CN109886963A - A kind of image processing method and system - Google Patents
A kind of image processing method and system Download PDFInfo
- Publication number
- CN109886963A CN109886963A CN201910243380.4A CN201910243380A CN109886963A CN 109886963 A CN109886963 A CN 109886963A CN 201910243380 A CN201910243380 A CN 201910243380A CN 109886963 A CN109886963 A CN 109886963A
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- picture
- fuzziness
- frame
- video
- 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
Landscapes
- Image Analysis (AREA)
- Television Signal Processing For Recording (AREA)
Abstract
The present invention provides a kind of image processing method and system, this method comprises: carrying out interval to video to be processed takes out frame processing, obtains N frame picture, N is positive integer.Using N frame picture as the input of image processing model, obtains length and be N and include the fuzzy degree series that every frame picture belongs to the probability value of fuzzy picture, wherein image processing model is based on picture sample training convolutional neural networks model and obtains.The fuzziness of every frame picture is compared with preset fuzziness threshold value, obtains the fuzzy fragment list of video to be processed.It in scheme provided by the invention, handles to obtain multiframe picture by carrying out pumping frame to video to be processed, multiframe picture is inputted in trained image processing model in advance and be handled, the fuzzy degree series that every frame picture belongs to the probability value of fuzzy picture are obtained.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing cost, cost of labor can be reduced and improve video treatment effeciency.
Description
Technical field
The present invention relates to technical field of image processing, and in particular to a kind of image processing method and system.
Background technique
With the development of science and technology, the video programs such as TV, film and variety early have become people life in can not or
Scarce a part, and being presented on the video program of spectators at the moment is video material after treatment.These raw video materials
Due to being the video recorded using means such as different seats in the plane and different angles, original video is plain during recording in this way
Material can have a large amount of flaw segment, so needing to carry out raw video material repeatedly processing after the completion of recording could be presented
Before spectators, one of the processing mode to raw video material is exactly to identify waste paper and cut off these useless video clips.
Wherein, most commonly seen waste paper form is exactly the fuzzy segment in raw video material.
Fuzzy segment, referring to causes since the camera lens of recorded video is acutely shaken or fast moving for photographic subjects
Record fuzzy pictures.The means of identification waste paper are usually first to go to identify fuzzy segment by primary editor and mark these moulds at present
Segment is pasted, finer below cut then is carried out to the raw video material after carrying out preliminary treatment by advanced editor again
Volume, so that video can be showed preferably.But during performance recording, the programs such as variety especially in recent years
It records, usually exists simultaneously tens even a seats in the plane up to a hundred and shooting, the program of finally show a hour may
It is to be come out from raw video material editing in up to a hundred hours, editor's needs primary in this way go to browse the original of up to a hundred hours time
Beginning video material goes identify fuzzy segment, marks the preliminary preparations such as fuzzy segment.Will cause in this way the plenty of time at
The waste of this and personnel cost, influences the working efficiency of post-production.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of image processing method and system, to solve existing image procossing
Video existing for method handles the problems such as time cost is high, high labor cost and working efficiency are low.
To achieve the above object, the embodiment of the present invention provides the following technical solutions:
First aspect of the embodiment of the present invention discloses a kind of image processing method, which comprises
Interval is carried out to video to be processed and takes out frame processing, obtains N frame picture, N is positive integer；
Using the N frame picture as the input of image processing model, the fuzzy degree series that length is N are obtained, it is described fuzzy
Degree series include the fuzziness of every frame picture, and the fuzziness is the probability value that every frame picture belongs to fuzzy picture, described image
Processing model is based on picture sample training convolutional neural networks model and obtains, and the fog-level of the fuzzy picture is to set in advance
It is fixed；
The fuzziness of every frame picture is compared with preset fuzziness threshold value, obtains the video to be processed
Fuzzy fragment list.
Preferably, described image processing model is obtained based on picture sample training convolutional neural networks model, comprising:
The picture sample is inputted in the convolutional neural networks model built in advance and is identified, is obtained described
The fuzziness of picture sample；
Calculate the error between the fuzziness of the picture sample and the realistic blur degree of the picture sample；
Judge whether the error is less than error threshold；
If so, calculating weight used in the fuzziness of the picture sample based on the convolutional neural networks, institute is established
State image processing model；
If it is not, the weight based on convolutional neural networks described in the error transfer factor, based on the convolutional Neural after adjustment weight
Network identifies the picture sample to obtain the fuzziness of picture sample again, calculates the picture identified again
Error between the fuzziness of sample and the realistic blur degree of the picture sample, until the error is less than the error threshold
Value calculates weight used in the fuzziness of picture sample based on current convolutional neural networks, establishes described image processing model.
Preferably, described to carry out interval pumping frame processing to video to be processed, the process for obtaining N frame picture includes:
Based on the length of the video to be processed, frame mode is taken out according to interval and obtains N frame picture, whereinL is the length of the video to be treated,To be rounded formula downwards.
Preferably, the fuzziness of every frame picture is compared with preset fuzziness threshold value, is obtained described wait locate
Manage fuzzy fragment list final in video, comprising:
Gaussian smoothing is carried out to the fuzzy degree series；
Obtain the fuzzy picture of M frame that fuzziness in the fuzzy sequence is greater than the fuzziness threshold value；
The M frame is obscured into picture and forms 0 or more the fuzzy segment comprising a fuzzy picture, or includes m serial number
The fuzzy segment of adjacent fuzzy picture, wherein M is the positive integer for being less than or equal to N more than or equal to 0, and m is the integer greater than 1；
The fuzzy segment is formed into fuzzy fragment list；
Fuzzy segment in the fuzzy fragment list is merged and deleted, the final fuzzy segment column are obtained
Table.
Preferably, the fuzzy segment in the fuzzy fragment list is merged and is deleted, comprising:
If there are the fuzzy segments that clip durations are less than preset duration in the fuzzy fragment list, delete described fuzzy
Segment；
If in the fuzzy fragment list, there are two adjacent fuzzy segments, and described two adjacent fuzzy pieces
It is separated by the picture of default frame number between section, then merging described two adjacent fuzzy segments becomes a fuzzy segment.
Second aspect of the embodiment of the present invention discloses a kind of image processing system, the system comprises:
Frame unit is taken out, frame processing is taken out for carrying out interval to video to be processed, obtains N frame picture, N is positive integer；
Computing unit, for the mould that length is N to be calculated using the N frame picture as the input of image processing model
Degree series are pasted, the fuzzy degree series include the fuzziness of every frame picture, and the fuzziness is that every frame picture belongs to fuzzy picture
Probability value, described image processing model be based on picture sample training convolutional neural networks model obtain, the fuzzy picture
Fog-level is to preset；
Processing unit obtains institute for the fuzziness of every frame picture to be compared with preset fuzziness threshold value
State the fuzzy fragment list of video to be processed.
Preferably, the computing unit includes:
Input module is carried out for inputting the picture sample in the convolutional neural networks model built in advance
Identification, obtains the fuzziness of the picture sample；
Error module, for calculating between the fuzziness of the picture sample and the realistic blur degree of the picture sample
Error；
Judgment module establishes module if so, executing, if it is not, then for judging whether the error is less than error threshold
Execute loop module；
Module is established, for calculating power used in the fuzziness of the picture sample based on the convolutional neural networks
Weight establishes described image processing model；
Loop module, for the weight based on convolutional neural networks described in the error transfer factor, after adjustment weight
Convolutional neural networks identify the picture sample to obtain the fuzziness of picture sample again, and calculating carries out identifying again
Error between the fuzziness of the picture sample arrived and the realistic blur degree of the picture sample, until the error is less than described
Error threshold calculates weight used in the fuzziness of picture sample based on current convolutional neural networks, establishes at described image
Manage model.
Preferably, the pumping frame unit is specifically used for the length based on the video to be processed, takes out frame mode according to interval
Obtain N frame picture, whereinL is the length of the video to be treated,For to
Lower rounding formula.
Preferably, the processing unit includes:
Gaussian smoothing module, for carrying out Gaussian smoothing to the fuzzy degree series；
Module is obtained, the M frame that the fuzziness threshold value is greater than for obtaining fuzziness in the fuzzy sequence obscures picture；
First composite module forms 0 or more the fuzzy piece comprising a fuzzy picture for the M frame to be obscured picture
Section, or the fuzzy segment comprising the adjacent fuzzy picture of m serial number, wherein M is the positive integer for being less than or equal to N more than or equal to 0, m
For the integer greater than 1；
Second composite module, for the fuzzy segment to be formed fuzzy fragment list；
Processing module, for the fuzzy segment in the fuzzy fragment list to be merged and is deleted, obtain it is described most
Whole fuzzy fragment list.
Preferably, the processing module includes:
Submodule is deleted, if for there are the fuzzy pieces that clip durations are less than preset duration in the fuzzy fragment list
Section, then delete the fuzzy segment；
Merge submodule, if in the fuzzy fragment list, there are two adjacent fuzzy segments, and described two
It is separated by the picture of default frame number between a adjacent fuzzy segment, then merging described two adjacent fuzzy segments becomes one
Fuzzy segment.
A kind of image processing method and system provided based on the embodiments of the present invention, this method comprises: to be processed
Video carries out interval and takes out frame processing, obtains N frame picture, N is positive integer.Using N frame picture as the input of image processing model, meter
It calculates and obtains the fuzzy degree series that length is N, which includes the fuzziness of every frame picture, and fuzziness is every frame picture category
In the probability value of fuzzy picture, image processing model is based on picture sample training convolutional neural networks model and obtains.Every frame is drawn
The fuzziness in face is compared with preset fuzziness threshold value, obtains the fuzzy fragment list of video to be processed.The present invention provides
Scheme in, by video to be processed carry out take out frame handle to obtain multiframe picture, by multiframe picture input in advance it is trained
It is handled in image processing model, obtains the fuzzy degree series that every frame picture belongs to the probability value of fuzzy picture.By every frame picture
Fuzziness and threshold value comparison obtain the fuzzy fragment list in video to be processed.Can reduce video processing cost, cost of labor and
Improve video treatment effeciency.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is a kind of flow chart of image processing method provided in an embodiment of the present invention；
Fig. 2 is that training convolutional neural networks provided in an embodiment of the present invention obtain the flow chart of image processing model；
Fig. 3 is the flow chart provided in an embodiment of the present invention for obtaining fuzzy fragment list final in video to be processed；
Fig. 4 is a kind of structural block diagram of image processing system provided in an embodiment of the present invention；
Fig. 5 is a kind of structural block diagram of image processing system provided in an embodiment of the present invention；
Fig. 6 is a kind of structural block diagram of image processing system provided in an embodiment of the present invention；
Fig. 7 is a kind of structural block diagram of image processing system provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
In this application, the terms "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion,
So that the process, method, article or equipment for including a series of elements not only includes those elements, but also including not having
The other element being expressly recited, or further include for elements inherent to such a process, method, article, or device.Do not having
There is the element limited in the case where more limiting by sentence "including a ...", it is not excluded that in the mistake including the element
There is also other identical elements in journey, method, article or equipment.
It can be seen from background technology that identify that the means of waste paper are usually first to be gone to identify fuzzy segment simultaneously by primary editor at present
These fuzzy segments are marked, then the raw video material after carrying out preliminary treatment is carried out below more by advanced editor again
Fine editing, so that video can be showed preferably.But during performance recording, especially in recent years comprehensive
The performance recordings such as skill usually exist simultaneously tens even a seats in the plane up to a hundred and are shooting, finally show a hour
Program may be to come out from raw video material editing in up to a hundred hours, and editor's needs primary in this way go to browse hours up to a hundred
The raw video material of time goes identify fuzzy segment, marks the preliminary preparations such as fuzzy segment.It will cause so big
The waste for measuring time cost and personnel cost, influences the working efficiency of post-production.
Therefore, a kind of image processing method disclosed by the embodiments of the present invention and system, this method comprises: by to be processed
Video carries out interval and takes out frame processing, obtains N frame picture, N is positive integer.Using N frame picture as the input of image processing model, meter
Calculation obtains length and is N and includes the fuzzy degree series of the fuzziness of every frame picture, wherein fuzziness is that every frame picture belongs to mould
The probability value of picture is pasted, image processing model is based on picture sample training convolutional neural networks model and obtains.By every frame picture
Fuzziness is compared with preset fuzziness threshold value, obtains the fuzzy fragment list of video to be processed.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
With reference to Fig. 1, a kind of flow chart of image processing method provided in an embodiment of the present invention is shown, the method includes
Following steps:
Step S101: interval is carried out to video to be processed and takes out frame processing, obtains N frame picture.
During implementing step S101, based on the length of the video to be processed, frame mode is taken out according to interval
N frame picture is obtained, N is positive integer.The value of N is calculated according to formula (1).When carrying out taking out frame processing, by every frame picture in institute
It states the temporal information in video to be processed and marks out.
The formula (1) are as follows:
Wherein, L is the length of the video to be treated,To be rounded formula downwards.
It should be noted that the downward rounding formula is downward round numbers, i.e., only retain integer part, deletes decimal point
Part, for example 3.3 take 3,4.7 to take 4.The above-mentioned video being related to takes out frame mode and includes but are not limited to be rounded formula downwards.
Step S102: using the N frame picture as the input of image processing model, the fuzziness that length is N is calculated
Sequence, the fuzzy sequence include the fuzziness of every frame picture.
During implementing step S102, the fuzziness is the probability value that every frame picture belongs to fuzzy picture,
Described image processing model is based on picture sample training convolutional neural networks model and obtains, and the fog-level of the fuzzy picture is
It presets.
Step S103: the fuzziness of every frame picture is compared with preset fuzziness threshold value, obtain it is described to
Handle the fuzzy fragment list of video.
During implementing step S103, Fuzzy Threshold is preset, fuzziness is greater than the fuzziness threshold
The picture of value handles obtained fuzzy picture to obtain final fuzzy fragment list as fuzzy picture.
It should be noted that corresponding fuzzy being specifically defined for picture is configured according to the actual situation by technical staff.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
What is be related in the step S102 that above-mentioned Fig. 1 is disclosed obtains figure based on picture sample training convolutional neural networks model
As the process of processing model shows training convolutional neural networks provided in an embodiment of the present invention and obtain image procossing with reference to Fig. 2
The flow chart of model, comprising the following steps:
Step S201: the picture sample is inputted in the convolutional neural networks model built in advance and is known
Not, the fuzziness of the picture sample is obtained.
During implementing step S201, the picture sample includes the clear picture and fuzzy graph of preset quantity
Piece is in advance labeled the fuzziness of the clear picture and blurred picture.
It should be noted that the acquisition of the clear picture and blurred picture is obtained according to the actual situation by technical staff
It takes.The selection of convolutional neural networks model is selected according to the actual situation by technical staff, such as selection low complex degree
MobileNet network.
Step S202: the mistake between the fuzziness of the picture sample and the realistic blur degree of the picture sample is calculated
Difference.
During implementing step S202, the realistic blur degree of the picture sample is in the step S201
In mark in advance.
Step S203: judging whether the error is less than error threshold, if so, S204 is thened follow the steps, if it is not, then executing
Step S205.
During implementing step S203, the error threshold is set according to the actual situation by technical staff
It sets.
Step S204: weight used in the fuzziness of the picture sample is calculated based on the convolutional neural networks, is built
Vertical described image handles model.
Step S205: the weight based on convolutional neural networks described in the error transfer factor, based on the convolution after adjustment weight
Neural network identifies the picture sample to obtain the fuzziness of picture sample again, what calculating was identified again
Error between the fuzziness of picture sample and the realistic blur degree of the picture sample, until the error is less than the error
Threshold value calculates weight used in the fuzziness of picture sample based on current convolutional neural networks, establishes described image processing mould
Type.
During implementing step S205, the convolutional neural networks mould is constantly trained using the picture sample
Type, until the error between the result and legitimate reading of convolutional neural networks model output is less than the error threshold, base
Weight used in the fuzziness of the picture sample is calculated in the convolutional neural networks last time, is established at described image
Manage model.
Preferably, when fuzzy segment in practical application described image processing model identification video, technical staff can be with
According to the weight of the corresponding practical fuzziness real-time update described image processing model of recognition result and the video, it is continuously improved
The accuracy of identification of described image processing model.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
Final fuzzy fragment list in the acquisition video to be processed being related in the step S103 that above-mentioned Fig. 1 is disclosed
Process show the stream provided in an embodiment of the present invention for obtaining fuzzy fragment list final in video to be processed with reference to Fig. 3
Cheng Tu, comprising the following steps:
Step S301: Gaussian smoothing is carried out to the fuzzy degree series.
During implementing step S301, Gaussian smoothing is carried out to the fuzzy degree series, removal is abnormal
Mutation value, for example remove between two frame pictures and occur that a fuzziness is very low or very high picture.
Step S302: the fuzzy picture of M frame that fuzziness in the fuzzy sequence is greater than the fuzziness threshold value is obtained.
During implementing step S302, by the fuzziness and mould of each frame picture in the fuzzy degree series
Paste degree threshold value is compared, and the M frame for obtaining fuzziness greater than the fuzziness threshold value obscures picture.M be more than or equal to 0 be less than etc.
In the positive integer of N.
Step S303: the M frame is obscured into picture and forms 0 or more the fuzzy segment comprising a fuzzy picture, or packet
The fuzzy segment of the adjacent fuzzy picture containing m serial number.
During implementing step S303, after obtaining the N frame picture, in order to every in N frame picture
One picture carries out serial number sequence.When the acquisition M frame obscures picture, the M frame is obscured into the m mould that serial number is adjacent in picture
It pastes picture and forms a fuzzy segment, form one with other fuzzy non-conterminous independent fuzzy pictures of panel number and obscure
Segment.M is the integer greater than 1.
Step S304: the fuzzy segment is formed into fuzzy fragment list.
During implementing step S304, fuzzy segment column will be made of fuzzy segment acquired in step s303
Table.
Step S305: merging and delete to the fuzzy segment in the fuzzy fragment list, obtains described final
Fuzzy fragment list.
During implementing step S305, if be less than default there are clip durations in the fuzzy fragment list
Long fuzzy segment then deletes the fuzzy segment.If in the fuzzy fragment list, there are two adjacent fuzzy segments,
And it is separated by the picture of default frame number between described two adjacent fuzzy segments, then closes described two adjacent fuzzy segments
And become a fuzzy segment.
It should be noted that the above-mentioned preset duration being related to and threshold value are carried out according to the actual situation by technical staff
Setting.
The final fuzzy fragment list how is obtained more preferably to illustrate, is lifted below by process A1-A3
Example explanation:
A1, video to be processed is carried out to obtain 9 frame pictures after taking out frame processing, after 9 frame picture input pictures are handled model
Obtain the fuzzy degree series comprising every frame fuzzy pictures degree are as follows: [0.7,0.8,0.5,0.3,0.5,0.9,0.1,0.2,0.9], sequence
Number be respectively 1-9.
A2, assume that fuzziness threshold value is 0.5, then the fuzzy picture more than or equal to 0.5 is [1,2,3,5,6,9].Composition 3
Fuzzy segment, respectively (start:1, stop:3), (start:5, stop:6) and (start:9, stop:9).
A3, assume to be separated by 1 frame picture between two fuzzy two adjacent fuzzy segments of segment, then it will be described two adjacent
Fuzzy segment merge become a fuzzy segment.Delete fuzzy segment of the clip durations less than 1 second.Assuming that fuzzy segment
The clip durations of (start:9, stop:9) are 0.5 second.Then by fuzzy segment (start:1, stop:3) and (start:5,
Stop:6 it) is merged into (start:1, stop:6), fuzzy segment (start:9, stop:9) is deleted.Final fuzzy segment column
Table includes fuzzy segment (start:1, stop:6).
It should be noted that the content that above process A1-A3 is related to is only limitted to for example, forming the side of fuzzy segment
Formula includes but is not limited only to the mode that above process A1-A3 is related to.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
Corresponding with a kind of image processing method that the embodiments of the present invention provide, with reference to Fig. 4, the embodiment of the present invention is also
Provide a kind of structural block diagram of image processing system, comprising: take out frame unit 401, computing unit 402 and processing unit 403.
Frame unit 401 is taken out, frame processing is taken out for carrying out interval to video to be processed, obtains N frame picture, N is positive integer.Tool
Hold the corresponding content of step S101 disclosed referring to embodiments of the present invention Fig. 1 in vivo.
Preferably, the pumping frame unit 401 be specifically used for the length based on the video to be processed, according to formula (1) into
Row takes out frame and handles to obtain N frame picture.
Computing unit 402, for using the N frame picture as the input of image processing model, it to be N's that length, which is calculated,
Fuzzy degree series, the fuzzy degree series include the fuzziness of every frame picture, wherein the fuzziness is that every frame picture belongs to mould
The probability value of picture is pasted, described image processing model is based on picture sample training convolutional neural networks model and obtains, described fuzzy
The fog-level of picture is to preset.Particular content is corresponding referring to the step S102 that embodiments of the present invention Fig. 1 is disclosed
Content.
Processing unit 403 is obtained for the fuzziness of every frame picture to be compared with preset fuzziness threshold value
The fuzzy fragment list of the video to be processed.The step S103 phase that particular content is disclosed referring to embodiments of the present invention Fig. 1
Corresponding content.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
With reference to Fig. 5, a kind of structural block diagram of image processing system provided in an embodiment of the present invention is shown, the calculating is single
Member 402 includes: input module 4021, error module 4022, judgment module 4023, establishes module 4024 and loop module 4025.
Input module 4021, for inputting the picture sample in the convolutional neural networks model built in advance
It is identified, obtains the fuzziness of the picture sample.The step that particular content is disclosed referring to embodiments of the present invention Fig. 2
The corresponding content of S201.
Error module 4022, for calculate the picture sample fuzziness and the picture sample realistic blur degree it
Between error.The corresponding content of step S202 that particular content is disclosed referring to embodiments of the present invention Fig. 2.
Judgment module 4023 establishes module if so, executing for judging whether the error is less than error threshold
4024, if it is not, then executing loop module 4025.The step S203 phase that particular content is disclosed referring to embodiments of the present invention Fig. 2
Corresponding content.
Module 4024 is established, for being calculated used in the fuzziness of the picture sample based on the convolutional neural networks
Weight establishes described image processing model.Particular content is corresponding referring to the step S204 that embodiments of the present invention Fig. 2 is disclosed
Content.
Loop module 4025, for the weight based on convolutional neural networks described in the error transfer factor, based on adjustment weight
Convolutional neural networks afterwards identify the picture sample to obtain the fuzziness of the picture sample, and calculating is known again
Error between the fuzziness for the picture sample not obtained and the realistic blur degree of the picture sample, until the error is less than
The error threshold calculates weight used in the fuzziness of picture sample based on current convolutional neural networks, establishes the figure
As processing model.The corresponding content of step S205 that particular content is disclosed referring to embodiments of the present invention Fig. 2.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
With reference to Fig. 6, a kind of structural block diagram of image processing system provided in an embodiment of the present invention is shown, the processing is single
Member 403 includes: Gaussian smoothing module 4031, obtains module 4032, the first composite module 4033, the second composite module 4034 and place
Manage module 4035.
Gaussian smoothing module 4031, for carrying out Gaussian smoothing to the fuzzy degree series.Particular content is referring to above-mentioned
The corresponding content of step S301 that inventive embodiments Fig. 3 is disclosed.
Module 4032 is obtained, the M frame that the fuzziness threshold value is greater than for obtaining fuzziness in the fuzzy sequence obscures
Picture.The corresponding content of step S302 that particular content is disclosed referring to embodiments of the present invention Fig. 3.
First composite module 4033 forms 0 or more the mould comprising a fuzzy picture for the M frame to be obscured picture
Paste segment, or the fuzzy segment comprising the adjacent fuzzy picture of m serial number, wherein M is just whole less than or equal to N more than or equal to 0
Number, m are the integer greater than 1.The corresponding content of step S303 that particular content is disclosed referring to embodiments of the present invention Fig. 3.
Second composite module 4034, for the fuzzy segment to be formed fuzzy fragment list.Particular content is referring to above-mentioned
The corresponding content of step S304 that Fig. 3 of the embodiment of the present invention is disclosed.
Processing module 4035 obtains institute for the fuzzy segment in the fuzzy fragment list to be merged and deleted
State final fuzzy fragment list.Particular content is corresponding interior referring to the step S305 that embodiments of the present invention Fig. 3 is disclosed
Hold.
In embodiments of the present invention, it handles to obtain multiframe picture by carrying out pumping frame to video to be processed, by multiframe picture
It inputs in trained image processing model in advance and handles, obtain the fuzziness sequence that every frame picture belongs to the probability value of fuzzy picture
Column.By the fuzziness and threshold value comparison of every frame picture, the fuzzy fragment list in video to be processed is obtained.Video processing can be reduced
Cost, cost of labor and raising video treatment effeciency.
With reference to Fig. 7, a kind of structural block diagram of image processing system provided in an embodiment of the present invention, the processing mould are shown
Block 4035 includes: to delete submodule 40351 and merging submodule 40352.
Submodule 40351 is deleted, if for there are the moulds that clip durations are less than preset duration in the fuzzy fragment list
Segment is pasted, then deletes the fuzzy segment.Particular content is corresponding referring to the step S305 that embodiments of the present invention Fig. 3 is disclosed
Content.
Merge submodule 40352, if in the fuzzy fragment list, there are two adjacent fuzzy segments, and
It is separated by the picture of default frame number between described two adjacent fuzzy segments, then is merged into described two adjacent fuzzy segments
For a fuzzy segment.The corresponding content of step S305 that particular content is disclosed referring to embodiments of the present invention Fig. 3.
In conclusion the present invention provides a kind of image processing method and system, this method comprises: carrying out to video to be processed
Frame processing is taken out at interval, obtains N frame picture, N is positive integer.Using N frame picture as the input of image processing model, obtaining length is
N and include the fuzzy degree series that every frame picture belongs to the probability value of fuzzy picture, wherein image processing model is based on picture sample
This training convolutional neural networks model obtains.The fuzziness of every frame picture is compared with preset fuzziness threshold value, is obtained
The fuzzy fragment list of video to be processed.In scheme provided by the invention, handle to obtain by carrying out pumping frame to video to be processed
Multiframe picture is inputted in trained image processing model in advance and is handled, obtained every frame picture and belong to fuzzy picture by multiframe picture
The fuzzy degree series of the probability value in face.By the fuzziness and threshold value comparison of every frame picture, the fuzzy piece in video to be processed is obtained
Duan Liebiao.Video processing cost, cost of labor can be reduced and improve video treatment effeciency.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for system or
For system embodiment, since it is substantially similar to the method embodiment, so describing fairly simple, related place is referring to method
The part of embodiment illustrates.System and system embodiment described above is only schematical, wherein the conduct
The unit of separate part description may or may not be physically separated, component shown as a unit can be or
Person may not be physical unit, it can and it is in one place, or may be distributed over multiple network units.It can root
According to actual need that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.Ordinary skill
Personnel can understand and implement without creative efforts.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think beyond the scope of this invention.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (10)
1. a kind of image processing method, which is characterized in that the described method includes:
Interval is carried out to video to be processed and takes out frame processing, obtains N frame picture, N is positive integer；
Using the N frame picture as the input of image processing model, the fuzzy degree series that length is N, the fuzziness sequence are obtained
Column include the fuzziness of every frame picture, and the fuzziness is the probability value that every frame picture belongs to fuzzy picture, described image processing
Model is based on picture sample training convolutional neural networks model and obtains, and the fog-level of the fuzzy picture is to preset；
The fuzziness of every frame picture is compared with preset fuzziness threshold value, obtains the fuzzy of the video to be processed
Fragment list.
2. the method according to claim 1, wherein described image processing model is based on picture sample training convolutional
Neural network model obtains, comprising:
The picture sample is inputted in the convolutional neural networks model built in advance and is identified, the picture is obtained
The fuzziness of sample；
Calculate the error between the fuzziness of the picture sample and the realistic blur degree of the picture sample；
Judge whether the error is less than error threshold；
If so, calculating weight used in the fuzziness of the picture sample based on the convolutional neural networks, the figure is established
As processing model；
If it is not, the weight based on convolutional neural networks described in the error transfer factor, based on the convolutional neural networks after adjustment weight
The picture sample is identified to obtain the fuzziness of picture sample again, calculates the picture sample identified again
Fuzziness and the picture sample realistic blur degree between error, until the error be less than the error threshold, base
Weight used in the fuzziness of picture sample is calculated in current convolutional neural networks, establishes described image processing model.
3. being obtained the method according to claim 1, wherein described carry out interval pumping frame processing to video to be processed
Process to N frame picture includes:
Based on the length of the video to be processed, frame mode is taken out according to interval and obtains N frame picture, wherein
L is the length of the video to be treated,To be rounded formula downwards.
4. the method according to claim 1, wherein by the fuzziness of every frame picture and preset fuzziness
Threshold value is compared, and obtains fuzzy fragment list final in the video to be processed, comprising:
Gaussian smoothing is carried out to the fuzzy degree series；
Obtain the fuzzy picture of M frame that fuzziness in the fuzzy sequence is greater than the fuzziness threshold value；
The M frame is obscured into picture and forms 0 or more the fuzzy segment comprising a fuzzy picture, or is adjacent comprising m serial number
Fuzzy picture fuzzy segment, wherein M is the positive integer for being less than or equal to N more than or equal to 0, and m is integer greater than 1；
The fuzzy segment is formed into fuzzy fragment list；
Fuzzy segment in the fuzzy fragment list is merged and deleted, the final fuzzy fragment list is obtained.
5. according to the method described in claim 4, it is characterized in that, the fuzzy segment in the fuzzy fragment list into
Row merges and deletes, comprising:
If deleting the fuzzy piece there are the fuzzy segment that clip durations are less than preset duration in the fuzzy fragment list
Section；
If in the fuzzy fragment list, there are two adjacent fuzzy segments, and described two adjacent fuzzy segments it
Between be separated by the picture of default frame number, then merging described two adjacent fuzzy segments becomes a fuzzy segment.
6. a kind of image processing system, which is characterized in that the system comprises:
Frame unit is taken out, frame processing is taken out for carrying out interval to video to be processed, obtains N frame picture, N is positive integer；
Computing unit, for the fuzziness that length is N to be calculated using the N frame picture as the input of image processing model
Sequence, the fuzzy degree series include the fuzziness of every frame picture, and the fuzziness is that every frame picture belongs to the general of fuzzy picture
Rate value, described image processing model are based on picture sample training convolutional neural networks model and obtain, and the fuzzy picture obscures
Degree is to preset；
Processing unit, for the fuzziness of every frame picture to be compared with preset fuzziness threshold value, obtain it is described to
Handle the fuzzy fragment list of video.
7. system according to claim 6, which is characterized in that the computing unit includes:
Input module is known for inputting the picture sample in the convolutional neural networks model built in advance
Not, the fuzziness of the picture sample is obtained；
Error module, for calculating the mistake between the fuzziness of the picture sample and the realistic blur degree of the picture sample
Difference；
Judgment module establishes module if so, executing, if it is not, then executing for judging whether the error is less than error threshold
Loop module；
Module is established, for calculating weight used in the fuzziness of the picture sample based on the convolutional neural networks, is built
Vertical described image handles model；
Loop module, for the weight based on convolutional neural networks described in the error transfer factor, based on the convolution after adjustment weight
Neural network identifies the picture sample to obtain the fuzziness of picture sample again, what calculating was identified again
Error between the fuzziness of picture sample and the realistic blur degree of the picture sample, until the error is less than the error
Threshold value calculates weight used in the fuzziness of picture sample based on current convolutional neural networks, establishes described image processing mould
Type.
8. system according to claim 6, which is characterized in that the pumping frame unit is specifically used for being based on the view to be processed
The length of frequency takes out frame mode according to interval and obtains N frame picture, whereinL is the view to be treated
The length of frequency,To be rounded formula downwards.
9. system according to claim 6, which is characterized in that the processing unit includes:
Gaussian smoothing module, for carrying out Gaussian smoothing to the fuzzy degree series；
Module is obtained, the M frame that the fuzziness threshold value is greater than for obtaining fuzziness in the fuzzy sequence obscures picture；
First composite module forms 0 or more the fuzzy segment comprising a fuzzy picture for the M frame to be obscured picture,
Or the fuzzy segment comprising the adjacent fuzzy picture of m serial number, wherein M is the positive integer for being less than or equal to N more than or equal to 0, and m is
Integer greater than 1；
Second composite module, for the fuzzy segment to be formed fuzzy fragment list；
Processing module obtains described final for the fuzzy segment in the fuzzy fragment list to be merged and deleted
Fuzzy fragment list.
10. system according to claim 9, which is characterized in that the processing module includes:
Submodule is deleted, if for the fuzzy segment in the fuzzy fragment list there are clip durations less than preset duration,
Delete the fuzzy segment；
Merge submodule, if in the fuzzy fragment list, there are two adjacent fuzzy segments, and described two phases
It is separated by the picture of default frame number between adjacent fuzzy segment, then merging described two adjacent fuzzy segments, which becomes one, obscures
Segment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910243380.4A CN109886963B (en) | 2019-03-28 | 2019-03-28 | Image processing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910243380.4A CN109886963B (en) | 2019-03-28 | 2019-03-28 | Image processing method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109886963A true CN109886963A (en) | 2019-06-14 |
| CN109886963B CN109886963B (en) | 2021-08-24 |
Family
ID=66934924
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910243380.4A Active CN109886963B (en) | 2019-03-28 | 2019-03-28 | Image processing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109886963B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021189827A1 (en) * | 2020-08-06 | 2021-09-30 | 平安科技（深圳）有限公司 | Method and apparatus for recognizing blurred image, and device and computer-readable storage medium |
| CN113792600A (en) * | 2021-08-10 | 2021-12-14 | 武汉光庭信息技术股份有限公司 | Video frame extraction method and system based on deep learning |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105635791A (en) * | 2015-12-30 | 2016-06-01 | 北京奇艺世纪科技有限公司 | Detection method and apparatus for blurred video image |
| CN106599783A (en) * | 2016-11-09 | 2017-04-26 | 浙江宇视科技有限公司 | Video occlusion detection method and device |
| CN107027071A (en) * | 2017-03-14 | 2017-08-08 | 深圳市创达天盛智能科技有限公司 | A kind of method and apparatus of video playback |
| CN107133948A (en) * | 2017-05-09 | 2017-09-05 | 电子科技大学 | Image blurring and noise evaluating method based on multitask convolutional neural networks |
| CN107197233A (en) * | 2017-06-23 | 2017-09-22 | 安徽大学 | Monitor video quality of data evaluating method and device based on edge calculations model |
| CN107240092A (en) * | 2017-05-05 | 2017-10-10 | 浙江大华技术股份有限公司 | A kind of image blur detection method and device |
| CN108322720A (en) * | 2018-03-01 | 2018-07-24 | 高新华 | Image shows content clarity lifting system |
| CN109299687A (en) * | 2018-09-18 | 2019-02-01 | 成都网阔信息技术股份有限公司 | A kind of fuzzy anomalous video recognition methods based on CNN |
-
2019
- 2019-03-28 CN CN201910243380.4A patent/CN109886963B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105635791A (en) * | 2015-12-30 | 2016-06-01 | 北京奇艺世纪科技有限公司 | Detection method and apparatus for blurred video image |
| CN106599783A (en) * | 2016-11-09 | 2017-04-26 | 浙江宇视科技有限公司 | Video occlusion detection method and device |
| CN107027071A (en) * | 2017-03-14 | 2017-08-08 | 深圳市创达天盛智能科技有限公司 | A kind of method and apparatus of video playback |
| CN107240092A (en) * | 2017-05-05 | 2017-10-10 | 浙江大华技术股份有限公司 | A kind of image blur detection method and device |
| CN107133948A (en) * | 2017-05-09 | 2017-09-05 | 电子科技大学 | Image blurring and noise evaluating method based on multitask convolutional neural networks |
| CN107197233A (en) * | 2017-06-23 | 2017-09-22 | 安徽大学 | Monitor video quality of data evaluating method and device based on edge calculations model |
| CN108322720A (en) * | 2018-03-01 | 2018-07-24 | 高新华 | Image shows content clarity lifting system |
| CN109299687A (en) * | 2018-09-18 | 2019-02-01 | 成都网阔信息技术股份有限公司 | A kind of fuzzy anomalous video recognition methods based on CNN |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021189827A1 (en) * | 2020-08-06 | 2021-09-30 | 平安科技（深圳）有限公司 | Method and apparatus for recognizing blurred image, and device and computer-readable storage medium |
| CN113792600A (en) * | 2021-08-10 | 2021-12-14 | 武汉光庭信息技术股份有限公司 | Video frame extraction method and system based on deep learning |
| CN113792600B (en) * | 2021-08-10 | 2023-07-18 | 武汉光庭信息技术股份有限公司 | Video frame extraction method and system based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109886963B (en) | 2021-08-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106612457B (en) | video sequence alignment method and system | |
| CN105225212B (en) | A kind of image processing method and device | |
| CN109886963A (en) | A kind of image processing method and system | |
| CN110060215A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN106845470A (en) | Map data collecting method and apparatus | |
| CN110765134A (en) | File establishing method, equipment and storage medium | |
| WO2019205971A1 (en) | Image processing method, apparatus and device, and image display method | |
| CN104700442A (en) | Image processing method and system for automatic filter and character adding | |
| CN112449117B (en) | Focusing step length determining method and device, storage medium and electronic device | |
| CN106650795A (en) | Sorting method of hotel room type images | |
| CN110222582A (en) | A kind of image processing method and camera | |
| JP3914747B2 (en) | Image search device | |
| CN109559365A (en) | File scanning method, device, mobile terminal and storage medium | |
| CN115996322B (en) | Image data management method for digital video shooting | |
| CN100593948C (en) | Method and device for jointing video | |
| CN113709562A (en) | Automatic clipping method, device and equipment based on baby motion video and storage medium | |
| CN110087097A (en) | It is a kind of that invalid video clipping method is automatically removed based on fujinon electronic video endoscope | |
| CN104361572B (en) | Image processing method and device | |
| CN108540817A (en) | Video data handling procedure, device, server and computer readable storage medium | |
| CN113676688B (en) | Periodic storage method, device, equipment and medium | |
| CN103685965B (en) | The processing method and equipment of a kind of image information | |
| CN109977745A (en) | Face image processing process and relevant apparatus | |
| CN106162158A (en) | A kind of method and device identifying lens shooting mode | |
| CN1322997A (en) | Name card exchanging apparatus, name card exchanging method and recording medium | |
| CN109993079A (en) | A kind of indoor design picture recognition and screening technique and its device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |