CN106530227A - Image restoration method and device - Google Patents
Image restoration method and device Download PDFInfo
- Publication number
- CN106530227A CN106530227A CN201610958087.2A CN201610958087A CN106530227A CN 106530227 A CN106530227 A CN 106530227A CN 201610958087 A CN201610958087 A CN 201610958087A CN 106530227 A CN106530227 A CN 106530227A
- Authority
- CN
- China
- Prior art keywords
- module
- mosaic
- image
- picture material
- resolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 96
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 57
- 238000012545 processing Methods 0.000 claims abstract description 30
- 239000000463 material Substances 0.000 claims description 97
- 230000008569 process Effects 0.000 claims description 58
- 238000012549 training Methods 0.000 claims description 33
- 238000011084 recovery Methods 0.000 claims description 13
- 238000001514 detection method Methods 0.000 claims description 9
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 15
- 238000005070 sampling Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 7
- 230000003287 optical effect Effects 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000000712 assembly Effects 0.000 description 3
- 238000000429 assembly Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000003475 lamination Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000007935 neutral effect Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- KLDZYURQCUYZBL-UHFFFAOYSA-N 2-[3-[(2-hydroxyphenyl)methylideneamino]propyliminomethyl]phenol Chemical compound OC1=CC=CC=C1C=NCCCN=CC1=CC=CC=C1O KLDZYURQCUYZBL-UHFFFAOYSA-N 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 201000001098 delayed sleep phase syndrome Diseases 0.000 description 1
- 208000033921 delayed sleep phase type circadian rhythm sleep disease Diseases 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000012092 media component Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4015—Demosaicing, e.g. colour filter array [CFA], Bayer pattern
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Abstract
The disclosure relates to an image restoration method and device. The method includes determining a mosaic region on a video image; inputting first image content of the mosaic region to a trained convolutional neural network, performing convolution processing on the first image content through a forward convolution module of the convolutional neural network, and obtaining a first feature representation of the first image content; performing deconvolution processing on the first feature representation through a deconvolution module in the convolutional neural network, and obtaining second image content before mosaic processing of the mosaic region, the resolution of the second image content being the same as that of the first image content; and displaying the second image content in the mosaic region. The technical scheme of the disclosure can enable original content on the video image to be displayed, ensures that visual experience of a user is not influenced by existence of the mosaic region on the video image, and improves video quality.
Description
Technical field
It relates to technical field of image processing, more particularly to a kind of image recovery method and device.
Background technology
When user watches the electricity that some television stations or video platform are provided by video website or video application
When shadow or TV play, it will usually which the station symbol in the upper left corner or the upper right corner of whole video pictures stamps mosaic, thus leads
Image information in cause video pictures is blurred, and reduces video quality.
The content of the invention
To overcome problem present in correlation technique, the embodiment of the present disclosure to provide a kind of image recovery method and device, use
To guarantee that the presence on video image no longer due to mosaic area affects the visual experience of user, video quality is lifted.
According to the first aspect of the embodiment of the present disclosure, there is provided a kind of image recovery method, including：
Determine the mosaic area on video image；
First picture material of the mosaic area is input into into the convolutional neural networks trained, by the volume
The positive convolution module of product neutral net carries out process of convolution to described first image content, obtains the of described first image content
One character representation；
The fisrt feature is represented by the warp volume module in the convolutional neural networks carries out deconvolution process, obtains
To the mosaic area by mosaic before the second picture material, the resolution of second picture material and first figure
The resolution of picture content is identical；
Second picture material is included in the mosaic area.
In one embodiment, methods described also includes：
By setting the mosaic sample of quantity to the respective corresponding spy of the positive convolution module and the warp volume module
Levy parameter to be trained, obtain the second of the corresponding fisrt feature parameter sets of the positive convolution module and the warp volume module
Set of characteristic parameters；
Based on the fisrt feature parameter sets that training is obtained, by the positive convolution module to the setting quantity
Mosaic sample carries out process of convolution, obtains second feature and represents；
Based on the second feature parameter sets that training is obtained, by the warp volume module to the second feature table
Showing carries out deconvolution process, obtains the reconstructed image of the mosaic sample of the setting quantity；
Determine the reconstructed image of the setting quantity and each self-corresponding original graph of mosaic sample of the setting quantity
Reset error as between；
When the reset error reaches the condition of convergence, control the convolutional neural networks deconditioning, obtain it is described
The convolutional neural networks of training.
In one embodiment, the output dimension of the warp volume module is identical with the input dimension of the positive convolution module.
In one embodiment, the mosaic area determined on video image, including：
Video image is input into the disaggregated model trained, the disaggregated model is led to by the mosaic sample for setting quantity
Cross to be trained default sorting algorithm and obtain；
The mosaic area on the video image is detected based on the disaggregated model trained.
In one embodiment, methods described also includes：
Determine the input dimension whether phase of resolution and the positive convolution module of first picture material of mosaic area
Together；
If the resolution of described first image content is differed with the input dimension, according to described first image content
Resolution and the input dimension, zoom in and out process to described first image content；
It is described that second picture material is included into that methods described also includes before the mosaic area：
According to the resolution and the input dimension of described first image content, by second picture material zoom to
Resolution before the not scaled process of described first image content.
According to the second aspect of the embodiment of the present disclosure, there is provided a kind of image restoration device, including：
First determining module, the mosaic area being configured to determine that on video image；
First processing module, is configured to first image of the mosaic area of first determining module determination
Content is input into into the convolutional neural networks trained, by the positive convolution module of the convolutional neural networks to first figure
As content carries out process of convolution, the fisrt feature for obtaining described first image content is represented；
Second processing module, the warp volume module being configured in the convolutional neural networks are processed to described first
The fisrt feature that module is obtained represents and carries out deconvolution process, the second figure before obtaining the mosaic area by mosaic
As content, the resolution of second picture material is identical with the resolution of described first image content；
Display module, is configured to include second picture material that the Second processing module is obtained in the horse
Match gram region.
In one embodiment, described device may also include：
Training module, is configured to set the mosaic sample of quantity to the positive convolution module and the deconvolution
The each self-corresponding characteristic parameter of module is trained, and obtains corresponding fisrt feature parameter sets of the positive convolution module and described
The second feature parameter sets of warp volume module；
3rd processing module, is configured to the fisrt feature parameter sets obtained based on training module training,
Process of convolution is carried out to the mosaic sample of the setting quantity by the positive convolution module, second feature is obtained and is represented；
Fourth processing module, is configured to the second feature parameter sets obtained based on training, by the warp
Volume module is represented to the second feature that the 3rd processing module is obtained carries out deconvolution process, obtains the setting quantity
Mosaic sample reconstructed image；
Second determining module, is configured to determine that the reconstructed image of the setting quantity that the fourth processing module is obtained
And the reset error between each self-corresponding original image of mosaic sample of the setting quantity；
Control module, when being configured as the reset error that second determining module determines and reaching the condition of convergence,
The convolutional neural networks deconditioning is controlled, the convolutional neural networks trained are obtained.
In one embodiment, the output dimension of the warp volume module is identical with the input dimension of the positive convolution module.
In one embodiment, first determining module may include：
Input submodule, is configured to video image is input into the disaggregated model trained, and the disaggregated model is by setting
The mosaic sample of fixed number amount is obtained by being trained to default sorting algorithm；
Detection sub-module, is configured to the institute of the input submodule input is detected based on the disaggregated model trained
State the mosaic area on video image.
In one embodiment, described device may also include：
3rd determining module, is configured to determine that the resolution and the positive convolution of first picture material of mosaic area
Whether the input dimension of module is identical；
Zoom module, if be configured to the 3rd determining module determine the resolution of described first image content with it is described
Input dimension is differed, according to resolution and the input dimension of described first image content, to described first image content
Zoom in and out process；
The Zoom module is additionally configured to resolution and the input dimension according to described first image content, by institute
State the second picture material and zoom to the resolution before the not scaled process with described first image content.
According to the third aspect of the embodiment of the present disclosure, there is provided a kind of image restoration device, including：
Processor；
For storing the memorizer of processor executable；
Wherein, the processor is configured to：
Determine the mosaic area on video image；
First picture material of the mosaic area is input into into the convolutional neural networks trained, by the volume
The positive convolution module of product neutral net carries out process of convolution to described first image content, obtains the of described first image content
One character representation；
The fisrt feature is represented by the warp volume module in the convolutional neural networks carries out deconvolution process, obtains
To the mosaic area by mosaic before the second picture material, the resolution of second picture material and first figure
The resolution of picture content is identical；
Second picture material is included in the mosaic area.
The technical scheme that embodiment of the disclosure is provided can include following beneficial effect：
Process of convolution is carried out to first picture material of mosaic area by the positive convolution module of CNN, by the anti-of CNN
Convolution module to convolution after the character representation of the first picture material carry out deconvolution process, obtain mosaic area by mosaic
The second front picture material, the second picture material is included in mosaic area, such that it is able to make to show on video image
The original contents gone out on video image, it is ensured that the presence on video image no longer due to mosaic area affects the vision body of user
Test, improve video quality.
It should be appreciated that the general description of the above and detailed description hereinafter are only exemplary and explanatory, not
The disclosure can be limited.
Description of the drawings
During accompanying drawing herein is merged in description and the part of this specification is constituted, show the enforcement for meeting the present invention
Example, and be used for explaining the principle of the present invention together with description.
Figure 1A is the flow chart of the image recovery method according to an exemplary embodiment.
Figure 1B is the frame diagram of the convolutional neural networks shown in Figure 1A illustrated embodiments.
Fig. 1 C are the schematic diagrams for not breaking mosaic on the video image shown in Figure 1A illustrated embodiments.
Fig. 1 D are the schematic diagrams of the mosaic on the video image shown in Figure 1A illustrated embodiments.
Fig. 1 E are the schematic diagrams after restoring to video image shown in Figure 1A illustrated embodiments.
Fig. 2A is the flow chart of the training convolutional neural networks according to an exemplary embodiment.
The frame diagram adopted when being the training convolutional neural networks according to an exemplary embodiment by Fig. 2 B.
Fig. 3 is the flow chart of the image recovery method for implementing to exemplify according to another exemplary.
Fig. 4 is a kind of block diagram of the image restoration device according to an exemplary embodiment.
Fig. 5 is the block diagram of another kind of image restoration device according to an exemplary embodiment.
Fig. 6 is a kind of block diagram suitable for image restoration device according to an exemplary embodiment.
Specific embodiment
Here in detail exemplary embodiment will be illustrated, its example is illustrated in the accompanying drawings.Explained below is related to
During accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous key element.Following exemplary embodiment
Described in embodiment do not represent and the consistent all embodiments of the present invention.Conversely, they be only with as appended by
The example of consistent apparatus and method in terms of some described in detail in claims, the present invention.
Figure 1A is the flow chart of the image recovery method according to an exemplary embodiment, and Figure 1B is implemented shown in Figure 1A
The Organization Chart of the convolutional neural networks for exemplifying, Fig. 1 C are not break mosaic on the video image shown in Figure 1A illustrated embodiments
Schematic diagram, Fig. 1 D are the schematic diagrams of the mosaic on the video image shown in Figure 1A illustrated embodiments, and Fig. 1 E are shown in Figure 1A
Enforcement exemplify video image is restored after schematic diagram；The image recovery method can be applied in electronic equipment
(for example：Smart mobile phone, panel computer) on, can be realized by way of Video Applications are installed on an electronic device, such as Figure 1A
Shown, the image recovery method comprises the following steps S101-S104：
In step S101, the mosaic area on video image is determined.
In one embodiment, can be by carrying out rough detection to the region that the station symbol on video image is located, positioning is gone into action
The approximate region that Sai Ke is located, for example, mosaic would generally be being detected substantially in the upper left side of screen image or upper right side
Behind region, the feature such as contrast, definition further according to mosaic itself identifies mosaic area.In another embodiment,
The method that pattern recognition can be adopted, is identified to mosaic area by the disaggregated model trained.
In step s 102, first picture material of mosaic area is input into into the convolutional neural networks trained,
Process of convolution is carried out to the first picture material by the positive convolution module of convolutional neural networks, the first of the first picture material is obtained
Character representation.
In one embodiment, as shown in Figure 1B, positive convolution module 11 of the first picture material output in CNN10, passes through
The process of convolution of positive convolution module 11, obtains the character representation of the first picture material.It will be appreciated by persons skilled in the art that
Positive convolution module 11 can include multiple convolutional layers and multiple sub-sampling layers, each convolutional layer and each sub-sampling in CNN
The parameter of layer is what CNN10 was trained to, and the fisrt feature obtained through 11 process of convolution of positive convolution module represent can
Accurately represent the first picture material, with regard to CNN in multiple convolutional layers and the concrete structure of multiple sub-sampling layers can be found in phase
The description of pass technology, disclosure here are not detailed.
In step s 103, fisrt feature is represented by the warp volume module in convolutional neural networks is carried out at deconvolution
Reason, the second picture material before obtaining mosaic area by mosaic, the resolution of the second picture material and the first picture material
Resolution it is identical.
In one embodiment, as shown in Figure 1B, the character representation of the first picture material can be input to warp volume module
12, processed by the deconvolution of warp volume module 12, the second picture material before restoring mosaic area by mosaic.This area
Technical staff is it is understood that warp volume module 12 can include multiple warp laminations and multiple up-sampling layers, up-sampling layer
Character representation for processing to deconvolution carries out up-sampling reconstruction, to guarantee the second picture material of the output of warp volume module 12
Resolution it is identical with the resolution of the first picture material.
In step S104, the second picture material is included in mosaic area.
In an exemplary scenario, the video image shown in Fig. 1 C has platform station symbol " 56 I find pleasure in ", positioned at whole video
The upper right side of image.In video display process, the station symbol in the video image of each frame is labeled with mosaic, if mosaic area
Domain has important picture material, influences whether that user watches the experience of video image, by the embodiment of the present disclosure, can be by horse
Match gram region restores not comprising the picture material for having station symbol, such that it is able to be able to ensure that the integrity of video image.As showing
Example, the resolution of the video image in Fig. 1 C is 1400*1000, by the mosaic area in Fig. 1 C that step S101 is identified
Size be 200*200, now first picture material of the 200*200 of mosaic area can be input into having trained
CNN10, by CNN10 in 11 pairs of the first picture materials of positive convolution module carry out process of convolution, obtain the first picture material
Fisrt feature represents, fisrt feature is represented and carry out deconvolution by warp volume module 12, export the second picture material, the second figure
As the resolution of content is 200*200, by the second picture material is included in mosaic area, the horse to video image is realized
The image restoration in match gram region, such that it is able to allow users to watch the normal picture for not having mosaic, lifts screen quality.
In the present embodiment, process of convolution is carried out to first picture material of mosaic area by the positive convolution module of CNN,
The fisrt feature of the first picture material after the warp volume module of CNN is to convolution represents and carries out deconvolution process, obtains horse
The second picture material before gram region is matched by mosaic, the second picture material is included in mosaic area, such that it is able to make to regard
The original contents on video image are can show that on frequency image, it is ensured that no longer due to the presence of mosaic area on video image
The visual experience of user is affected, video quality is improved.
In one embodiment, method also includes：
Align convolution module and warp volume module each corresponding characteristic parameter enters by setting the mosaic sample of quantity
Row training, obtains the second feature parameter sets of the corresponding fisrt feature parameter sets of positive convolution module and warp volume module；
Based on the fisrt feature parameter sets that training is obtained, the mosaic sample for setting quantity is entered by positive convolution module
Row process of convolution, obtains second feature and represents；
Based on the second feature parameter sets that training is obtained, second feature is represented by warp volume module carries out deconvolution
Process, obtain the reconstructed image of the mosaic sample of setting quantity；
It is determined that between each self-corresponding original image of mosaic sample of the reconstructed image of setting quantity and setting quantity
Reset error；
When reset error reaches the condition of convergence, convolutional neural networks deconditioning is controlled, the convolution god for having been trained
Jing networks.
In one embodiment, the output dimension of warp volume module and the input dimension of positive convolution module are identical.
In one embodiment, determine the mosaic area on video image, including：
Video image is input into the disaggregated model trained, disaggregated model passes through right by the mosaic sample for setting quantity
Default sorting algorithm is trained and obtains；
Based on the mosaic area on the disaggregated model detection video image trained.
In one embodiment, method also includes：
Determine whether the resolution of first picture material of mosaic area is identical with the input dimension of positive convolution module；
If the resolution of the first picture material is differed with input dimension, according to resolution and the input of the first picture material
Dimension, zooms in and out process to the first picture material.
So far, the said method that the embodiment of the present disclosure is provided, can make to can show that on video image on video image
Original contents, it is ensured that on video image no longer due to mosaic area presence affect user visual experience, lifted video
Quality.
Below with specific embodiment come illustrate the embodiment of the present disclosure provide technical scheme.
Fig. 2A is the flow chart of the training convolutional neural networks according to an exemplary embodiment, and Fig. 2 B are shown according to one
Example property implements the frame diagram adopted during the training convolutional neural networks for exemplifying；The present embodiment is provided using the embodiment of the present disclosure
Said method, it is illustrative as a example by training convolutional neural networks by how, as shown in Figure 2 A, comprise the steps：
In step s 201, by setting, the mosaic sample of quantity aligns convolution module and warp volume module is each corresponded to
Characteristic parameter be trained, obtain the second feature of the corresponding fisrt feature parameter sets of positive convolution module and warp volume module
Parameter sets.
In one embodiment, the mosaic sample is the figure that obtained after mosaic of station symbol of television station or the network platform
Piece sample.In one embodiment, batch training can be carried out to CNN, participates in the setting quantity of the mosaic sample of training every time
Concrete number be not limited, for example, set quantity as 128,64, etc..In one embodiment, in fisrt feature parameter sets
Characteristic parameter be weight parameter during each convolutional layer and sub-sampling layer in positive convolution module 11 is, second feature parameter set
Characteristic parameter in conjunction is the weight parameter in each warp lamination in warp volume module 12 and up-sampling layer, fisrt feature ginseng
The quantity of the parameter included in manifold conjunction and second feature parameter sets can be according to each convolutional layer and sub-sampling layer, each warp
Determining, the disclosure is not limited the quantity of the neuron in lamination and up-sampling layer to the particular number of parameter.
In step S202, based on the fisrt feature parameter sets that training is obtained, by positive convolution module to setting quantity
Mosaic sample carry out process of convolution, obtain second feature and represent.
In step S203, based on the second feature parameter sets that training is obtained, by warp volume module to second feature
Expression carries out deconvolution process, obtains the reconstructed image of the mosaic sample of setting quantity.
Positive convolution module carries out process of convolution and warp volume module to second feature to the mosaic sample for setting quantity
Expression carries out the description that the description of deconvolution process may refer to above-mentioned Figure 1A illustrated embodiments, will not be described in detail herein.
In step S204, it is determined that the reconstructed image of setting quantity and each self-corresponding original of mosaic sample for setting quantity
Reset error between beginning image, when reset error reaches the condition of convergence, execution step S205, when reset error is not up to received
When holding back condition, execution step S201 continues to train to CNN.
In one embodiment, the recovery between the pre- original image of reconstructed image can be calculated by the loss function of CNN10
Error, the concrete formula of the loss function of CNN10 may refer to description of related art, will not be described in detail herein.
In step S205, when reset error reaches the condition of convergence, convolutional neural networks deconditioning is controlled, is obtained
The convolutional neural networks of training.
It is in an exemplary scenario, illustrative as a example by 128 to set quantity, 128 mosaic samples pair
The original images of 128 station symbols comprising television station or the network platform are answered, when 128 mosaic samples are input to
After CNN10,128 mosaic samples can align convolution module 11 and each self-corresponding characteristic parameter of warp volume module 12 enters
Row training, obtains the second feature parameter set of 11 corresponding fisrt feature parameter sets of positive convolution module and warp volume module 12
Close.Afterwards, by the second feature parameter sets pair of the fisrt feature parameter sets and warp volume module 12 of positive convolution module 11
128 mosaic samples carry out convolution and deconvolution, obtain each self-corresponding restored image of 128 mosaic samples, lead to
Cross reset error computing module 13 and calculate each self-corresponding restored image of 128 mosaic samples and corresponding original image
Reset error, when reset error reaches the condition of convergence, controls CNN10 deconditionings, the convolutional neural networks trained.
In the present embodiment, by being trained to CNN, CNN is made to be provided with the function of positive convolution and reverse convolution, when
When reset error reaches the condition of convergence, the CNN accurate recoveries after training can be enable to go out the original image of mosaic area, really
Protect the integrity of video image.
Fig. 3 is the flow chart of the image recovery method for implementing to exemplify according to another exemplary；The present embodiment utilizes this public affairs
The said method of embodiment offer is opened, how to determine the mosaic area on video image the first figure to mosaic area
As content carry out it is illustrative as a example by pretreatment, as shown in figure 3, comprising the steps：
In step S301, video image is input into the disaggregated model trained, disaggregated model is by the horse for setting quantity
Match gram sample is obtained by being trained to default sorting algorithm.
In one embodiment, default sorting algorithm can be adaboost, support vector machine (Support Vector
Machine, referred to as SVM), CNN etc., default sorting algorithm is instructed by mosaic sample and non-mosaic sample
Practice, obtain for detecting the disaggregated model of mosaic area.
In step s 302, based on the mosaic area on the disaggregated model detection video image trained.
In one embodiment, mosaic area can include the center of mosaic area and the of mosaic area
The resolution of one picture material.
In step S303, the input of the resolution and convolutional neural networks of first picture material of mosaic area is determined
Whether dimension is identical.
In step s 304, if the resolution of the first picture material is differed with input dimension, according to the first picture material
Resolution with input dimension, process is zoomed in and out to the first picture material.
In step S305, after the second picture material is reset into, according to resolution and the input dimension of the first picture material
Second picture material is zoomed to the resolution before the not scaled process with the first picture material by degree.
For example, determine that the resolution of the first picture material is 200*200 by above-mentioned steps S302, the input of CNN10 is tieed up
Spend for 100*100, then the diminution that the first picture material carry out 1/4 can be processed, it is real shown in above-mentioned Figure 1A such that it is able to pass through
The description for applying example is restored to the first picture material, is restoring the second picture material by above-mentioned Figure 1A illustrated embodiments
Afterwards, the resolution of the second picture material is 100*100, it is therefore desirable to the second picture material is carried out 2*2 times and is amplified, so as to incite somebody to action
Second picture material of 100*100 sizes is amplified to the resolution before the not scaled process with the first picture material, i.e.,
200*200。
In the present embodiment, the mosaic area in video image is identified by the disaggregated model trained, can be with
Improve the efficiency of mosaic area's detection；By process is zoomed in and out to mosaic area, it can be ensured that the Marseille of arbitrary size
Gram region can carry out image restoration by CNN.
Fig. 4 is a kind of block diagram of the image restoration device according to an exemplary embodiment, as shown in figure 4, image is multiple
Original device includes：
First determining module 41, the mosaic area being configured to determine that on video image；
First processing module 42, is configured to first picture material of mosaic area for determining the first determining module 41
It is input into into the convolutional neural networks trained, the first picture material is rolled up by the positive convolution module of convolutional neural networks
Product is processed, and the fisrt feature for obtaining the first picture material is represented；
Second processing module 43, is configured to the warp volume module 42 in convolutional neural networks to first processing module
The fisrt feature for obtaining is represented carries out deconvolution process, the second picture material before obtaining mosaic area by mosaic, and second
The resolution of picture material is identical with the resolution of the first picture material；
Display module 44, is configured to include the second picture material that Second processing module 43 is obtained in mosaic area
Domain.
Fig. 5 is the block diagram of another kind of image restoration device according to an exemplary embodiment, as shown in figure 5, upper
On the basis of stating embodiment illustrated in fig. 4, in one embodiment, device may also include：
Training module 45, the mosaic sample for being configured to set quantity align convolution module and warp volume module is each
Self-corresponding characteristic parameter is trained, and obtains the of the corresponding fisrt feature parameter sets of positive convolution module and warp volume module
Two set of characteristic parameters；
3rd processing module 46, is configured to the fisrt feature parameter sets obtained based on the training of training module 45, is passed through
Positive convolution module carries out process of convolution to the mosaic sample for setting quantity, obtains second feature and represents；
Fourth processing module 47, is configured to the second feature parameter sets obtained based on training, by warp volume module
The second feature that 3rd processing module 46 is obtained is represented carries out deconvolution process, obtains the weight of the mosaic sample of setting quantity
Composition picture；
Second determining module 48, is configured to determine that the reconstructed image of the setting quantity that fourth processing module 47 is obtained and sets
Reset error between each self-corresponding original image of mosaic sample of fixed number amount；
Control module 49, is configured as the reset error of the second determining module determination when reaching the condition of convergence, control volume
Product neutral net deconditioning, the convolutional neural networks trained.
In one embodiment, the output dimension of warp volume module and the input dimension of positive convolution module are identical.
In one embodiment, the first determining module 41 may include：
Input submodule 411, is configured to video image is input into the disaggregated model trained, and disaggregated model is by setting
The mosaic sample of quantity is obtained by being trained to default sorting algorithm；
Detection sub-module 412, is configured to the video based on the disaggregated model detection input of input submodule 411 trained
Mosaic area on image.
In one embodiment, device may also include：
3rd determining module 50, is configured to determine that in first image of the mosaic area that the first determining module 41 determines
Whether the resolution of appearance is identical with the input dimension of positive convolution module；
Zoom module 51, if being configured to resolution and input dimension that the 3rd determining module 50 determines the first picture material
Differ, according to resolution and the input dimension of the first picture material, process is zoomed in and out to the first picture material.
Zoom module 51 is additionally configured to include the second picture material before mosaic area in display module 44, root
According to the resolution and input dimension of the first picture material, the second picture material is zoomed to not scaled with the first picture material
Resolution before process.
With regard to the device in above-described embodiment, wherein modules perform the concrete mode of operation in relevant the method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
Fig. 6 is a kind of block diagram suitable for image restoration device according to an exemplary embodiment.For example, device
600 can be mobile phone, and computer, digital broadcast terminal, messaging devices, game console, tablet device, medical treatment set
It is standby, body-building equipment, personal digital assistant etc..
With reference to Fig. 6, device 600 can include following one or more assemblies：Process assembly 602, memorizer 604, power supply
Component 606, multimedia groupware 608, audio-frequency assembly 610, the interface 612 of input/output (I/O), sensor cluster 614, and
Communication component 616.
The integrated operation of 602 usual control device 600 of process assembly, such as with display, call, data communication, phase
Machine operates and records the associated operation of operation.Treatment element 602 can refer to perform including one or more processors 620
Order, to complete all or part of step of above-mentioned method.Additionally, process assembly 602 can include one or more modules, just
Interaction between process assembly 602 and other assemblies.For example, processing component 602 can include multi-media module, many to facilitate
Interaction between media component 608 and process assembly 602.
Memorizer 604 is configured to store various types of data to support the operation in equipment 600.These data are shown
Example includes the instruction of any application program or method for operating on device 600, and contact data, telephone book data disappear
Breath, picture, video etc..Memorizer 604 can be by any kind of volatibility or non-volatile memory device or their group
Close and realize, such as static RAM (SRAM), Electrically Erasable Read Only Memory (EEPROM) is erasable to compile
Journey read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash
Device, disk or CD.
Electric power assembly 606 provides electric power for the various assemblies of device 600.Electric power assembly 606 can include power management system
System, one or more power supplys, and other generate, manage and distribute the component that electric power is associated with for device 600.
Multimedia groupware 608 is included in the screen of one output interface of offer between described device 600 and user.One
In a little embodiments, screen can include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen
Curtain may be implemented as touch screen, to receive the input signal from user.Touch panel includes one or more touch sensings
Device is with the gesture on sensing touch, slip and touch panel.The touch sensor can not only sensing touch or sliding action
Border, but also detect and the touch or slide related persistent period and pressure.In certain embodiments, many matchmakers
Body component 608 includes a front-facing camera and/or post-positioned pick-up head.When equipment 600 be in operator scheme, such as screening-mode or
During video mode, front-facing camera and/or post-positioned pick-up head can receive outside multi-medium data.Each front-facing camera and
Post-positioned pick-up head can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio-frequency assembly 610 is configured to output and/or input audio signal.For example, audio-frequency assembly 610 includes a Mike
Wind (MIC), when device 600 is in operator scheme, such as call model, logging mode and speech recognition mode, mike is matched somebody with somebody
It is set to reception external audio signal.The audio signal for being received can be further stored in memorizer 604 or via communication set
Part 616 sends.In certain embodiments, audio-frequency assembly 610 also includes a speaker, for exports audio signal.
, for interface is provided between process assembly 602 and peripheral interface module, above-mentioned peripheral interface module can for I/O interfaces 612
To be keyboard, click wheel, button etc..These buttons may include but be not limited to：Home button, volume button, start button and lock
Determine button.
Sensor cluster 614 includes one or more sensors, and the state for various aspects are provided for device 600 is commented
Estimate.For example, sensor cluster 614 can detect the opening/closed mode of equipment 600, and the relative localization of component is for example described
Display and keypad of the component for device 600, sensor cluster 614 can be with 600 1 components of detection means 600 or device
Position change, user is presence or absence of with what device 600 was contacted, 600 orientation of device or acceleration/deceleration and device 600
Temperature change.Sensor cluster 614 can include proximity transducer, be configured to detect when not having any physical contact
The presence of object nearby.Sensor cluster 614 can also include optical sensor, such as CMOS or ccd image sensor, for into
As used in application.In certain embodiments, the sensor cluster 614 can also include acceleration transducer, gyro sensors
Device, Magnetic Sensor, pressure transducer or temperature sensor.
Communication component 616 is configured to facilitate the communication of wired or wireless way between device 600 and other equipment.Device
600 can access the wireless network based on communication standard, such as WiFi, 2G or 3G, or combinations thereof.In an exemplary enforcement
In example, communication component 616 receives the broadcast singal or broadcast related information from external broadcasting management system via broadcast channel.
In one exemplary embodiment, the communication component 616 also includes near-field communication (NFC) module, to promote junction service.Example
Such as, NFC module can be based on RF identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra broadband (UWB) technology,
Bluetooth (BT) technology and other technologies are realizing.
In the exemplary embodiment, device 600 can be by one or more application specific integrated circuits (ASIC), numeral letter
Number processor (DSP), digital signal processing appts (DSPD), PLD (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components realizations, for performing said method.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
Such as include the memorizer 604 of instruction, above-mentioned instruction can be performed to complete said method by the processor 620 of device 600.For example,
The non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, floppy disk
With optical data storage devices etc..
Those skilled in the art will readily occur to its of the disclosure after considering description and putting into practice disclosure disclosed herein
Its embodiment.The application is intended to any modification, purposes or the adaptations of the disclosure, these modifications, purposes or
Person's adaptations follow the general principle of the disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.Description and embodiments are considered only as exemplary, and the true scope of the disclosure and spirit are by following
Claim is pointed out.
It should be appreciated that the disclosure is not limited to the precision architecture for being described above and being shown in the drawings, and
And various modifications and changes can be being carried out without departing from the scope.The scope of the present disclosure is limited only by appended claim.
Claims (11)
1. a kind of image recovery method, it is characterised in that methods described includes：
Determine the mosaic area on video image；
First picture material of the mosaic area is input into into the convolutional neural networks trained, by convolution god
The positive convolution module of Jing networks carries out process of convolution to described first image content, and obtain described first image content first is special
Levy expression；
The fisrt feature is represented by the warp volume module in the convolutional neural networks carries out deconvolution process, obtains institute
The second picture material before mosaic area is stated by mosaic, in the resolution and described first image of second picture material
The resolution of appearance is identical；
Second picture material is included in the mosaic area.
2. method according to claim 1, it is characterised in that methods described also includes：
By setting the mosaic sample of quantity, to the positive convolution module and the warp volume module, each corresponding feature is joined
Number is trained, and obtains the second feature of the corresponding fisrt feature parameter sets of the positive convolution module and the warp volume module
Parameter sets；
Based on the fisrt feature parameter sets that training is obtained, by Marseille of the positive convolution module to the setting quantity
Gram sample carries out process of convolution, obtains second feature and represents；
Based on the second feature parameter sets that obtain of training, by the warp volume module second feature is represented into
Row deconvolution is processed, and obtains the reconstructed image of the mosaic sample of the setting quantity；
Determine it is described setting quantity reconstructed image and it is described setting quantity each self-corresponding original image of mosaic sample it
Between reset error；
When the reset error reaches the condition of convergence, the convolutional neural networks deconditioning is controlled, obtain described training
Convolutional neural networks.
3. method according to claim 2, it is characterised in that the output dimension of the warp volume module and the positive convolution
The input dimension of module is identical.
4. method according to claim 1, it is characterised in that the mosaic area on the determination video image, including：
Video image is input into the disaggregated model trained, the disaggregated model passes through right by the mosaic sample for setting quantity
Default sorting algorithm is trained and obtains；
The mosaic area on the video image is detected based on the disaggregated model trained.
5. method according to claim 1, it is characterised in that methods described also includes：
Determine whether the resolution of first picture material of mosaic area is identical with the input dimension of the positive convolution module；
If the resolution of described first image content is differed with the input dimension, according to the resolution of described first image content
Rate and the input dimension, zoom in and out process to described first image content；
It is described that second picture material is included into that methods described also includes before the mosaic area：
According to the resolution and the input dimension of described first image content, by second picture material zoom to it is described
Resolution before the not scaled process of the first picture material.
6. a kind of image restoration device, it is characterised in that described device includes：
First determining module, the mosaic area being configured to determine that on video image；
First processing module, is configured to first picture material of the mosaic area of first determining module determination
It is input into into the convolutional neural networks trained, by the positive convolution module of the convolutional neural networks in described first image
Appearance carries out process of convolution, and the fisrt feature for obtaining described first image content is represented；
Second processing module, is configured to the warp volume module in the convolutional neural networks to the first processing module
The fisrt feature for obtaining is represented and carries out deconvolution process, in the second image before obtaining the mosaic area by mosaic
Hold, the resolution of second picture material is identical with the resolution of described first image content；
Display module, is configured to include second picture material that the Second processing module is obtained in the mosaic
Region.
7. device according to claim 6, it is characterised in that described device also includes：
Training module, is configured to set the mosaic sample of quantity to the positive convolution module and the warp volume module
Each self-corresponding characteristic parameter is trained, and obtains the corresponding fisrt feature parameter sets of the positive convolution module and the warp
The second feature parameter sets of volume module；
3rd processing module, is configured to the fisrt feature parameter sets obtained based on training module training, is passed through
The positive convolution module carries out process of convolution to the mosaic sample of the setting quantity, obtains second feature and represents；
Fourth processing module, is configured to the second feature parameter sets obtained based on training, by the warp product module
Block is represented to the second feature that the 3rd processing module is obtained carries out deconvolution process, obtains the horse of the setting quantity
The reconstructed image of match gram sample；
Second determining module, is configured to determine that reconstructed image and the institute of the setting quantity that the fourth processing module is obtained
State the reset error between each self-corresponding original image of mosaic sample of setting quantity；
Control module, when being configured as the reset error that second determining module determines and reaching the condition of convergence, control
The convolutional neural networks deconditioning, obtains the convolutional neural networks trained.
8. device according to claim 7, it is characterised in that the output dimension of the warp volume module and the positive convolution
The input dimension of module is identical.
9. device according to claim 6, it is characterised in that first determining module includes：
Input submodule, is configured to video image is input into the disaggregated model trained, and the disaggregated model is by setting number
The mosaic sample of amount is obtained by being trained to default sorting algorithm；
Detection sub-module, is configured to detect that the input submodule is regarded described in being input into based on the disaggregated model trained
Mosaic area on frequency image.
10. device according to claim 6, it is characterised in that described device also includes：
3rd determining module, is configured to determine that the resolution and the positive convolution module of first picture material of mosaic area
Input dimension it is whether identical；
Zoom module, if being configured to the 3rd determining module determines the resolution of described first image content and the input
Dimension is differed, and according to resolution and the input dimension of described first image content, described first image content is carried out
Scaling process；
The Zoom module is additionally configured to include second picture material in the mosaic area in the display module
Before domain, according to resolution and the input dimension of described first image content, by second picture material zoom to
Resolution before the not scaled process of described first image content.
11. a kind of image restoration devices, it is characterised in that described device includes：
Processor；
For storing the memorizer of processor executable；
Wherein, the processor is configured to：
Determine the mosaic area on video image；
First picture material of the mosaic area is input into into the convolutional neural networks trained, by convolution god
The positive convolution module of Jing networks carries out process of convolution to described first image content, and obtain described first image content first is special
Levy expression；
The fisrt feature is represented by the warp volume module in the convolutional neural networks carries out deconvolution process, obtains institute
The second picture material before mosaic area is stated by mosaic, in the resolution and described first image of second picture material
The resolution of appearance is identical；
Second picture material is included in the mosaic area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610958087.2A CN106530227B (en) | 2016-10-27 | 2016-10-27 | Image recovery method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610958087.2A CN106530227B (en) | 2016-10-27 | 2016-10-27 | Image recovery method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106530227A true CN106530227A (en) | 2017-03-22 |
| CN106530227B CN106530227B (en) | 2019-08-06 |
Family
ID=58325864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610958087.2A Active CN106530227B (en) | 2016-10-27 | 2016-10-27 | Image recovery method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106530227B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016657A (en) * | 2017-04-07 | 2017-08-04 | 河北工业大学 | The restorative procedure of the face picture covered by reticulate pattern |
| CN107578054A (en) * | 2017-09-27 | 2018-01-12 | 北京小米移动软件有限公司 | Image processing method and device |
| CN107992894A (en) * | 2017-12-12 | 2018-05-04 | 北京小米移动软件有限公司 | Image-recognizing method, device and computer-readable recording medium |
| CN108010538A (en) * | 2017-12-22 | 2018-05-08 | 北京奇虎科技有限公司 | Audio data processing method and device, computing device |
| CN108230253A (en) * | 2017-05-08 | 2018-06-29 | 北京市商汤科技开发有限公司 | Image recovery method, device, electronic equipment and computer storage media |
| CN108734667A (en) * | 2017-04-14 | 2018-11-02 | Tcl集团股份有限公司 | A kind of image processing method and system |
| CN108805884A (en) * | 2018-06-13 | 2018-11-13 | 北京搜狐新媒体信息技术有限公司 | A kind of mosaic area's detection method, device and equipment |
| CN108876726A (en) * | 2017-12-12 | 2018-11-23 | 北京旷视科技有限公司 | Method, apparatus, system and the computer storage medium of image procossing |
| CN109410123A (en) * | 2018-10-15 | 2019-03-01 | 深圳市能信安科技股份有限公司 | The method, apparatus and electronic equipment of removal mosaic based on deep learning |
| WO2019047655A1 (en) * | 2017-09-05 | 2019-03-14 | 百度在线网络技术（北京）有限公司 | Method and apparatus for use in determining driving behavior of driverless vehicle |
| CN109564638A (en) * | 2018-01-15 | 2019-04-02 | 深圳鲲云信息科技有限公司 | Artificial intelligence process device and its applied processing method |
| CN110111282A (en) * | 2019-05-09 | 2019-08-09 | 杭州电子科技大学上虞科学与工程研究院有限公司 | A kind of video deblurring method based on motion vector and CNN |
| CN111351502A (en) * | 2018-12-21 | 2020-06-30 | 赫尔环球有限公司 | Method, apparatus and computer program product for generating an overhead view of an environment from a perspective view |
| CN111401453A (en) * | 2020-03-18 | 2020-07-10 | 西安电子科技大学 | Mosaic image classification and identification method and system |
| CN112184597A (en) * | 2020-11-05 | 2021-01-05 | 温州大学大数据与信息技术研究院 | Image restoration device and method |
| CN112329634A (en) * | 2020-11-05 | 2021-02-05 | 华中师范大学 | Classroom behavior recognition method and device, electronic equipment and storage medium |
| WO2021135614A1 (en) * | 2019-12-31 | 2021-07-08 | 杭州海康威视数字技术股份有限公司 | Video processing method, apparatus, and system |
| US11145029B2 (en) * | 2019-07-25 | 2021-10-12 | Rovi Guides, Inc. | Automated regeneration of low quality content to high quality content |
| CN113905276A (en) * | 2021-09-24 | 2022-01-07 | 威视芯半导体(合肥)有限公司 | Video coding processing method, system, equipment and storage medium |
| US11256863B2 (en) | 2019-07-19 | 2022-02-22 | Rovi Guides, Inc. | Systems and methods for generating content for a screenplay |
| US11336943B2 (en) * | 2020-02-24 | 2022-05-17 | Realtek Semiconductor Corporation | Method and electronic device for detecting resolution |
| US11562016B2 (en) | 2019-06-26 | 2023-01-24 | Rovi Guides, Inc. | Systems and methods for generating supplemental content for media content |
| US11604827B2 (en) | 2020-02-21 | 2023-03-14 | Rovi Guides, Inc. | Systems and methods for generating improved content based on matching mappings |
| US11895376B2 (en) | 2019-03-25 | 2024-02-06 | Rovi Guides, Inc. | Systems and methods for creating customized content |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102131079A (en) * | 2011-04-20 | 2011-07-20 | 杭州华三通信技术有限公司 | Method and device for eliminating motion blur of image |
| CN104200224A (en) * | 2014-08-28 | 2014-12-10 | 西北工业大学 | Valueless image removing method based on deep convolutional neural networks |
| US20150178591A1 (en) * | 2013-12-18 | 2015-06-25 | New York University | System, method and computer-accessible medium for restoring an image taken through a window |
| CN105096279A (en) * | 2015-09-23 | 2015-11-25 | 成都融创智谷科技有限公司 | Digital image processing method based on convolutional neural network |
-
2016
- 2016-10-27 CN CN201610958087.2A patent/CN106530227B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102131079A (en) * | 2011-04-20 | 2011-07-20 | 杭州华三通信技术有限公司 | Method and device for eliminating motion blur of image |
| US20150178591A1 (en) * | 2013-12-18 | 2015-06-25 | New York University | System, method and computer-accessible medium for restoring an image taken through a window |
| CN104200224A (en) * | 2014-08-28 | 2014-12-10 | 西北工业大学 | Valueless image removing method based on deep convolutional neural networks |
| CN105096279A (en) * | 2015-09-23 | 2015-11-25 | 成都融创智谷科技有限公司 | Digital image processing method based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 吕国豪: "基于卷积神经网络的正则化方法", 《计算机研究与发展》 * |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016657B (en) * | 2017-04-07 | 2019-05-28 | 河北工业大学 | The restorative procedure of the face picture covered by reticulate pattern |
| CN107016657A (en) * | 2017-04-07 | 2017-08-04 | 河北工业大学 | The restorative procedure of the face picture covered by reticulate pattern |
| CN108734667A (en) * | 2017-04-14 | 2018-11-02 | Tcl集团股份有限公司 | A kind of image processing method and system |
| CN108230253A (en) * | 2017-05-08 | 2018-06-29 | 北京市商汤科技开发有限公司 | Image recovery method, device, electronic equipment and computer storage media |
| WO2019047655A1 (en) * | 2017-09-05 | 2019-03-14 | 百度在线网络技术（北京）有限公司 | Method and apparatus for use in determining driving behavior of driverless vehicle |
| CN107578054A (en) * | 2017-09-27 | 2018-01-12 | 北京小米移动软件有限公司 | Image processing method and device |
| CN107992894A (en) * | 2017-12-12 | 2018-05-04 | 北京小米移动软件有限公司 | Image-recognizing method, device and computer-readable recording medium |
| CN108876726A (en) * | 2017-12-12 | 2018-11-23 | 北京旷视科技有限公司 | Method, apparatus, system and the computer storage medium of image procossing |
| CN107992894B (en) * | 2017-12-12 | 2022-02-08 | 北京小米移动软件有限公司 | Image recognition method, image recognition device and computer-readable storage medium |
| CN108010538A (en) * | 2017-12-22 | 2018-05-08 | 北京奇虎科技有限公司 | Audio data processing method and device, computing device |
| CN108010538B (en) * | 2017-12-22 | 2021-08-24 | 北京奇虎科技有限公司 | Audio data processing method and device and computing equipment |
| CN109564638A (en) * | 2018-01-15 | 2019-04-02 | 深圳鲲云信息科技有限公司 | Artificial intelligence process device and its applied processing method |
| WO2019136762A1 (en) * | 2018-01-15 | 2019-07-18 | 深圳鲲云信息科技有限公司 | Artificial intelligence processor and processing method applied thereto |
| CN108805884A (en) * | 2018-06-13 | 2018-11-13 | 北京搜狐新媒体信息技术有限公司 | A kind of mosaic area's detection method, device and equipment |
| CN109410123B (en) * | 2018-10-15 | 2023-08-18 | 深圳市能信安科技股份有限公司 | Deep learning-based mosaic removing method and device and electronic equipment |
| CN109410123A (en) * | 2018-10-15 | 2019-03-01 | 深圳市能信安科技股份有限公司 | The method, apparatus and electronic equipment of removal mosaic based on deep learning |
| CN111351502A (en) * | 2018-12-21 | 2020-06-30 | 赫尔环球有限公司 | Method, apparatus and computer program product for generating an overhead view of an environment from a perspective view |
| US11720992B2 (en) | 2018-12-21 | 2023-08-08 | Here Global B.V. | Method, apparatus, and computer program product for generating an overhead view of an environment from a perspective image |
| US11895376B2 (en) | 2019-03-25 | 2024-02-06 | Rovi Guides, Inc. | Systems and methods for creating customized content |
| CN110111282A (en) * | 2019-05-09 | 2019-08-09 | 杭州电子科技大学上虞科学与工程研究院有限公司 | A kind of video deblurring method based on motion vector and CNN |
| CN110111282B (en) * | 2019-05-09 | 2021-05-11 | 杭州电子科技大学上虞科学与工程研究院有限公司 | Video deblurring method based on motion vector and CNN |
| US11562016B2 (en) | 2019-06-26 | 2023-01-24 | Rovi Guides, Inc. | Systems and methods for generating supplemental content for media content |
| US11256863B2 (en) | 2019-07-19 | 2022-02-22 | Rovi Guides, Inc. | Systems and methods for generating content for a screenplay |
| US11934777B2 (en) | 2019-07-19 | 2024-03-19 | Rovi Guides, Inc. | Systems and methods for generating content for a screenplay |
| US11145029B2 (en) * | 2019-07-25 | 2021-10-12 | Rovi Guides, Inc. | Automated regeneration of low quality content to high quality content |
| WO2021135614A1 (en) * | 2019-12-31 | 2021-07-08 | 杭州海康威视数字技术股份有限公司 | Video processing method, apparatus, and system |
| US11604827B2 (en) | 2020-02-21 | 2023-03-14 | Rovi Guides, Inc. | Systems and methods for generating improved content based on matching mappings |
| US11914645B2 (en) | 2020-02-21 | 2024-02-27 | Rovi Guides, Inc. | Systems and methods for generating improved content based on matching mappings |
| US11336943B2 (en) * | 2020-02-24 | 2022-05-17 | Realtek Semiconductor Corporation | Method and electronic device for detecting resolution |
| CN111401453A (en) * | 2020-03-18 | 2020-07-10 | 西安电子科技大学 | Mosaic image classification and identification method and system |
| CN112329634A (en) * | 2020-11-05 | 2021-02-05 | 华中师范大学 | Classroom behavior recognition method and device, electronic equipment and storage medium |
| CN112184597A (en) * | 2020-11-05 | 2021-01-05 | 温州大学大数据与信息技术研究院 | Image restoration device and method |
| CN112329634B (en) * | 2020-11-05 | 2024-04-02 | 华中师范大学 | Classroom behavior identification method and device, electronic equipment and storage medium |
| CN113905276A (en) * | 2021-09-24 | 2022-01-07 | 威视芯半导体(合肥)有限公司 | Video coding processing method, system, equipment and storage medium |
| CN113905276B (en) * | 2021-09-24 | 2023-11-14 | 威视芯半导体(杭州)有限公司 | Video coding processing method, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106530227B (en) | 2019-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106530227A (en) | Image restoration method and device | |
| CN106339680B (en) | Face key independent positioning method and device | |
| CN104731688B (en) | Point out the method and device of reading progress | |
| CN105809704A (en) | Method and device for identifying image definition | |
| CN106651955A (en) | Method and device for positioning object in picture | |
| CN107798669A (en) | Image defogging method, device and computer-readable recording medium | |
| CN106682736A (en) | Image identification method and apparatus | |
| CN107948708A (en) | Barrage methods of exhibiting and device | |
| CN106993229A (en) | Interactive attribute methods of exhibiting and device | |
| CN106778773A (en) | The localization method and device of object in picture | |
| CN104077563B (en) | Face identification method and device | |
| CN105354560A (en) | Fingerprint identification method and device | |
| CN107832746A (en) | Expression recognition method and device | |
| CN106228556A (en) | Image quality analysis method and device | |
| CN105094539B (en) | Reference information display methods and device | |
| CN106919629A (en) | The method and device of information sifting is realized in group chat | |
| CN105975540A (en) | Information display method and device | |
| CN107463903A (en) | Face key independent positioning method and device | |
| CN107563994A (en) | The conspicuousness detection method and device of image | |
| CN107797741A (en) | Method for showing interface and device | |
| CN107729880A (en) | Method for detecting human face and device | |
| CN107967459A (en) | convolution processing method, device and storage medium | |
| CN107943550A (en) | Method for showing interface and device | |
| CN106534951A (en) | Method and apparatus for video segmentation | |
| CN105825490A (en) | Gaussian blur method and device of image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |