], therefore distribute more weight by blocking effect, such as formula (2-33).This value is larger, and video quality is poorer.
IMB=0.5I+0.3M+0.2B formula (2-33)
Above-described embodiment is only be described the preferred embodiment of the present invention; not scope of the present invention is limited; under not departing from the present invention and designing the prerequisite of spirit; the various distortion that the common engineers and technicians in this area make technical scheme of the present invention and improvement, all should fall in protection range that claims of the present invention determine.
Claims (2)
1. a method for evaluating video quality, is characterized in that:
Step 1, video quality subjective assessment process, it adopts nine grades of grading systems processed to evaluate, and the mean subjective suggestion calculating each scene in multiple video scene is divided, and completes video quality subjective assessment, obtains subjective evaluation result;
Step 2, video quality objective assessment preliminary treatment, it comprises Y-PSNR process, the process of marginal texture similitude, definition process, the process of Space-time domain value;
Described Y-PSNR process, it passes through formula:
Calculate Y-PSNR PSNR, wherein, wherein x
iwith
represent the pixel value in original video and distortion frame of video, M, N are the length of video and wide;
The process of described marginal texture similitude, it is based on luminance video, contrast and structural similarity three calculation of parameter marginal texture similitude ESSIM;
Described definition process, it passes through formula:
calculate definition, wherein, m, n are that video is long and wide, and df is the amplitude of variation of luminance video, and dx is the distance increment between video frame pixel;
The process of described Space-time domain value, it is based on the motility index calculate Space-time domain index IMB in the ambiguity in spatial domain, blocking effect and time domain;
Step 3, video quality objective assessment process, it specifically comprises:
Step a1, Evaluations matrix build, and it is to the original data processing comprising Y-PSNR, marginal texture similitude, definition, Space-time domain value, to build Evaluations matrix;
Step a2, normalized matrix build, and it uses utility function to carry out standardization conversion, to build normalized matrix to the Evaluations matrix comprising initial data;
Step a3, weight calculation, it carries out weight calculation to each standardized index in normalized matrix, completes the calculating of evaluation index entropy and the calculating of evaluation index entropy power;
Step a4, objective evaluation data processing, its by initial data Evaluations matrix build, normalized matrix build and weight calculation result process so that obtain objective evaluation result;
Step 4, the process of video quality overall merit, it, in conjunction with subjective evaluation result and objective evaluation result, obtains synthesis result by COMPREHENSIVE CALCULATING.
2. method according to claim 1, is characterized in that, video quality overall merit treatment step comprises further:
Step b1, reception subjective evaluation result and objective evaluation result are as the second initial data;
Step b2, the second initial data Evaluations matrix build, and it is to the second original data processing comprising subjective evaluation result and objective evaluation result, to build the second Evaluations matrix;
Step b3, the second normalized matrix build, and it uses the second utility function to carry out standardization conversion to the second Evaluations matrix comprising the second initial data, to build the second normalized matrix;
Step b4, the second weight calculation, it carries out weight calculation to each standardized index in the second normalized matrix, completes the calculating of the second evaluation index entropy and the calculating of the second evaluation index entropy power;
Step b5, the process of video quality overall merit, its by second initial data Evaluations matrix build, second normalized matrix build and the second weight calculation result process so that obtain synthesis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510435863.6A CN105100789B (en) | 2015-07-22 | 2015-07-22 | A kind of method for evaluating video quality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510435863.6A CN105100789B (en) | 2015-07-22 | 2015-07-22 | A kind of method for evaluating video quality |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105100789A true CN105100789A (en) | 2015-11-25 |
| CN105100789B CN105100789B (en) | 2018-05-15 |
Family
ID=54580199
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510435863.6A Active CN105100789B (en) | 2015-07-22 | 2015-07-22 | A kind of method for evaluating video quality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105100789B (en) |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105516713A (en) * | 2015-12-24 | 2016-04-20 | 招商局重庆交通科研设计院有限公司 | Image quality evaluation method of road traffic closed-circuit television based on machine vision |
| CN105635727A (en) * | 2015-12-29 | 2016-06-01 | 北京大学 | Subjective image quality evaluation method based on paired comparison and device thereof |
| CN106060537A (en) * | 2016-08-04 | 2016-10-26 | 腾讯科技（深圳）有限公司 | Method and device for assessing video quality |
| CN106803272A (en) * | 2017-01-13 | 2017-06-06 | 惠州Tcl移动通信有限公司 | A kind of camera evaluating method and system |
| CN107249127A (en) * | 2017-05-18 | 2017-10-13 | 深圳众厉电力科技有限公司 | One kind assesses accurate network video quality assessment system |
| CN107743226A (en) * | 2017-11-06 | 2018-02-27 | 潘柏霖 | One kind monitors accurate environmental monitoring system |
| CN107806863A (en) * | 2017-11-07 | 2018-03-16 | 潘柏霖 | A kind of unattended reservoir monitoring system |
| CN107832944A (en) * | 2017-11-07 | 2018-03-23 | 龚土婷 | A kind of natural hybridized orbit monitoring system |
| CN107833433A (en) * | 2017-11-06 | 2018-03-23 | 龚土婷 | A kind of accurate mountain area natural calamity early warning system of early warning |
| CN107862281A (en) * | 2017-11-06 | 2018-03-30 | 钟永松 | A kind of smart city planning system |
| CN107886234A (en) * | 2017-11-07 | 2018-04-06 | 钟永松 | A kind of efficient Urban Planning System |
| CN107948649A (en) * | 2016-10-12 | 2018-04-20 | 北京金山云网络技术有限公司 | A kind of method for video coding and device based on subjective quality model |
| CN108109147A (en) * | 2018-02-10 | 2018-06-01 | 北京航空航天大学 | A kind of reference-free quality evaluation method of blurred picture |
| CN109086995A (en) * | 2018-08-01 | 2018-12-25 | 广州虎牙信息科技有限公司 | Evaluation method, device, equipment and the storage medium of application program download path |
| CN109982069A (en) * | 2017-12-28 | 2019-07-05 | 浙江宇视科技有限公司 | The measurement method and system of cell breath |
| CN110232719A (en) * | 2019-06-21 | 2019-09-13 | 腾讯科技（深圳）有限公司 | A kind of classification method of medical image, model training method and server |
| CN110505472A (en) * | 2019-07-15 | 2019-11-26 | 武汉大学 | A kind of H.265 ultra high-definition method for evaluating video quality |
| CN110913207A (en) * | 2019-12-03 | 2020-03-24 | 华南理工大学 | Video transmission quality evaluation method based on multitask deep learning |
| CN111046893A (en) * | 2018-10-12 | 2020-04-21 | 富士通株式会社 | Image similarity determining method and device, and image processing method and device |
| CN111311584A (en) * | 2020-02-24 | 2020-06-19 | 北京百度网讯科技有限公司 | Video quality evaluation method and device, electronic equipment and readable medium |
| CN111711816A (en) * | 2020-07-08 | 2020-09-25 | 福州大学 | Video objective quality evaluation method based on observable coding effect intensity |
| CN111479106B (en) * | 2020-03-12 | 2021-06-29 | 上海交通大学 | Two-dimensional quality descriptor fused audio and video joint quality evaluation method and terminal |
| CN113298338A (en) * | 2020-10-19 | 2021-08-24 | 阿里巴巴集团控股有限公司 | Quality evaluation method and device |
| CN114584849A (en) * | 2019-09-24 | 2022-06-03 | 腾讯科技（深圳）有限公司 | Video quality evaluation method and device, electronic equipment and computer storage medium |
| CN114615484A (en) * | 2022-03-08 | 2022-06-10 | 常山县亿思达电子有限公司 | Vision tracking and positioning system based on retina monitoring |
| CN115225961A (en) * | 2022-04-22 | 2022-10-21 | 上海赛连信息科技有限公司 | No-reference network video quality evaluation method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1809175A (en) * | 2005-01-17 | 2006-07-26 | 华为技术有限公司 | Video quality evaluation method |
| CN101312544A (en) * | 2008-06-25 | 2008-11-26 | 厦门大学 | Video quality automatic evaluation system oriented to wireless network and evaluation method thereof |
| CN101426150A (en) * | 2008-12-08 | 2009-05-06 | 青岛海信电子产业控股股份有限公司 | Video image quality evaluation method and system |
| CN101742355A (en) * | 2009-12-24 | 2010-06-16 | 厦门大学 | Method for partial reference evaluation of wireless videos based on space-time domain feature extraction |
| EP2119248B1 (en) * | 2007-03-07 | 2011-01-12 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Concept for determining a video quality measure for block coded images |
| WO2011097697A1 (en) * | 2010-02-12 | 2011-08-18 | Ecole De Technologie Superieure | Method and system for determining a quality measure for an image using multi-level decomposition of images |
| CN102984540A (en) * | 2012-12-07 | 2013-03-20 | 浙江大学 | Video quality assessment method estimated on basis of macroblock domain distortion degree |
| CN104346809A (en) * | 2014-09-23 | 2015-02-11 | 上海交通大学 | Image quality evaluation method for image quality dataset adopting high dynamic range |
-
2015
- 2015-07-22 CN CN201510435863.6A patent/CN105100789B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1809175A (en) * | 2005-01-17 | 2006-07-26 | 华为技术有限公司 | Video quality evaluation method |
| EP2119248B1 (en) * | 2007-03-07 | 2011-01-12 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Concept for determining a video quality measure for block coded images |
| CN101312544A (en) * | 2008-06-25 | 2008-11-26 | 厦门大学 | Video quality automatic evaluation system oriented to wireless network and evaluation method thereof |
| CN101426150A (en) * | 2008-12-08 | 2009-05-06 | 青岛海信电子产业控股股份有限公司 | Video image quality evaluation method and system |
| CN101742355A (en) * | 2009-12-24 | 2010-06-16 | 厦门大学 | Method for partial reference evaluation of wireless videos based on space-time domain feature extraction |
| WO2011097697A1 (en) * | 2010-02-12 | 2011-08-18 | Ecole De Technologie Superieure | Method and system for determining a quality measure for an image using multi-level decomposition of images |
| CN102984540A (en) * | 2012-12-07 | 2013-03-20 | 浙江大学 | Video quality assessment method estimated on basis of macroblock domain distortion degree |
| CN104346809A (en) * | 2014-09-23 | 2015-02-11 | 上海交通大学 | Image quality evaluation method for image quality dataset adopting high dynamic range |
Non-Patent Citations (1)
| Title |
|---|
| 原菲等: "移动智能终端视频质量评价的应用研究", 《包装世界》 * |
Cited By (40)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105516713A (en) * | 2015-12-24 | 2016-04-20 | 招商局重庆交通科研设计院有限公司 | Image quality evaluation method of road traffic closed-circuit television based on machine vision |
| CN105635727A (en) * | 2015-12-29 | 2016-06-01 | 北京大学 | Subjective image quality evaluation method based on paired comparison and device thereof |
| CN105635727B (en) * | 2015-12-29 | 2017-06-16 | 北京大学 | Evaluation method and device based on the image subjective quality for comparing in pairs |
| CN106060537A (en) * | 2016-08-04 | 2016-10-26 | 腾讯科技（深圳）有限公司 | Method and device for assessing video quality |
| CN106060537B (en) * | 2016-08-04 | 2019-08-13 | 腾讯科技（深圳）有限公司 | A kind of video quality evaluation method and device |
| CN107948649A (en) * | 2016-10-12 | 2018-04-20 | 北京金山云网络技术有限公司 | A kind of method for video coding and device based on subjective quality model |
| CN106803272A (en) * | 2017-01-13 | 2017-06-06 | 惠州Tcl移动通信有限公司 | A kind of camera evaluating method and system |
| CN107249127A (en) * | 2017-05-18 | 2017-10-13 | 深圳众厉电力科技有限公司 | One kind assesses accurate network video quality assessment system |
| CN107249127B (en) * | 2017-05-18 | 2019-01-22 | 南京和电科技有限公司 | A kind of accurate network video quality assessment system of assessment |
| CN107743226A (en) * | 2017-11-06 | 2018-02-27 | 潘柏霖 | One kind monitors accurate environmental monitoring system |
| CN107862281A (en) * | 2017-11-06 | 2018-03-30 | 钟永松 | A kind of smart city planning system |
| CN107833433A (en) * | 2017-11-06 | 2018-03-23 | 龚土婷 | A kind of accurate mountain area natural calamity early warning system of early warning |
| CN107886234A (en) * | 2017-11-07 | 2018-04-06 | 钟永松 | A kind of efficient Urban Planning System |
| CN107832944A (en) * | 2017-11-07 | 2018-03-23 | 龚土婷 | A kind of natural hybridized orbit monitoring system |
| CN107806863A (en) * | 2017-11-07 | 2018-03-16 | 潘柏霖 | A kind of unattended reservoir monitoring system |
| CN109982069A (en) * | 2017-12-28 | 2019-07-05 | 浙江宇视科技有限公司 | The measurement method and system of cell breath |
| CN109982069B (en) * | 2017-12-28 | 2020-12-29 | 浙江宇视科技有限公司 | Method and system for measuring respiratory effect |
| CN108109147A (en) * | 2018-02-10 | 2018-06-01 | 北京航空航天大学 | A kind of reference-free quality evaluation method of blurred picture |
| CN108109147B (en) * | 2018-02-10 | 2022-02-18 | 北京航空航天大学 | No-reference quality evaluation method for blurred image |
| CN109086995A (en) * | 2018-08-01 | 2018-12-25 | 广州虎牙信息科技有限公司 | Evaluation method, device, equipment and the storage medium of application program download path |
| CN111046893B (en) * | 2018-10-12 | 2024-02-02 | 富士通株式会社 | Image similarity determining method and device, image processing method and device |
| CN111046893A (en) * | 2018-10-12 | 2020-04-21 | 富士通株式会社 | Image similarity determining method and device, and image processing method and device |
| US11954852B2 (en) | 2019-06-21 | 2024-04-09 | Tencent Technology (Shenzhen) Company Limited | Medical image classification method, model training method, computing device, and storage medium |
| CN110232719A (en) * | 2019-06-21 | 2019-09-13 | 腾讯科技（深圳）有限公司 | A kind of classification method of medical image, model training method and server |
| CN110505472A (en) * | 2019-07-15 | 2019-11-26 | 武汉大学 | A kind of H.265 ultra high-definition method for evaluating video quality |
| CN110505472B (en) * | 2019-07-15 | 2021-01-15 | 武汉大学 | Quality evaluation method for H.265 ultra-high-definition video |
| CN114584849A (en) * | 2019-09-24 | 2022-06-03 | 腾讯科技（深圳）有限公司 | Video quality evaluation method and device, electronic equipment and computer storage medium |
| CN114584849B (en) * | 2019-09-24 | 2023-05-05 | 腾讯科技（深圳）有限公司 | Video quality evaluation method, device, electronic equipment and computer storage medium |
| CN110913207A (en) * | 2019-12-03 | 2020-03-24 | 华南理工大学 | Video transmission quality evaluation method based on multitask deep learning |
| CN110913207B (en) * | 2019-12-03 | 2021-07-13 | 华南理工大学 | Video transmission quality evaluation method based on multitask deep learning |
| CN111311584A (en) * | 2020-02-24 | 2020-06-19 | 北京百度网讯科技有限公司 | Video quality evaluation method and device, electronic equipment and readable medium |
| CN111311584B (en) * | 2020-02-24 | 2023-06-23 | 北京百度网讯科技有限公司 | Video quality evaluation method and device, electronic equipment and readable medium |
| CN111479106B (en) * | 2020-03-12 | 2021-06-29 | 上海交通大学 | Two-dimensional quality descriptor fused audio and video joint quality evaluation method and terminal |
| CN111711816A (en) * | 2020-07-08 | 2020-09-25 | 福州大学 | Video objective quality evaluation method based on observable coding effect intensity |
| CN111711816B (en) * | 2020-07-08 | 2022-11-11 | 福州大学 | Video objective quality evaluation method based on observable coding effect intensity |
| CN113298338A (en) * | 2020-10-19 | 2021-08-24 | 阿里巴巴集团控股有限公司 | Quality evaluation method and device |
| CN114615484B (en) * | 2022-03-08 | 2022-11-01 | 常山县亿思达电子有限公司 | Vision tracking and positioning system based on retina monitoring |
| CN114615484A (en) * | 2022-03-08 | 2022-06-10 | 常山县亿思达电子有限公司 | Vision tracking and positioning system based on retina monitoring |
| CN115225961A (en) * | 2022-04-22 | 2022-10-21 | 上海赛连信息科技有限公司 | No-reference network video quality evaluation method and device |
| CN115225961B (en) * | 2022-04-22 | 2024-01-16 | 上海赛连信息科技有限公司 | No-reference network video quality evaluation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105100789B (en) | 2018-05-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105100789A (en) | Method for evaluating video quality | |
| CN105163106A (en) | Multi-data-processing video quality evaluation system | |
| CN100559881C (en) | A kind of method for evaluating video quality based on artificial neural net | |
| CN108428227B (en) | No-reference image quality evaluation method based on full convolution neural network | |
| Wang et al. | Multiscale structural similarity for image quality assessment | |
| CN100559880C (en) | A kind of highly-clear video image quality evaluation method and device based on self-adapted ST area | |
| CN101562675B (en) | No-reference image quality evaluation method based on Contourlet transform | |
| CN101783971B (en) | Method for subjective testing of video quality | |
| CN102740114B (en) | A kind of nothing ginseng appraisal procedure of Subjective video quality | |
| CN103475897B (en) | Adaptive image quality evaluation method based on distortion type judgment | |
| CN109379153A (en) | A kind of frequency spectrum sensing method | |
| CN106780446A (en) | It is a kind of to mix distorted image quality evaluating method without reference | |
| CN104486618A (en) | Video image noise detection method and device | |
| CN103856775A (en) | Processing method for subjective evaluation result of stereo video quality | |
| CN106375754B (en) | View-based access control model stimulates the video quality evaluation without reference method of attenuation characteristic | |
| CN108055533A (en) | A kind of subjective quality assessment method for panoramic video | |
| CN103077514B (en) | A kind of visual perception image quality evaluation method based on full variation | |
| CN101895787B (en) | Method and system for subjectively evaluating video coding performance | |
| CN105959685B (en) | A kind of compression bit rate Forecasting Methodology based on video content and cluster analysis | |
| CN116955068A (en) | Sequence similarity calculation and alarm processing method, device and storage medium | |
| CN106341683A (en) | Panoramic video quality judgment method and panoramic video quality judgment system | |
| Seshadrinathan et al. | New vistas in image and video quality | |
| Zhu et al. | Siqd: Surveillance image quality database and performance evaluation for objective algorithms | |
| CN114915777A (en) | Non-reference ultrahigh-definition video quality objective evaluation method based on deep reinforcement learning | |
| Loukil et al. | A new image quality metric using system visual human characteristics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |