CN109495772A - A kind of sort method and system of video quality - Google Patents
A kind of sort method and system of video quality Download PDFInfo
- Publication number
- CN109495772A CN109495772A CN201710811287.XA CN201710811287A CN109495772A CN 109495772 A CN109495772 A CN 109495772A CN 201710811287 A CN201710811287 A CN 201710811287A CN 109495772 A CN109495772 A CN 109495772A
- Authority
- CN
- China
- Prior art keywords
- video data
- video
- evaluation information
- centering
- training sample
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/262—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists
- H04N21/26258—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists for generating a list of items to be played back in a given order, e.g. playlist, or scheduling item distribution according to such list
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/458—Scheduling content for creating a personalised stream, e.g. by combining a locally stored advertisement with an incoming stream; Updating operations, e.g. for OS modules ; time-related management operations
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/81—Monomedia components thereof
- H04N21/8126—Monomedia components thereof involving additional data, e.g. news, sports, stocks, weather forecasts
- H04N21/8133—Monomedia components thereof involving additional data, e.g. news, sports, stocks, weather forecasts specifically related to the content, e.g. biography of the actors in a movie, detailed information about an article seen in a video program
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/83—Generation or processing of protective or descriptive data associated with content; Content structuring
- H04N21/845—Structuring of content, e.g. decomposing content into time segments
- H04N21/8456—Structuring of content, e.g. decomposing content into time segments by decomposing the content in the time domain, e.g. in time segments
Abstract
The application embodiment discloses the sort method and system of a kind of video quality, wherein the method provides sets of video data, and it includes at least one set of video data pair that the video data, which is concentrated,；The described method includes: reading one group of target video data pair that the video data is concentrated；The user is received for the target video data to the evaluation information delivered；Video data based on the target video data centering constructs training sample, and by the training sample input sequencing model, obtains ranking results, the ranking results are for characterizing the target video data centering quality preferably video data；The order models are corrected according to the difference value between the ranking results and the evaluation information, so that obtained ranking results are consistent with the evaluation information when training sample to be inputted to the order models after correction again.Technical solution provided by the present application can be improved the efficiency of sequence.
Description
Technical field
This application involves field of computer technology, in particular to the sort method and system of a kind of video quality.
Background technique
With the continuous development of computer technology, many video editing techniques are emerged.These video editing techniques are for example
Including technologies such as video code conversion, video compress, video beautifications.
Currently, when the original video for recording video recording device is uploaded to video playback website, due to video playing
Website would generally data volume to video and format limit, it is therefore desirable to original video is carried out at compression and/or transcoding
Reason.Using different compressions and transcoding technology, treated that video would generally show different effects.Such as it is clear in picture
It can change in terms of degree and/or speech recognition.
In addition, landscaping treatment can be carried out to original video in order to enable video pictures more attract user.Similarly,
Different effects can also be showed using the video after different beautification technical treatments.
It, currently can be in order to pick out the video editing techniques that can be easily accepted by a user from numerous video editing techniques
It is realized by the method for video quality sequence.Specifically, it can be provided simultaneously to user by video editing techniques processing
The original video editing video and being handled without video editing techniques.User watched editor video and original video it
Afterwards, can the image quality that presents of the two basic videos, respectively editor's video and original video give a mark, thus according to
The result of marking is ranked up the two videos.In this way, can judge that currently employed video is compiled according to ranking results
Whether the technology of collecting is easily accepted by a user.
However, the sort method of this video quality in the prior art is usually by manually by way of watching video
It is ranked up, not only cost is larger, and efficiency is relatively low.
Summary of the invention
The purpose of the application embodiment is to provide the sort method and system of a kind of video quality, can be improved sequence
Efficiency.
To achieve the above object, the application embodiment provides a kind of sort method of video quality, provides video data
Collection, it includes at least one set of video data pair that the video data, which is concentrated, and two video datas of the video data centering have
Identical original video data；Wherein, at least one video data is the original video number in described two video datas
According to what is obtained by processing, which comprises read one group of target video data pair that the video data is concentrated, and successively
The video that two video datas of target video data centering characterize is shown to user；The user is received for the target
Video data is to the evaluation information delivered, and the evaluation information is for determining that the target video data centering quality preferably regards
Frequency evidence；Video data based on the target video data centering constructs training sample, and the training sample is inputted and is arranged
Sequence model obtains ranking results, and the ranking results are for characterizing the target video data centering quality preferably video counts
According to；The order models are corrected according to the difference value between the ranking results and the evaluation information, so that will
When the training sample inputs the order models after correction again, obtained ranking results are consistent with the evaluation information.
To achieve the above object, the application embodiment also provides a kind of ordering system of video quality, the system packet
Memory and processor are included, computer program and sets of video data are stored in the memory, the video data concentrates packet
At least one set of video data pair is included, two video datas of the video data centering have identical original video data；Its
In, at least one video data is that the original video data is obtained by processing in described two video datas；It is described
When computer program is executed by the processor, following functions are realized: reading one group of target video that the video data is concentrated
Data pair, and the video of two video datas of target video data centering characterization is successively shown to user；Receive the use
Family is for the target video data to the evaluation information delivered, and the evaluation information is for determining the target video data pair
Middle quality preferably video data；Video data based on the target video data centering constructs training sample, and will be described
Training sample input sequencing model obtains ranking results, and the ranking results are for characterizing the target video data centering matter
Amount preferably video data；The order models are carried out according to the difference value between the ranking results and the evaluation information
Correction, so that when the training sample to be inputted to the order models after correction again, obtained ranking results and the evaluation
Information is consistent.
Therefore technical solution provided by the present application, two views in one group of target video data pair are being shown to user
After the video of frequency data characterization, user can receive for the target video data to the evaluation information delivered.Evaluation letter
Breath can characterize user and feel quality preferably video data.If the quality preferably video data is treated video
Data then show the video editing techniques used when processing by customer acceptance.It is on the contrary then show that user does not recognize adoptable view
Frequency editing technique.In this way, the video data pair that can largely compare can be provided a user, and receive the evaluation of user feedback
Information.Then it can use the video data of these video data centerings as training sample, order models be trained.?
When training, order models can export the ranking results for video data pair.But since initial order models may be less
Accurately, the evaluation information that will lead to ranking results and user feedback is inconsistent.In such a case, it is possible to be tied according to the sequence
Difference value between fruit and the evaluation information is corrected the order models, so that the order models after correction can
The ranking results of the correctly predicted video data pair.In this way, after completing training process, when needs are to new video data
When being ranked up to the quality of the video of characterization, can directly by new video data to the input data as order models,
So as to export the ranking results of the new video data pair by the order models for completing training.Therefore the application
The technical solution of offer can train the higher sequence of precision in conjunction with the method for the evaluation information and machine learning of user
Model, so as to automatically to new video data to being ranked up, improve the efficiency of video quality sequence.
Detailed description of the invention
It, below will be to embodiment in order to illustrate more clearly of the application embodiment or technical solution in the prior art
Or attached drawing needed to be used in the description of the prior art is briefly described, it should be apparent that, the accompanying drawings in the following description is only
It is some embodiments as described in this application, for those of ordinary skill in the art, in not making the creative labor property
Under the premise of, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the sort method flow chart of video quality in the application embodiment；
Fig. 2 is the interception schematic diagram of video segment data in the application embodiment；
Fig. 3 is the judgement schematic diagram of multiple video segment datas in the application embodiment；
Fig. 4 is the structural schematic diagram of the ordering system of video quality in the application embodiment.
Specific embodiment
In order to make those skilled in the art better understand the technical solutions in the application, below in conjunction with the application reality
The attached drawing in mode is applied, the technical solution in the application embodiment is clearly and completely described, it is clear that described
Embodiment is only a part of embodiment of the application, rather than whole embodiments.Based on the embodiment party in the application
Formula, every other embodiment obtained by those of ordinary skill in the art without making creative efforts, is all answered
When the range for belonging to the application protection.
The application provides a kind of sort method of video quality, and the method can be applied to have data processing function
In terminal device.The terminal device for example can be desktop computer, laptop, tablet computer, work station etc..In this reality
It applies in mode, sets of video data can be provided in the terminal device, it includes at least one set of video counts that the video data, which is concentrated,
According to right.It may include two video datas for being able to carry out comparison in the video data pair.The two of the video data centering
A video data can have identical original video data.The video data of the video data centering can be described original
Video data itself is also possible to what the original video data was obtained by processing.For example, the original video data is logical
The data of video recording device recording are crossed, which does not pass through video editing processing.So in the video data pair, one
A video data can be the original video data, another video data then can be the original video data by view
The video data that frequency editing technique obtains after being handled.Certainly, two video datas of the video data centering can also be with
It is all the video data obtained after video editing techniques are handled, only video editing skill used by two video datas
Art can be different.For example, a video data of the video data centering is to carry out picture pressure to the original video data
The video data obtained after contracting, another video data then can be for the views obtained after initial data progress picture beautification
Frequency evidence.In practical application scene, in order to embody the difference of two video datas of video data centering, it is ensured that
At least one video data is that the original video data is obtained by processing in described two video datas.Due to two
Video data corresponds to the same original video data, therefore the corresponding video content of the two video datas and video length are all
It can be identical, the picture or sound effect only shown has difference.
Referring to Fig. 1, the sort method of video quality provided by the present application may comprise steps of.
S1: one group of target video data pair that the video data is concentrated is read, and successively shows the target to user
The video of two video datas of video data centering characterization.
In the present embodiment, terminal device can read one group of target view that the video data is concentrated from memory
Frequency is according to right.The target video data concentrates a random set video data pair to can be the video data.It is reading
The target video data can successively render two video datas wherein included to later, so as to
The video of video data characterization is shown in display.In this way, user can be corresponding by the viewing of the display of terminal device
Video.In addition, user can also use the electronic equipment different from the terminal device.Terminal device is reading the target
After video data, which can be sent at the electronic equipment that user uses, so as in electronic equipment
Place is rendered and is shown.
In the present embodiment, to user show two video datas sequence can be it is random, to avoid user
In the case of the video data corresponding to the currently watched video is guessed and causes ranking results inaccurate.In addition, working as
Some video data, by after user viewing, no longer can provide the video data pair to the user to the video of characterization,
It is not accurate enough familiar with rear made evaluation to video content to avoid user.In addition, in video display process, Ke Yijin
Only suspend, review and F.F., to guarantee the integrality of video and guarantee that video can not be repeated viewing.
In one embodiment, it is contemplated that if allowing the longer video of user's viewing time, user is likely to only remember
The firmly shorter a part of video content of last duration, and evaluation information can be made based on this partial video content.In consideration of it, at this
The shorter video content of duration can be provided a user in embodiment every time, to guarantee the standard for the evaluation information that user makes
True property.Specifically, terminal device is reading the target video data for preparing to be supplied to user to later, it can be determined that the target
Whether the corresponding playing duration of video data centering video data is more than or equal to specified duration threshold value.The specified duration threshold
Value can be pre-set time constant.For example, the specified duration threshold value can be 20 seconds.When the target video data
When the corresponding playing duration of centering video data is more than or equal to specified duration threshold value, one can be intercepted out from video data
Section video segment data is supplied to user.It specifically, referring to Fig. 2, can be from two videos of the target video data pair
Interception has the video segment data of identical play position and identical playing duration respectively in data, constitutes video segment data
It is right.Described have identical play position and identical playing duration and can refer to two video clips of the video segment data centering
The video that data are characterized has identical initial time node and terminates timing node.For example, the target that terminal device is read
The corresponding playing duration of two video datas of video data centering is 5 minutes, and the specified duration threshold value is 30 seconds, then
Play time can be intercepted in the two video datas respectively at this time in 5 minutes to 5 points 25 seconds video segment datas, thus
Obtain the video segment data pair being made of two video segment datas.It should be noted that in practical application scene, in order to
The randomness for guaranteeing the video segment data of interception can determine the initial time node of the video of interception at random, and one
The video length of interception is determined within the scope of fixed duration at random.For example, the duration range can be 10 seconds to 30 seconds, in this way,
The duration of the video finally intercepted can be a random value in 10 seconds to 30 seconds.
In the present embodiment, intercepting out the video segment data to later, can successively show to user described in
The video clip of two video segment datas of video segment data centering characterization.Specifically, described two piece of video can be rendered
Segment data, and be supplied to user according to random order and watch.
S3: the user is received for the target video data to the evaluation information delivered, the evaluation information is used for
Determine the target video data centering quality preferably video data.
In the present embodiment, user can deliver after having viewed target video data to two videos of characterization
For the evaluation information of the target video data pair.The evaluation information can be used for determining the target video data centering
Quality preferably video data.Specifically, user can be selected after having watched two videos for the option provided
It selects.The option for example may include " first video is preferable ", " second video is preferable " and " two videos are not obvious
Difference ".After having selected corresponding option, terminal device can receive corresponding with the option of user's selection user
Evaluation information, so as to know the user to the evaluation result of two video datas of the target video data centering.
In one embodiment, when in the target video data the corresponding playing duration of video data it is longer, need
When intercepting video segment data from video data, obtained two piece of video number of segment of video segment data centering can will be intercepted
User is showed to watch according to the video clip of characterization.In this way, user after having watched video clip, equally can choose accordingly
Option, obtain the Segment evaluation information of the video data pair.The Segment evaluation information can be used for determining the piece of video
Segment data centering quality preferably video segment data.
In the present embodiment, quality preferably video segment data in the video segment data pair is determined in user
It afterwards, then can be according to judgement as a result, determining the target video data centering quality preferably video data.In an applied field
Jing Zhong only can intercept one group of video segment data from target video data centering to accelerate the determination flow of video quality
It is right, and user can be directed to the video segment data to the Segment evaluation information delivered as the target video data pair
Evaluation information.It specifically, can be using video data belonging to the quality preferably video segment data as the target
Video data centering quality preferably video data.For example, the target video data centering includes video data A and video counts
According to B, video segment data A1 can be intercepted from video data A, and video segment data B1 can be intercepted from video data B.
If the bright video segment data A1 of the Segment evaluation information table of user is better than video segment data B1, it is considered that video counts
It is better than video data B according to A.
In another embodiment, in order to avoid evaluating the error caused by carrying out single to video segment data,
Multiple groups video segment data pair can be intercepted from target video data centering, and can be in conjunction with the piece of multiple groups video segment data pair
Section evaluation information, the comprehensive evaluation information for obtaining target video data pair.
In the present embodiment, can be from the quantity of the video segment data pair of target video data centering interception
At least two groups.In this way, being directed to every group of video segment data pair, the available user is directed to the video segment data pair
The local evaluation information delivered.Wherein, the local evaluation information can be used for characterizing quality preferably video segment data institute
The video data of category.For example, 10 groups of video segment datas pair have been intercepted from target video data centering, for every group of video clip
Data pair, with local evaluation information can be delivered per family.The part evaluation information is in addition to that can determine video segment data centering
Quality preferably video segment data, while can also determine video data belonging to the quality preferably video segment data.
Such as in the example of above embodiment, when video segment data A1 is better than video segment data B1, the part evaluation letter
Breath can characterize quality preferably video segment data and belong to video clip A.
In this way, to each group video segment data to evaluating after, available multiple local evaluation informations, these
Local evaluation information can characterize respective video data.In the present embodiment, identical local evaluation information can be counted
Quantity.Wherein, identical local evaluation information can refer to that the video data of characterization is the part evaluation of the same video data
Information.For example, referring to Fig. 3, can count in 4 local evaluation informations and obtain characterization quality preferably piece of video number of segment
Quantity according to the local evaluation information for belonging to video data A is 3, and characterizes quality preferably video segment data and belong to video counts
Quantity according to the local evaluation information of B is 1.At this point it is possible to which the most part of quantity is commented using the principle that the minority is subordinate to the majority
The video data of valence information representation is as the target video data centering quality preferably video data.Example as escribed above
In, it can be using video data A as quality preferably video data.Certainly, in practical application scene, it is possible that quantity
Most local evaluation information more than one, at this point it is possible to the method evaluated by continuing growing user, until being counted
Until the most local evaluation information of amount is one.
S5: the video data based on the target video data centering constructs training sample, and the training sample is defeated
Enter order models, obtain ranking results, the ranking results are preferably regarded for characterizing the target video data centering quality
Frequency evidence.
In the present embodiment, a large amount of video data pair can be provided in step S1, in step s3 so as to
To with these video datas to corresponding evaluation information.These video datas pair and corresponding evaluation information can be used as machine
The training sample of study, so that training obtains the model that can meet user's evaluation behavioural habits, which can be to input
Video data is evaluated, and exports the ranking results of video data.
In the present embodiment, training sample can be constructed based on the video data of the target video data centering, and
By the training sample input sequencing model.In practical application scene, the order models can use RankSVM model,
The training sample then can be constructed as the training sample suitable for RankSVM model.Specifically, when constructing training sample,
Can extract respectively first video data of target video data centering first eigenvector and the second video data
Two feature vectors.Described eigenvector can be using convolutional neural networks (Convolutional Neural Network,
CNN it) is obtained after the picture progress feature extraction in the video characterized to video data.Obtaining the first eigenvector
It, can the first eigenvector and the positive training sample of second feature vector building based on extraction after second feature vector
Sheet and reverse train sample.Wherein, the sequence of feature vector can be in the positive training sample and the reverse train sample
It is different.Specifically, in the positive training sample, the first eigenvector is located at before the second feature vector.Example
Such as, i-th of the building positive training sample can be expressed as (xi,yi), wherein xiIndicate first eigenvector, yiIt indicates
Second feature vector.Correspondingly, in the reverse train sample, the second feature vector can be located at the fisrt feature
Before vector.For example, i-th of reverse train sample of building can be expressed as (yi,xi)。
In the present embodiment, the positive training sample and the reverse train sample can respectively with respective theory
Value is associated.Specifically, the positive training sample is associated with the first theoretical value, the reverse train sample and the second theory
Value is associated.Wherein, the theoretical value can characterize theory corresponding to the sortord according to feature vector in training sample
Ranking results.For example, if user shows the first view for the evaluation information that the first video data and the second video data are delivered
The quality of frequency evidence is more excellent, then before coming Second Eigenvalue due to the First Eigenvalue in positive training sample, then described
Positive associated first theoretical value of training sample can be 1, show in positive training sample characteristic value put in order with it is actual
The quality-ordered of video data is identical.And before Second Eigenvalue comes the First Eigenvalue in the reverse train sample, therefore
Associated second theoretical value of reverse train sample can be -1, indicate that characteristic value puts in order and reality in reverse train sample
The quality-ordered of the video data on border is opposite.Therefore, first theoretical value and second theoretical value can be different.
In the present embodiment, after obtaining positive training sample and reverse train sample, can respectively by it is described just
To training sample and the reverse train sample input sequencing model, corresponding first actual value of the positive training sample is obtained
And corresponding second actual value of the reverse train sample.It should be noted that in the incipient training stage, the sequence
Parameter in model may be inaccurate, therefore the first obtained actual value may be different from the first theoretical value, the second actual value
It may also be different from the second theoretical value.
S7: being corrected the order models according to the difference value between the ranking results and the evaluation information,
So that when the training sample to be inputted to the order models after correction again, obtained ranking results and the evaluation information one
It causes.
In the present embodiment, for the order models in building, parameter therein is all the default value of setting, these are silent
Recognizing value may can not be adapted with current training sample completely.The ranking results that therefore, it is necessary to be exported according to order models
Order models are corrected with the actual evaluation information of user.
In the present embodiment, the ranking results of order models output can be a numerical value, which can indicate two
The sortord of a video data.Such as before 1 can indicate that the first video data comes the second video data, -1 can be indicated
Second video data comes before the first video data.When the ranking results of order models output and user are directed to the two videos
When the actual evaluation information of data announces is inconsistent, then show that the parameter in order models needs to adjust.Specifically, it can calculate
Difference value between the ranking results and the evaluation information, and the sequence mould is inputted using the difference value as feedback information
Type, so that order models can be adjusted internal parameter.In practical application scene, inside the order models
Parameter can be one or more parameter matrix, the numerical value in parameter matrix can be adjusted according to certain interval of values
It is whole.In this way, when order models receive the difference value between ranking results and evaluation information, it can be according to the big of the difference value
The small and described interval of values determines numerical value that needs adjust.For example, the difference value is 3 times that error allows maximum value,
3 times for needing the numerical value that adjusts as the interval of values can so be set.In this way, difference value is bigger, what setting needed to adjust
Numerical value can also be bigger.
In the present embodiment, when the training sample of input is positive training sample and reverse train sample, Ke Yifen
Order models are not corrected using the training result of the positive training sample and the training result of reverse train sample.
Specifically, by the positive training sample and the reverse train sample input sequencing model, the positive training sample is obtained
After this corresponding first actual value and corresponding second actual value of the reverse train sample, it is practical that described first can be calculated
The first difference value between value and first theoretical value, and using first difference value to the parameter in the order models
It is corrected.Likewise it is possible to calculate the second difference value between second actual value and second theoretical value, and utilize
Second difference value is corrected the parameter in the order models.
It in the present embodiment, can be defeated by the training sample again after being corrected to the order models
Order models after entering correction, then can be corrected again according to obtained ranking results, until inputting the trained sample
After this, until obtained ranking results are consistent with the evaluation information.In this way, can be made by the training of mass data
Parameter in the order models can be suitable for most of video data.
In one embodiment, after completing the training stage to the order models, the order models can be utilized
Two video datas for treating sequence are ranked up.Two video datas to be sorted can be to be evaluated without user
Data, since the parameter in order models have passed through adjustment in the training stage, order models can based on with
The compatible parameter of evaluation behavior two video datas to be sorted to this in family are handled, to obtain described to be sorted two
The corresponding ranking results of a video data.
In one embodiment, can using various video edit methods to the same original video data at
Reason, to obtain different treated video datas.It, can be by right in order to judge the superiority and inferiority of these video editing methods
Treated video data is ranked up to realize.At this point, the quantity of video data to be sorted can more than two, and
It is at least three.It in this case, can be by the video counts to be sorted in order to guarantee that order models can work normally
According to composition sub-video data pair two-by-two.For example, currently total, there are three video data A, B, C to be sorted, then may be constructed
Three sub-video datas pair of (A, B), (A, C) and (B, C).It may include two video counts in each sub-video data pair
According to.In this way, can be using above-mentioned order models successively to these sub-video datas to being ranked up.Specifically, Ke Yiyi
It is secondary by the sub-video data to input correction after order models, obtain two video datas of the sub-video data centering
Ranking results.For example, the ranking results of three above-mentioned sub-video datas pair can be respectively A excellent better than C and B better than B, A
In C.So, the ranking results based on the sub-video data pair can be ranked up the video data to be sorted.
For example, final ranking results can be, A is optimal, and B takes second place, and C is worst.In this way, by being ranked up to video quality, so as to
To embody the superiority and inferiority of various video editing methods.
In one embodiment, it is contemplated that environment of the user when watching video may be to final evaluation information structure
At influence.For example, being directed to two different video datas, carried out by the higher equipment of resolution ratio and the lower equipment of resolution ratio
When broadcasting, the result of presentation may be entirely different.When user watches the two video datas in the lower equipment of resolution ratio
When the video of characterization, it is apparent for may feeling the difference between two videos not.However, when user sets resolution ratio is higher
When the video of standby upper the two video datas of viewing characterization, obvious difference can be found.Therefore, the use is being received
When family is directed to the target video data to the evaluation information delivered, the target video data can also be recorded together to characterization
Video when playing locating environmental information, and it is the environmental information is associated with the evaluation information.The environment letter
Breath for example can be above-mentioned resolution information, can also be the type information for playing the terminal device of video, can also be aobvious
The dimension information etc. of display screen.In this way, after introducing environmental information, the order models corrected based on the evaluation information are just
It can be used for the video data in the environment in environmental information characterization to being ranked up.For example, order models are needles
The playing environment of large-screen mobile phone is trained, is ranked up then the order models are subsequent to video data
When, it can be dedicated for being ranked up to the video played in large-screen mobile phone, so as to have higher sequence precision.
Referring to Fig. 4, the application also provides a kind of ordering system of video quality, the system comprises memories and processing
Device stores computer program and sets of video data in the memory, and it includes at least one set of video that the video data, which is concentrated,
Two video datas of data pair, the video data centering have identical original video data；Wherein, described two videos
At least one video data is that the original video data is obtained by processing in data；The computer program is described
When processor executes, following functions are realized.
S1: one group of target video data pair that the video data is concentrated is read, and successively shows the target to user
The video of two video datas of video data centering characterization.
S3: the user is received for the target video data to the evaluation information delivered, the evaluation information is used for
Determine the target video data centering quality preferably video data.
S5: the video data based on the target video data centering constructs training sample, and the training sample is defeated
Enter order models, obtain ranking results, the ranking results are preferably regarded for characterizing the target video data centering quality
Frequency evidence.
S7: being corrected the order models according to the difference value between the ranking results and the evaluation information,
So that when the training sample to be inputted to the order models after correction again, obtained ranking results and the evaluation information one
It causes.
In the present embodiment, the video of two video datas characterization of the target video data centering is shown to user
Include:
When the corresponding playing duration of the target video data centering video data is more than or equal to specified duration threshold value
When, interception has identical play position and identical playing duration respectively from two video datas of the target video data pair
Video segment data pair, and successively show view that two video segment datas of video segment data centering characterize to user
Frequency segment.
In the present embodiment, when the computer program is executed by the processor, following functions are also realized:
The Segment evaluation information that the user is directed to the video segment data pair is received, the Segment evaluation information is used for
Determine the video segment data centering quality preferably video segment data；
Using video data belonging to the quality preferably video segment data as the target video data centering matter
Amount preferably video data.
In the present embodiment, when the computer program is executed by the processor, following functions are also realized:
The first eigenvector and the second video data of first video data of target video data centering are extracted respectively
Second feature vector, and the first eigenvector based on extraction and second feature vector construct positive training sample and anti-
To training sample；Wherein, first eigenvector described in the positive training sample is located at before the second feature vector, institute
Second feature vector described in reverse train sample is stated to be located at before the first eigenvector.
In the present embodiment, the positive training sample is associated with the first theoretical value, the reverse train sample and
Second theoretical value is associated；Wherein, first theoretical value and second theoretical value be not identical；
Correspondingly, when the computer program is executed by the processor, following functions are also realized:
Respectively by the positive training sample and the reverse train sample input sequencing model, the positive training is obtained
Corresponding first actual value of sample and corresponding second actual value of the reverse train sample；
The first difference value between first actual value and first theoretical value is calculated, and utilizes first difference
Value is corrected the parameter in the order models；
The second difference value between second actual value and second theoretical value is calculated, and utilizes second difference
Value is corrected the parameter in the order models.
In the present embodiment, when the computer program is executed by the processor, following functions are also realized:
Video of the target video data to the characterization environmental information locating when playing is recorded, and the environment is believed
It ceases associated with the evaluation information；
Correspondingly, it is used for based on the order models that the evaluation information corrects in environmental information characterization
Video data in environment is to being ranked up.
In the present embodiment, the memory includes but is not limited to random access memory (Random Access
Memory, RAM), read-only memory (Read-Only Memory, ROM), caching (Cache), hard disk (Hard Disk
Drive, HDD) or storage card (Memory Card).
In the present embodiment, the processor can be implemented in any suitable manner.For example, the processor can be with
Take such as microprocessor or processor and storage can by (micro-) processor execute computer readable program code (such as
Software or firmware) computer-readable medium, logic gate, switch, specific integrated circuit (Application Specific
Integrated Circuit, ASIC), programmable logic controller (PLC) and the form etc. for being embedded in microcontroller.
The specific function that the ordering system for the video quality that this specification embodiment provides, memory and processor are realized
Can, explanation can be contrasted with the aforementioned embodiments in this specification, and the technical effect of aforementioned embodiments can be reached,
Here it just repeats no more.
Therefore technical solution provided by the present application, two views in one group of target video data pair are being shown to user
After the video of frequency data characterization, user can receive for the target video data to the evaluation information delivered.Evaluation letter
Breath can characterize user and feel quality preferably video data.If the quality preferably video data is treated video
Data then show the video editing techniques used when processing by customer acceptance.It is on the contrary then show that user does not recognize adoptable view
Frequency editing technique.In this way, the video data pair that can largely compare can be provided a user, and receive the evaluation of user feedback
Information.Then it can use the video data of these video data centerings as training sample, order models be trained.?
When training, order models can export the ranking results for video data pair.But since initial order models may be less
Accurately, the evaluation information that will lead to ranking results and user feedback is inconsistent.In such a case, it is possible to be tied according to the sequence
Difference value between fruit and the evaluation information is corrected the order models, so that the order models after correction can
The ranking results of the correctly predicted video data pair.In this way, after completing training process, when needs are to new video data
When being ranked up to the quality of the video of characterization, can directly by new video data to the input data as order models,
So as to export the ranking results of the new video data pair by the order models for completing training.Therefore the application
The technical solution of offer can train the higher sequence of precision in conjunction with the method for the evaluation information and machine learning of user
Model, so as to automatically to new video data to being ranked up, improve the efficiency of video quality sequence.
In the 1990s, the improvement of a technology can be distinguished clearly be on hardware improvement (for example,
Improvement to circuit structures such as diode, transistor, switches) or software on improvement (improvement for method flow).So
And with the development of technology, the improvement of current many method flows can be considered as directly improving for hardware circuit.
Designer nearly all obtains corresponding hardware circuit by the way that improved method flow to be programmed into hardware circuit.Cause
This, it cannot be said that the improvement of a method flow cannot be realized with hardware entities module.For example, programmable logic device
(Programmable Logic Device, PLD) (such as field programmable gate array (Field Programmable Gate
Array, FPGA)) it is exactly such a integrated circuit, logic function determines device programming by user.By designer
Voluntarily programming comes a digital display circuit " integrated " on a piece of PLD, designs and makes without asking chip maker
Dedicated IC chip.Moreover, nowadays, substitution manually makes IC chip, this programming is also used instead mostly " is patrolled
Volume compiler (logic compiler) " software realizes that software compiler used is similar when it writes with program development,
And the source code before compiling also write by handy specific programming language, this is referred to as hardware description language
(Hardware Description Language, HDL), and HDL is also not only a kind of, but there are many kind, such as ABEL
(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description
Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL
(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby
Hardware Description Language) etc., VHDL (Very-High-Speed is most generally used at present
Integrated Circuit Hardware Description Language) and Verilog2.Those skilled in the art
It will be apparent to the skilled artisan that only needing method flow slightly programming in logic and being programmed into integrated circuit with above-mentioned several hardware description languages
In, so that it may it is readily available the hardware circuit for realizing the logical method process.
It is also known in the art that the sequence in addition to realizing video quality in a manner of pure computer readable program code
Other than system, completely can by by method and step carry out programming in logic come so that video quality ordering system with logic gate,
Switch, specific integrated circuit, programmable logic controller (PLC) and insertion microcontroller etc. form realize identical function.Therefore this
The ordering system of kind of video quality is considered a kind of hardware component, and to including for realizing various functions in it
Device can also be considered as the structure in hardware component.Or even, both can may be used being considered as realizing the device of various functions
To be that the software module of implementation method can be the structure in hardware component again.
As seen through the above description of the embodiments, those skilled in the art can be understood that the application can
It realizes by means of software and necessary general hardware platform.Based on this understanding, the technical solution essence of the application
On in other words the part that contributes to existing technology can be embodied in the form of software products, the computer software product
It can store in storage medium, such as ROM/RAM, magnetic disk, CD, including some instructions are used so that a computer equipment
(can be personal computer, server or the network equipment etc.) executes each embodiment of the application or embodiment
Method described in certain parts.
Each embodiment in this specification is described in a progressive manner, same and similar between each embodiment
Part may refer to each other, what each embodiment stressed is the difference with other embodiments.In particular, needle
For the embodiment of the ordering system of video quality, it is referred to the introduction control solution of the embodiment of preceding method
It releases.
The application can describe in the general context of computer-executable instructions executed by a computer, such as program
Module.Generally, program module includes routines performing specific tasks or implementing specific abstract data types, programs, objects, group
Part, data structure etc..The application can also be practiced in a distributed computing environment, in these distributed computing environments, by
Task is executed by the connected remote processing devices of communication network.In a distributed computing environment, program module can be with
In the local and remote computer storage media including storage equipment.
Although depicting the application by embodiment, it will be appreciated by the skilled addressee that there are many deformations by the application
With variation without departing from spirit herein, it is desirable to which the attached claims include these deformations and change without departing from the application
Spirit.
Claims (15)
1. a kind of sort method of video quality, which is characterized in that provide sets of video data, it includes extremely that the video data, which is concentrated,
Few one group of video data pair, two video datas of the video data centering have identical original video data；Wherein, institute
Stating at least one video data in two video datas is that the original video data is obtained by processing, the method packet
It includes:
One group of target video data pair that the video data is concentrated is read, and successively shows the target video data to user
The video of two video datas of centering characterization；
The user is received for the target video data to the evaluation information delivered, the evaluation information is described for determining
Target video data centering quality preferably video data；
Video data based on the target video data centering constructs training sample, and by the training sample input sequencing mould
Type obtains ranking results, and the ranking results are for characterizing the target video data centering quality preferably video data；
The order models are corrected according to the difference value between the ranking results and the evaluation information, so that will
When the training sample inputs the order models after correction again, obtained ranking results are consistent with the evaluation information.
2. the method according to claim 1, wherein showing two of the target video data centering to user
Video data characterization video include:
When the corresponding playing duration of the target video data centering video data is more than or equal to specified duration threshold value, from
Interception has the view of identical play position and identical playing duration respectively in two video datas of the target video data pair
Frequency fragment data pair, and the piece of video of two video segment datas of video segment data centering characterization is successively shown to user
Section.
3. according to the method described in claim 2, it is characterized in that, receiving the user for the target video data pair
Evaluation information includes:
The Segment evaluation information that the user is directed to the video segment data pair is received, the Segment evaluation information is for determining
The video segment data centering quality preferably video segment data；
Using video data belonging to the quality preferably video segment data as the target video data centering quality compared with
Excellent video data.
4. according to the method described in claim 2, it is characterized in that, the quantity of the video segment data pair is at least two groups；
Correspondingly, the method also includes:
The user is obtained for the video segment data to the local evaluation information delivered, the part evaluation information is used for
Characterize video data belonging to quality preferably video segment data；
Count the quantity of identical local evaluation information, and using the video data of the most local evaluation information characterization of quantity as
The target video data centering quality preferably video data.
5. the method according to claim 1, wherein the video data structure based on the target video data centering
Building training sample includes:
Extract respectively first video data of target video data centering first eigenvector and the second video data
Two feature vectors, and the first eigenvector based on extraction and second feature vector construct positive training sample and reversed instruction
Practice sample；Wherein, first eigenvector described in the positive training sample is located at before the second feature vector, described anti-
It is located at before the first eigenvector to second feature vector described in training sample.
6. according to the method described in claim 5, it is characterized in that, it is described forward direction training sample it is associated with the first theoretical value,
The reverse train sample is associated with the second theoretical value；Wherein, first theoretical value and second theoretical value be not identical；
Correspondingly, by the training sample input sequencing model, obtaining ranking results includes:
Respectively by the positive training sample and the reverse train sample input sequencing model, the positive training sample is obtained
Corresponding first actual value and corresponding second actual value of the reverse train sample；
Correspondingly, packet is corrected to the order models according to the difference value between the ranking results and the evaluation information
It includes:
The first difference value between first actual value and first theoretical value is calculated, and utilizes first difference value pair
Parameter in the order models is corrected；
The second difference value between second actual value and second theoretical value is calculated, and utilizes second difference value pair
Parameter in the order models is corrected.
7. the method according to claim 1, wherein being carried out by the training sample to the order models
After correction, the method also includes:
By the order models after two video datas input correction wait sort, to obtain two video datas to be sorted
Corresponding ranking results.
8. method according to claim 1 or claim 7, which is characterized in that when the quantity of video data to be sorted is at least three
When a, the method also includes:
The video data to be sorted is formed into sub-video data pair two-by-two, and successively by the sub-video data to input school
Order models after just obtain the ranking results of two video datas of sub-video data centering；
Based on the ranking results of the sub-video data pair, the video data to be sorted is ranked up.
9. the method according to claim 1, wherein being directed to the target video data pair receiving the user
When the evaluation information delivered, the method also includes:
Record the target video data environmental information locating when playing to the video of characterization, and by the environmental information with
The evaluation information is associated；
Correspondingly, it is used for based on the order models that the evaluation information corrects to the environment in environmental information characterization
In video data to being ranked up.
10. a kind of ordering system of video quality, which is characterized in that the system comprises memory and processor, the storage
Computer program and sets of video data are stored in device, it includes at least one set of video data pair that the video data, which is concentrated, described
Two video datas of video data centering have identical original video data；Wherein, in described two video datas at least
One video data is that the original video data is obtained by processing；The computer program is executed by the processor
When, realize following functions:
One group of target video data pair that the video data is concentrated is read, and successively shows the target video data to user
The video of two video datas of centering characterization；
The user is received for the target video data to the evaluation information delivered, the evaluation information is described for determining
Target video data centering quality preferably video data；
Video data based on the target video data centering constructs training sample, and by the training sample input sequencing mould
Type obtains ranking results, and the ranking results are for characterizing the target video data centering quality preferably video data；
The order models are corrected according to the difference value between the ranking results and the evaluation information, so that will
When the training sample inputs the order models after correction again, obtained ranking results are consistent with the evaluation information.
11. system according to claim 10, which is characterized in that show the two of the target video data centering to user
The video of a video data characterization includes:
When the corresponding playing duration of the target video data centering video data is more than or equal to specified duration threshold value, from
Interception has the view of identical play position and identical playing duration respectively in two video datas of the target video data pair
Frequency fragment data pair, and the piece of video of two video segment datas of video segment data centering characterization is successively shown to user
Section.
12. system according to claim 11, which is characterized in that when the computer program is executed by the processor,
Also realize following functions:
The Segment evaluation information that the user is directed to the video segment data pair is received, the Segment evaluation information is for determining
The video segment data centering quality preferably video segment data；
Using video data belonging to the quality preferably video segment data as the target video data centering quality compared with
Excellent video data.
13. system according to claim 10, which is characterized in that when the computer program is executed by the processor,
Also realize following functions:
Extract respectively first video data of target video data centering first eigenvector and the second video data
Two feature vectors, and the first eigenvector based on extraction and second feature vector construct positive training sample and reversed instruction
Practice sample；Wherein, first eigenvector described in the positive training sample is located at before the second feature vector, described anti-
It is located at before the first eigenvector to second feature vector described in training sample.
14. system according to claim 13, which is characterized in that the forward direction training sample is related to the first theoretical value
Connection, the reverse train sample are associated with the second theoretical value；Wherein, first theoretical value and second theoretical value not phase
Together；
Correspondingly, when the computer program is executed by the processor, following functions are also realized:
Respectively by the positive training sample and the reverse train sample input sequencing model, the positive training sample is obtained
Corresponding first actual value and corresponding second actual value of the reverse train sample；
The first difference value between first actual value and first theoretical value is calculated, and utilizes first difference value pair
Parameter in the order models is corrected；
The second difference value between second actual value and second theoretical value is calculated, and utilizes second difference value pair
Parameter in the order models is corrected.
15. system according to claim 10, which is characterized in that when the computer program is executed by the processor,
Also realize following functions:
Record the target video data environmental information locating when playing to the video of characterization, and by the environmental information with
The evaluation information is associated；
Correspondingly, it is used for based on the order models that the evaluation information corrects to the environment in environmental information characterization
In video data to being ranked up.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710811287.XA CN109495772B (en) | 2017-09-11 | 2017-09-11 | Video quality sequencing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710811287.XA CN109495772B (en) | 2017-09-11 | 2017-09-11 | Video quality sequencing method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109495772A true CN109495772A (en) | 2019-03-19 |
| CN109495772B CN109495772B (en) | 2021-10-15 |
Family
ID=65687646
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710811287.XA Active CN109495772B (en) | 2017-09-11 | 2017-09-11 | Video quality sequencing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109495772B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114741048A (en) * | 2022-05-20 | 2022-07-12 | 中译语通科技股份有限公司 | Sample sorting method and device, computer equipment and readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101282481A (en) * | 2008-05-09 | 2008-10-08 | 中国传媒大学 | Method for evaluating video quality based on artificial neural net |
| 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 |
| CN101540048A (en) * | 2009-04-21 | 2009-09-23 | 北京航空航天大学 | Image quality evaluating method based on support vector machine |
| CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
| CN104318562A (en) * | 2014-10-22 | 2015-01-28 | 百度在线网络技术（北京）有限公司 | Method and device for confirming quality of internet images |
| US20150120624A1 (en) * | 2013-10-30 | 2015-04-30 | Sony Corporation | Apparatus and method for information processing |
-
2017
- 2017-09-11 CN CN201710811287.XA patent/CN109495772B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101282481A (en) * | 2008-05-09 | 2008-10-08 | 中国传媒大学 | Method for evaluating video quality based on artificial neural net |
| 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 |
| CN101540048A (en) * | 2009-04-21 | 2009-09-23 | 北京航空航天大学 | Image quality evaluating method based on support vector machine |
| CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
| US20150120624A1 (en) * | 2013-10-30 | 2015-04-30 | Sony Corporation | Apparatus and method for information processing |
| CN104318562A (en) * | 2014-10-22 | 2015-01-28 | 百度在线网络技术（北京）有限公司 | Method and device for confirming quality of internet images |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114741048A (en) * | 2022-05-20 | 2022-07-12 | 中译语通科技股份有限公司 | Sample sorting method and device, computer equipment and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109495772B (en) | 2021-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107707931B (en) | Method and device for generating interpretation data according to video data, method and device for synthesizing data and electronic equipment | |
| CN110012302A (en) | A kind of network direct broadcasting monitoring method and device, data processing method | |
| CN110730387B (en) | Video playing control method and device, storage medium and electronic device | |
| US9002175B1 (en) | Automated video trailer creation | |
| CN107707967A (en) | The determination method, apparatus and computer-readable recording medium of a kind of video file front cover | |
| CN109168026A (en) | Instant video display methods, device, terminal device and storage medium | |
| CN107615766A (en) | System and method for creating and distributing content of multimedia | |
| CN106303669B (en) | A kind of video clipping method and device | |
| CN105874780A (en) | Method and apparatus for generating a text color for a group of images | |
| CN104091607B (en) | Video editing method and device based on IOS equipment | |
| CN110019933A (en) | Video data handling procedure, device, electronic equipment and storage medium | |
| WO2019169979A1 (en) | Film and television works production method, apparatus, and device | |
| CN106572380A (en) | User terminal and video dynamic thumbnail generating method | |
| US20190364211A1 (en) | System and method for editing video contents automatically technical field | |
| US20140104494A1 (en) | Relational display of images | |
| CN104038848A (en) | Video processing method and video processing device | |
| CN107920274A (en) | A kind of method for processing video frequency, client and server | |
| CN107197381A (en) | Temperature curve generation method, device and the readable storage medium storing program for executing of television video | |
| CN105979346A (en) | Method and device for playing streaming media | |
| CN109922334A (en) | A kind of recognition methods and system of video quality | |
| CN104091608B (en) | A kind of video editing method and device based on ios device | |
| CN104581224B (en) | Switch the method, apparatus and terminal of broadcasting content | |
| CN111077992B (en) | Click-to-read method, electronic equipment and storage medium | |
| CN106412674B (en) | The screenshot method and its device of dynamic video | |
| Husa et al. | HOST-ATS: automatic thumbnail selection with dashboard-controlled ML pipeline and dynamic user survey |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200514 Address after: 310052 room 508, floor 5, building 4, No. 699, Wangshang Road, Changhe street, Binjiang District, Hangzhou City, Zhejiang Province Applicant after: Alibaba (China) Co.,Ltd. Address before: 100080 Beijing Haidian District city Haidian street A Sinosteel International Plaza No. 8 block 5 layer A, C Applicant before: Youku network technology (Beijing) Co., Ltd |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |