CN105357526B - The mobile phone football video quality assessment device considering scene classification based on compression domain and method - Google Patents

Abstract

The invention discloses a kind of mobile phone football video quality assessment device considering scene classification based on compression domain and method.This apparatus for evaluating includes the compression coding module being sequentially connected with, the module that gets parms and quality assessment modules.This appraisal procedure comprises the following steps: 1) compression coding module carries out compression coding to the original video of input；2) video after compression coding is carried out half decoding and extracts compression domain parameter and process these parameters by the module that gets parms, and the compression domain parameter after process is for follow-up video quality assessment；3) quality assessment modules calculates the quality of mobile phone football video according to the video evaluation model set up.The present invention is during carrying out quality evaluation to football video, consider that three kinds of different scenes of football video improve the accuracy of assessment, football video scene is classified and the football video of different scenes is carried out quality evaluation by the parameter utilizing compression domain, thus is effectively improved feasibility and the real-time of assessment.

Description

The mobile phone football video quality assessment device considering scene classification based on compression domain and method

Technical field:

The invention belongs to video technique field, be specifically related to a kind of mobile phone football video quality assessment device considering scene classification based on compression domain and method.

Background technology:

Along with mobile device and the development of Video Applications, video watched by increasing user's selection mobile phone, but the viewing screen of mobile phone is less and transmission bandwidth be limited in the viewing experience that largely have impact on user.For football video, football video exists motion and less object faster, these all strong influence viewing experiences of user.

For problem above, current many researchs all concentrate on set up accurately video quality assessment model to improve the quality of video.Traditional video quality evaluation method according to the information relevant to original video utilized number can be divided into full reference, half reference and without reference.Without the appraisal procedure of reference owing to need not to utilize the information in original video to have higher real-time and practicality.Can be divided into again without reference method based on pixel domain and method based on compression domain, method based on compression domain need not complete decoded video streams just can obtain the information of compression domain, therefore has relatively low computation complexity and real-time.At present about the research of video quality evaluation although it is contemplated that user is experienced the impact of quality by video content, but for this particular kind of video of football, these researchs do not consider user's difference for its different scene focus.

Summary of the invention:

It is an object of the invention to for the current deficiency to mobile phone football video method for evaluating quality, provide a kind of mobile phone football video quality assessment device considering scene classification based on compression domain and method, for mobile phone football video carries out the scene classification accuracy with raising assessment, and use the method for compression domain to reduce the complexity calculated.

For reaching above-mentioned purpose, the present invention adopts the following technical scheme that and realizes:

The mobile phone football video quality assessment device considering scene classification based on compression domain, including the compression coding module being sequentially connected with, the module that gets parms and quality assessment modules；Wherein,

Compression coding module, for carrying out compression coding to the original video of input；

Get parms module, extracts compression domain parameter for the video after compression coding is carried out half decoding and processes these parameters, and the compression domain parameter after process is for follow-up video quality assessment；

Quality assessment modules, for calculating the quality of mobile phone football video according to the video evaluation model set up.

The present invention is further improved by, and in the module that gets parms, the compression domain parameter eventually for video quality assessment includes:

4*4 coded macroblocks ratio IMPI, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter STcloseup, middle scape critical parameter STmedium and distant view critical parameter STlong in logarithm coding bit rate LBR, resolution SR, I frame residual values RPVI, motion vector size MVM, I frame.

The mobile phone football video method for evaluating quality considering scene classification based on compression domain, comprises the following steps:

1) compression coding module carries out compression coding to the original video of input；

2) video after compression coding is carried out half decoding and extracts compression domain parameter and process these parameters by the module that gets parms, and the compression domain parameter after process is for follow-up video quality assessment；

3) quality assessment modules calculates the quality of mobile phone football video according to the video evaluation model set up.

The present invention is further improved by, step 1) concrete methods of realizing as follows:

101) original video of input is cut into the video segment of 8s；

102) utilizing H.264 form to encode, every eight frames include an I frame, seven P frames；

103) being encoded under three kinds of resolution and various code rate by each original video, wherein, three kinds of resolution are respectively 1280*720,848*480 and 576*320.

The present invention is further improved by, step 2) in the compression domain parameter eventually for video quality assessment that gets parms in module include:

4*4 coded macroblocks ratio IMPI, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter STcloseup, middle scape critical parameter STmedium and distant view critical parameter STlong in logarithm coding bit rate LBR, resolution SR, I frame residual values RPVI, motion vector size MVM, I frame.

The present invention is further improved by, and close shot critical parameter STcloseup, middle scape critical parameter STmedium and distant view critical parameter STlong are obtained through scene classification algorithm by compression domain parameter, specifically comprise the following steps that

201) calculated level or vertical motion vector ratio HMVR and MVM

Wherein,WithIt is respectively the number of the motion vector of the horizontal direction in jth frame and vertical direction,For the number of motion vectors whole in jth frame, N_sqFor the whole frame numbers in video sequence；

202) preliminary scene judges

When video HMVR is less than or equal to 70%, video is judged to close shot video or middle scape video；When video HMVR is less than 60%, video is judged to close shot video；When video HMVR is more than 70%, video is judged to distant view video or middle scape video；When video HMVR is more than 90%, video is judged to distant view video；

203) combine MVM and carry out further scene judgement

Coefficient of determination cla₁Computing formula be cla₁=HMVR × h₁+ MVM, it is determined that coefficient cla₂Computing formula be cla₂=HMVR × h₂-MVM；Work as cla₁＜ thr₁Time, video is judged to close shot；Work as cla₁≥thr₁, video is judged to middle scape；Work as cla₂＞ thr₂Time, video is judged to distant view；Work as cla₂≤thr₂Time, video is judged to middle scape；Wherein h₁=20, thr₁=45, h₂=80, thr₂=30；

Finally, above step the STcloseup of the close shot video determined is 1, STmedium and STlong is 0；The STmedium of middle scape video is 1, STcloseup and STlong is 0；The STlong of distant view video is 1, STcloseup and STmedium is 0.

The present invention is further improved by, and in I frame residual values RPVI, motion vector size MVM, I frame, 4*4 coded macroblocks ratio IMPI and P frame inter-coded macroblocks ratio IMPP are content characteristic parameter, by compression domain parameter through being calculated, specific as follows:

I frame residual values RPVI

Wherein Z represents the matrix of quantization parameter,Show the element in the matrix of the left side and the element multiplication of same position in the matrix of the right；HT is hada matrix；E_MBIt it is the estimated value of a matrix energy；SI is the set of macro block in I frame, and N is the number of pixel；

Motion vector size MVM

Wherein x_iAnd y_iFor i-th motion vector horizontal and the size of vertical direction, N_tNumber for total movement vector；

4*4 coded macroblocks ratio IMPI in I frame

Wherein N_iFor the number of I frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence,For the number of 4*4 coded macroblocks in frame in i-th I frame；

P frame inter-coded macroblocks ratio IMPP

Wherein N_pFor the number of P frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence,For the number of inter-coded macroblocks in i-th P frame.

The present invention is further improved by, and the mobile phone football video evaluation model Q in quality assessment modules is for calculating the quality score of video, and the computing formula of its primitive form is:

Wherein α, β, γ are the coefficient relevant to compression domain parameter.

The present invention is further improved by, and the computing formula of coefficient relevant to compression domain parameter in mobile phone football video evaluation model is:

α=a₁+b₁×SR+c₁×STcloseup+d₁×STlong+e₁×STmedium

β=a₂+b₂×SR+c₂×MVM+d₂×IMPI

γ=a₃+b₃×SR+c₃×RPVI+d₃×IMPI+e₃×IMPP+f₃×STcloseup+g₃×STlong+h₃×STmedium

Factor alpha and SR, STcloseup, STlong, STmedium are relevant, factor beta and SR, and MVM, IMPI are relevant, coefficient gamma and SR, and RPVI, IMPI, IMPP, STcloseup, STlong, STmedium are relevant；

Wherein a₁,b₁,c₁,d₁,e₁,a₂,b₂,c₂,d₂,a₃,b₃,c₃,d₃,e₃,f₃,g₃,h₃For coefficient to be trained.

The present invention is further improved by, and in mobile phone football video evaluation model, the grading system to video is 0-10 point of system；0-2 divides expression video quality extreme difference, and user cannot stand；3-4 divides expression video quality poor, and user cannot accept；5-6 divides expression video quality general, and user can accept and be ready to watch for a long time；7-8 divides expression video quality preferable, and user has preferably experience in watching process；9-10 divides expression video quality fabulous, and user is the most satisfied.

Compared with existing mobile phone football video quality assessment device and method, the method have the advantages that

1, the present invention mobile phone football video quality assessment device considering scene classification based on compression domain and method, consider the impact on video quality of three kinds of scenes of football video, and propose the scene classification method of a kind of football video, improve the accuracy to football video quality evaluation with this.

2, the parameter in football video Evaluation Model on Quality all uses compression domain parameter, compression domain parameter contains the compression domain parameter of spatial domain and time domain, these compression domain parameters are used to not only increase the accuracy of assessment, and be not required to video be decoded completely and can obtain, improve practicality and real-time, reduce computation complexity to a great extent.

3, saying on the whole, the mobile phone football video apparatus for evaluating of the present invention and method have ideal lifting in assessment performance and are easily achieved, and have good using value.

Accompanying drawing illustrates:

Fig. 1 is mobile phone football video apparatus for evaluating schematic diagram in the present invention.

Fig. 2 is compression domain parameter schematic diagram in the present invention.

Fig. 3 is football video scene classification process figure in the present invention.

Fig. 4 is the scatterplot testing subjective testing mark and the calculated mark of apparatus for evaluating obtained.

Detailed description of the invention:

The present invention will be described in detail below in conjunction with the accompanying drawings:

As it is shown in figure 1, the present invention mobile phone football video quality assessment device considering scene classification based on compression domain, including the compression coding module being sequentially connected with, the module that gets parms and quality assessment modules；Wherein,

Compression coding module, for carrying out compression coding to the original video of input；

Get parms module, extracts compression domain parameter for the video after compression coding is carried out half decoding and processes these parameters, and the compression domain parameter after process is for follow-up video quality assessment；

Quality assessment modules, for calculating the quality of mobile phone football video according to the video evaluation model set up.

The present invention mobile phone football video method for evaluating quality considering scene classification based on compression domain, it is characterised in that comprise the following steps:

1) compression coding module carries out compression coding to the original video of input；Concrete methods of realizing is as follows:

101) original video of input is cut into the video segment of 8s；

102) utilizing H.264 form to encode, every eight frames include an I frame, seven P frames；

103) being encoded under three kinds of resolution and various code rate by each original video, wherein, three kinds of resolution are respectively 1280*720,848*480 and 576*320.

2) video after compression coding is carried out half decoding and extracts compression domain parameter and process these parameters by the module that gets parms, and the compression domain parameter after process is for follow-up video quality assessment；

3) quality assessment modules calculates the quality of mobile phone football video according to the video evaluation model set up.

Above-mentioned steps 2) in the compression domain parameter eventually for video quality assessment that gets parms in module include: 4*4 coded macroblocks ratio IMPI, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter STcloseup, middle scape critical parameter STmedium and distant view critical parameter STlong in logarithm coding bit rate LBR, resolution SR, I frame residual values RPVI, motion vector size MVM, I frame.

Wherein, the parameter of extracting directly can there be coded macroblocks type and the discrete cosine transform coefficient of coding bit rate LBR, resolution SR, motion vector size MVM, I frame and P frame from compression domain.Specific as follows:

(1) logarithm coding bit rate LBR and resolution SR are video coding parameter, can directly obtain from compression domain parameter.

(2) in I frame residual values RPVI, motion vector size MVM, I frame, 4*4 coded macroblocks ratio IMPI and P frame inter-coded macroblocks ratio IMPP are content characteristic parameter, need to be by compression domain parameter through being calculated.

A) I frame residual values RPVI

Wherein Z represents the matrix of quantization parameter,Show the element in the matrix of the left side and the element multiplication of same position in the matrix of the right.HT is hada matrix.E_MBIt it is the estimated value of a matrix energy.SI is the set of macro block in I frame, and N is the number of pixel.

B) motion vector size MVM

Wherein x_iAnd y_iFor i-th motion vector horizontal and the size of vertical direction, N_tNumber for total movement vector.

C) 4*4 coded macroblocks ratio IMPI in I frame

Wherein N_iFor the number of I frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence,For the number of 4*4 coded macroblocks in frame in i-th I frame.

D) P frame inter-coded macroblocks ratio IMPP

Wherein N_pFor the number of P frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence,For the number of inter-coded macroblocks in i-th P frame.

(3) close shot critical parameter STcloseup, middle scape critical parameter STmedium, distant view critical parameter STlong are semantic scene parameter, need to be obtained through scene classification algorithm by compression domain parameter.

A) calculated level or vertical motion vector ratio HMVR and MVM

WhereinWithFor the horizontal direction in jth frame and the number of the motion vector of vertical direction,For the number of motion vectors whole in jth frame, N_sqFor the whole frame numbers in video sequence.

B) preliminary scene judges

When HMVR is less than 70%, video can be judged to close shot video or middle scape video；When video HMVR is less than 60%, directly video can be judged to close shot video.When HMVR is more than 70%, video can be judged to distant view video or middle scape video；When HMVR is more than 90%, video directly can be judged to distant view video.

C) combine MVM and carry out concrete scene judgement

Coefficient of determination cla₁Computing formula be cla₁=HMVR × h₁+ MVM, it is determined that coefficient cla₂Computing formula be cla₂=HMVR × h₂-MVM.Work as cla₁＜ thr₁Time, video can be judged as close shot；Work as cla₂≤thr₂, video can be judged as middle scape.Work as cla₂＞ thr₂Time, video can be judged as distant view；Work as cla₂≤thr₂Time, video can be judged as middle scape.Wherein h₁=20, thr₁=45, h₂=80, thr₂=30.

Finally, above step the STcloseup of the close shot video determined is 1, STmedium and STlong is 0；The STmedium of middle scape video is 1, STcloseup and STlong is 0；The STlong of distant view video is 1, STcloseup and STmedium is 0.

Above-mentioned steps 3) in mobile phone football video evaluation model in quality assessment modules for calculating the quality score of video, the computing formula of its primitive form is:

Wherein α, beta, gamma is the coefficient relevant to compression domain parameter, and computing formula is:

α=a₁+b₁×SR+c₁×STcloseup+d₁×STlong+e₁×STmedium

β=a₂+b₂×SR+c₂×MVM+d₂×IMPI

γ=a₃+b₃×SR+c₃×RPVI+d₃×IMPI+e₃×IMPP+f₃×STcloseup+g₃×STlong+h₃×STmedium

Factor alpha and SR, STcloseup, STlong, STmedium are relevant, factor beta and SR, and MVM, IMPI are relevant, coefficient gamma and SR, and RPVI, IMPI, IMPP, STcloseup, STlong, STmedium are relevant.

Wherein a₁,b₁,c₁,d₁,e₁,a₂,b₂,c₂,d₂,a₃,b₃,c₃,d₃,e₃,f₃,g₃,h₃For coefficient to be trained.Through experiment, obtaining the subjective testing mark of test video, the coefficient value finally trained is: a₁=7.600, b₁=1.9876, c₁=0.300, d₁=-0.950, e₁=-0.450, a₂=2.840, b₂=0.0026, c₂=0.5293, d₂=0.0919, a₃=0.2909, b₃=-0.0418, c₃=-3.4*10^-6,d₃=-0.0783, e₃=0.0327, f₃=0.2858, g₃=0.2798, h₃=0.3071.

In this Evaluation Model on Quality, the grading system to video is 0-10 point of system.0-2 divides expression video quality extreme difference, and user cannot stand；3-4 divides expression video quality poor, and user cannot accept；5-6 divides expression video quality general, and user can accept and be ready to watch for a long time；7-8 divides expression video quality preferable, and user has preferably experience in watching process；9-10 divides expression video quality fabulous, and user is the most satisfied.

In order to test this apparatus for evaluating effectiveness for video quality assessment, calculated for the apparatus for evaluating in this programme quality score is compared by inventor with subjective testing mark.Fig. 4 is that mobile phone football video quality can be relatively accurately estimated by the scatterplot of subjective testing mark mark calculated with apparatus for evaluating will become apparent from this device.

Claims (7)

1. the mobile phone football video quality assessment device considering scene classification based on compression domain, it is characterised in that Including the compression coding module being sequentially connected with, the module that gets parms and quality assessment modules；Wherein,

Compression coding module, for carrying out compression coding to the original video of input；

Get parms module, extracts compression domain parameter also for the video after compression coding is carried out half decoding Processing these parameters, the compression domain parameter after process is for follow-up video quality assessment；

Quality assessment modules, for calculating the quality of mobile phone football video according to the video evaluation model set up；

Wherein, in the module that gets parms, the compression domain parameter eventually for video quality assessment includes:

Logarithm coding bit rate LBR, resolution SR, I frame residual values RPVI, motion vector size MVM, 4*4 coded macroblocks ratio IMPI, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter in I frame STcloseup, middle scape critical parameter STmedium and distant view critical parameter STlong.

2. the mobile phone football video method for evaluating quality considering scene classification based on compression domain, it is characterised in that Comprise the following steps:

1) compression coding module carries out compression coding to the original video of input；Concrete methods of realizing is as follows:

101) original video of input is cut into the video segment of 8s；

102) utilizing H.264 form to encode, every eight frames include an I frame, seven P frames；

103) each original video is encoded under three kinds of resolution and various code rate, wherein, three kinds points Resolution is respectively 1280*720,848*480 and 576*320；

2) video after compression coding is carried out half decoding and extracts compression domain parameter and to this by the module that gets parms A little parameters process, and the compression domain parameter after process, for follow-up video quality assessment, wherein obtains ginseng In digital-to-analogue block, the compression domain parameter eventually for video quality assessment includes: logarithm coding bit rate LBR, point 4*4 coded macroblocks ratio IMPI in resolution SR, I frame residual values RPVI, motion vector size MVM, I frame, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter STcloseup, middle scape critical parameter STmedium And distant view critical parameter STlong；

3) quality assessment modules calculates the quality of mobile phone football video according to the video evaluation model set up.

The mobile phone football video quality considering scene classification based on compression domain the most according to claim 2 Appraisal procedure, it is characterised in that close shot critical parameter STcloseup, middle scape critical parameter STmedium and Distant view critical parameter STlong is obtained through scene classification algorithm by compression domain parameter, specifically comprises the following steps that

201) calculated level or vertical motion vector ratio HMVR and MVM

$HMVR=\frac{\underset{j=1}{\overset{N_{sq}}{\Σ}}(N_{hor}^j+N_{ver}^j)/N_{fra}^j}{N_{sq}}$

Wherein,WithIt is respectively the number of the motion vector of the horizontal direction in jth frame and vertical direction,For the number of motion vectors whole in jth frame, N_sqFor the whole frame numbers in video sequence；

202) preliminary scene judges

When video HMVR is less than or equal to 70%, video is judged to close shot video or middle scape video；When regarding Frequently, when HMVR is less than 60%, video is judged to close shot video；When video HMVR is more than 70%, Video is judged to distant view video or middle scape video；When video HMVR is more than 90%, video is judged to Distant view video；

203) combine MVM and carry out further scene judgement

Coefficient of determination cla₁Computing formula be cla₁=HMVR × h₁+ MVM, it is determined that coefficient cla₂Computing formula be cla₂=HMVR × h₂-MVM；Work as cla₁＜ thr₁Time, video is judged to close shot；Work as cla₁≥thr₁, video is judged to middle scape； Work as cla₂＞ thr₂Time, video is judged to distant view；Work as cla₂≤thr₂Time, video is judged to middle scape；Wherein h₁=20, thr₁=45, h₂=80, thr₂=30；

Finally, above step the STcloseup of the close shot video determined is 1, STmedium and STlong It is 0；The STmedium of middle scape video is 1, STcloseup and STlong is 0；The STlong of distant view video Being 1, STcloseup and STmedium is 0.

The mobile phone football video quality considering scene classification based on compression domain the most according to claim 3 Appraisal procedure, it is characterised in that in I frame residual values RPVI, motion vector size MVM, I frame, 4*4 compiles Decoding macroblock ratio IMPI and P frame inter-coded macroblocks ratio IMPP are content characteristic parameter, compression domain join Number is through being calculated, specific as follows:

I frame residual values RPVI

$E_{MB}=\left\{\begin{array}{cc}{2}^{(Q_{P,MB}-4)/3}\×sum(Z_{MB}\⊗Z_{MB})& HTOFF\\ 2^{(Q_{P,MB}-4)/3}\×\left(sum\right(Z_{MB}^{\′}\⊗Z_{MB}^{\′})+sum(Z_D^{\′\′}\⊗Z_D^{\′\′}\left)\right)& HTON\end{array}\right.$

$RPVI=\sqrt{\frac{1}{N}\underset{n\∈S_1}{\Σ}{E}_{MB,n}}$

Wherein, Q_p,MBFor the quantization parameter of macro block, Z represents the matrix of quantization parameter, Z_MBQuantization system for macro block Matrix number, Z_DFor the quantization parameter matrix through Hadamard transform,Show the element in the matrix of the left side and the right square The element multiplication of same position in Zhen；HT is hada matrix, and HT ON indicates the situation of Hadamard transform, HT OFF indicates the situation without Hadamard transform；E_MBIt it is the estimated value of a matrix energy；E_MB,nFor being numbered The estimated value of the macro block energy of n；SI is the set of macro block in I frame, and N is the number of pixel；

Motion vector size MVM

$MVM=\frac{1}{N_t}\underset{i}{\overset{N_t}{\Σ}}\sqrt{x_i^2+y_i^2}$

Wherein x_iAnd y_iFor i-th motion vector horizontal and the size of vertical direction, N_tFor total movement vector Number；

4*4 coded macroblocks ratio IMPI in I frame

$IMPI=\frac{1}{N_i}\underset{i}{\overset{N_i}{\Σ}}\frac{N_{\mathrm{int}ra4*4}^i}{N_{mb}}$

Wherein N_iFor the number of I frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence, For the number of 4*4 coded macroblocks in frame in i-th I frame；

P frame inter-coded macroblocks ratio IMPP

$IMPP=\frac{1}{N_p}\underset{i}{\overset{N_p}{\Σ}}\frac{N_{\mathrm{int}ra}^i}{N_{mb}}$

Wherein N_pFor the number of P frame, N in video sequence_mbFor the number of macro block in the every frame of video sequence, For the number of inter-coded macroblocks in i-th P frame.

The mobile phone football video quality considering scene classification based on compression domain the most according to claim 2 Appraisal procedure, it is characterised in that the mobile phone football video evaluation model Q in quality assessment modules regards for calculating The quality score of frequency, the computing formula of its primitive form is:

$Q=\mathrm{\α}\×\exp {(-\frac{\min (LBR-\mathrm{\β},0)}{\mathrm{\γ}})}^2$

Wherein α, beta, gamma is the coefficient relevant to compression domain parameter, and LBR is logarithm coding bit rate.

The mobile phone football video quality considering scene classification based on compression domain the most according to claim 5 Appraisal procedure, it is characterised in that coefficient relevant to compression domain parameter in mobile phone football video evaluation model Computing formula is:

α=a₁+b₁×SR+c₁×STcloseup+d₁×STlong+e₁×STmedium

β=a₂+b₂×SR+c₂×MVM+d₂×IMPI

γ=a₃+b₃×SR+c₃×RPVI+d₃×IMPI+e₃×IMPP+f₃×STcloseup+g₃×STlong+h₃×STmedium

Factor alpha and resolution SR, close shot critical parameter STcloseup, distant view critical parameter STlong and Middle scape critical parameter STmedium is relevant, factor beta and resolution SR, motion vector size MVM and I In frame, 4*4 coded macroblocks ratio IMPI is relevant, coefficient gamma and resolution SR, I frame residual values RPVI, I frame Interior 4*4 coded macroblocks ratio IMPI, P frame inter-coded macroblocks ratio IMPP, close shot critical parameter STcloseup, distant view critical parameter STlong and middle scape critical parameter STmedium are relevant；

Wherein a₁,b₁,c₁,d₁,e₁,a₂,b₂,c₂,d₂,a₃,b₃,c₃,d₃,e₃,f₃,g₃,h₃For coefficient to be trained.

The mobile phone football video quality considering scene classification based on compression domain the most according to claim 5 Appraisal procedure, it is characterised in that in mobile phone football video evaluation model, the grading system to video is that 0-10 divides System；0-2 divides expression video quality extreme difference, and user cannot stand；3-4 divides expression video quality poor, user Cannot accept；5-6 divides expression video quality general, and user can accept and be ready to watch for a long time；7-8 divides Representing that video quality is preferable, user has preferably experience in watching process；9-10 divides expression video quality pole Good, user is the most satisfied.