CN110062282A - A kind of super-resolution video method for reconstructing, device and electronic equipment - Google Patents
A kind of super-resolution video method for reconstructing, device and electronic equipment Download PDFInfo
- Publication number
- CN110062282A CN110062282A CN201910204955.1A CN201910204955A CN110062282A CN 110062282 A CN110062282 A CN 110062282A CN 201910204955 A CN201910204955 A CN 201910204955A CN 110062282 A CN110062282 A CN 110062282A
- Authority
- CN
- China
- Prior art keywords
- resolution
- component
- texture
- gpu
- model data
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 15
- 238000007667 floating Methods 0.000 claims abstract description 9
- 238000013500 data storage Methods 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 5
- 230000015572 biosynthetic process Effects 0.000 claims 1
- 230000008569 process Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000003362 replicative effect Effects 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/4402—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/4402—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display
- H04N21/440218—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display by transcoding between formats or standards, e.g. from MPEG-2 to MPEG-4
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/4402—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display
- H04N21/440263—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display by altering the spatial resolution, e.g. for displaying on a connected PDA
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Generation (AREA)
Abstract
The embodiment of the invention provides a kind of super-resolution video method for reconstructing, device and electronic equipments, the video to be processed of the acquisition first resolution of CPU first, each video frame of first resolution in video to be processed is converted into yuv format, Y-component Texture and UV the component Texture under the first resolution of the video frame is obtained, and is sent to GPU；GPU accesses the GPU model data saved, carries out convolution algorithm to Y-component Texture and GPU model data, obtains the Y-component Texture of second resolution；Then, the UV component Texture of first resolution is converted into second resolution, then merged with the Y-component Texture of second resolution, the video frame of the second resolution after being rebuild.Since the model data of Super-resolution reconstruction established model to be loaded as to the GPU model data of floating point type, therefore it can be accessed by GPU, and degree of parallelism is high when performing image processing by GPU, Caton phenomenon can occurs to avoid playing when video, the real-time super-resolution rate for realizing video is rebuild.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of super-resolution video method for reconstructing, device and
Electronic equipment.
Background technique
As client shows that the resolution ratio of equipment is higher and higher, user requires also gradually to mention to the clarity of viewing video
It is high.But it is limited sometimes by upstream bandwidth or flow, user can only but select the video-see of low resolution, this is just unrestrained in vain
The high definition screen of display equipment is taken.
Super-resolution rebuilding is the item of image amplifying technique rapidly developed in recent years, can utilize one or more
Low-resolution image restores high-definition picture, is apparent from fuzzy image.Currently, being carried out on picture based on nerve net
The super-resolution rebuilding algorithm of network is existing many.
However, different from picture, in video display process, the decoding and display of video need to consume the regular hour, because
This is very high to the requirement of real-time of super-resolution rebuilding algorithm in order to keep broadcasting pictures smooth, and will be neural network based super
When resolution reconstruction algorithm is applied to video, due to algorithm complexity, the super-resolution rebuilding process of each frame is all quite time-consuming, leads
Cause occur Caton in the playing process of video.
Summary of the invention
The embodiment of the present invention is designed to provide a kind of super-resolution video method for reconstructing, device and electronic equipment, with
It realizes and the real-time super-resolution rate of video is rebuild.Specific technical solution is as follows:
In a first aspect, the embodiment of the invention provides a kind of super-resolution video method for reconstructing, which is characterized in that be applied to
Terminal, the terminal include central processor CPU and graphics processor GPU, which comprises
The CPU obtains the video to be processed of first resolution；
The CPU obtains each video frame of the first resolution in the video to be processed, by the lattice of the video frame
Formula is converted to yuv format, obtains the Y-component texture Texture of the first resolution of the video frame and UV of first resolution
Component texture Texture；
The UV component Texture of the Y-component Texture of first resolution and first resolution is sent to by the CPU
GPU；
The GPU model data of the GPU access Saved Presets Super-resolution reconstruction established model；
The GPU carries out convolution algorithm to Y-component Texture and GPU model data, and the after obtaining super-resolution rebuilding
The Y-component Texture of two resolution ratio；
The UV component Texture of first resolution is converted to the UV component Texture of second resolution by the GPU；
The GPU merges the Y-component Texture of second resolution with the UV component Texture of second resolution, obtains
The video frame of second resolution；The second resolution is higher than the first resolution.
Optionally, the GPU model data is saved using following steps to GPU:
The model data that the CPU loads the Super-resolution reconstruction established model is floating type float, obtains float type
Model data；
The model data of the float type is saved as GPU model data to GPU.
Optionally, the GPU model data is saved using following steps to GPU:
The model data that the CPU loads the Super-resolution reconstruction established model is float type, obtains float type
Model data；
Using tri- components of R, G, B, the model data of the float type is stored, forms model data Texture；
It saves model data Texture as GPU model data to GPU.
Optionally, described to utilize tri- components of R, G, B, the model data of the float type is stored, model data is formed
The step of Texture, comprising:
Every 3 model values in the model data of the float type are stored to the R component of a pixel, G
In component and B component, model data Texture is formed.
Optionally, it by every 3 model values in the model data of the float type, stores to the R of a pixel
After step in component, G component and B component, further includes:
By each pixel duplication n × n, the model data Texture of the float type of n × n structure is formed；
Described the step of saving model data Texture to GPU as GPU model data, comprising:
The model data Texture of the float type of n × n structure is saved as GPU model data to GPU.
Optionally, the step of GPU model data for the Super-resolution reconstruction established model that the GPU access has saved, comprising:
The GPU is accessed and is indexed in the model data Texture according to the index of preset each available pixel point
R component, G component and the corresponding model value of B component of corresponding each available pixel point；
The GPU carries out convolution algorithm to Y-component Texture and GPU model data, and the after obtaining super-resolution rebuilding
The step of Y-component Texture of two resolution ratio, comprising:
The GPU is to indexing corresponding each available picture in the Y-component Texture and the model data Texture
R component, G component and the corresponding model value progress convolution algorithm of B component of vegetarian refreshments, second point after obtaining super-resolution rebuilding
The Y-component Texture of resolution.
Optionally, the UV component Texture of first resolution is converted to the UV component of second resolution by the GPU
The step of Texture, comprising:
The GPU carries out interpolation according to interpolation algorithm, to the UV component Texture of the first resolution, obtains second
The UV component Texture of resolution ratio.
Optionally, the CPU is by the UV component Texture of the Y-component Texture of first resolution and first resolution
Before the step of being sent to GPU, further includes:
Whether the super-resolution rebuilding function in CPU detection video-see interface is in the open state；If it is,
Execute the step that the UV component Texture by the Y-component Texture of first resolution and first resolution is sent to GPU
Suddenly.
Second aspect, the embodiment of the invention provides a kind of super-resolution video reconstructing devices, which is characterized in that is applied to
Terminal, described device include:
Video obtains module, for obtaining the video to be processed of first resolution；
Format converting module will be described for obtaining each video frame of the first resolution in the video to be processed
The format of video frame is converted to yuv format, and the Y-component Texture and first for obtaining the first resolution of the video frame is differentiated
The UV component Texture of rate；
Sending module, for sending out the UV component Texture of the Y-component Texture of first resolution and first resolution
Give GPU；
Access modules, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；
Convolution module obtains super-resolution rebuilding for carrying out convolution algorithm to Y-component Texture and GPU model data
The Y-component Texture of second resolution afterwards；
Resolution converting block, for the UV component Texture of first resolution to be converted to UV points of second resolution
Measure Texture；
Merging module, for closing the UV component Texture of the Y-component Texture of second resolution and second resolution
And obtain the video frame of second resolution；The second resolution is higher than the first resolution.
Optionally, described device, further includes:
First data storage module, for saving the GPU model data to GPU；
First data storage module, comprising:
First load submodule, the model data for loading the Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
First saves submodule, for saving the model data of the float type as GPU model data to GPU.
Optionally, described device, further includes:
Second data storage module, for saving the GPU model data to GPU；
Second data storage module, comprising:
Second load submodule, the model data for loading the Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
Texture forms submodule, for storing the model data of the float type using tri- components of R, G, B,
Form model data Texture；
Second saves submodule, for saving model data Texture as GPU model data to GPU.
Optionally, the Texture forms submodule, comprising:
Numerical value sub-module stored, for by every 3 model values in the model data of the float type, store to
In the R component of one pixel, G component and B component, model data Texture is formed.
Optionally, the Texture forms submodule, further includes:
Data replicate submodule, will be in the model data of the float type for numerical value sub-module stored execution
Every 3 model values, after storing the step into the R component of a pixel, G component and B component, by each pixel
N × n are replicated, the model data Texture of the float type of n × n structure is formed；
Described second saves submodule, is specifically used for making the model data Texture of the float type of n × n structure
It saves for GPU model data to GPU.
Optionally, the access modules access the mould specifically for the index according to preset each available pixel point
R component, G component and the corresponding model value of B component of corresponding each available pixel point are indexed in type data Texture；
The convolution module is specifically used for index pair in the Y-component Texture and the model data Texture
R component, G component and the corresponding model value of B component for each available pixel point answered carry out convolution algorithm, obtain super-resolution
The Y-component Texture of second resolution after reconstruction.
Optionally, the resolution converting block is specifically used for according to interpolation algorithm, to UV points of the first resolution
It measures Texture and carries out interpolation, obtain the UV component Texture of second resolution.
Optionally, described device, further includes:
Detection module, it is described by the Y-component Texture of first resolution and first point for being executed in the sending module
Before the step of UV component Texture of resolution is sent to GPU, the super-resolution rebuilding function of detecting in video-see interface is
It is no in the open state；If it is, trigger the sending module execute the Y-component Texture by first resolution and
The step of UV component Texture of first resolution is sent to GPU.
The third aspect, the embodiment of the invention provides a kind of electronic equipment, including CPU and GPU；
The CPU, for obtaining the video to be processed of first resolution；First obtained in the video to be processed is differentiated
The format of the video frame is converted to yuv format, obtains the Y of the first resolution of the video frame by each video frame of rate
The UV component Texture of component Texture and first resolution；By the Y-component Texture and first resolution of first resolution
UV component Texture be sent to GPU；
The GPU, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；To Y-component
Texture and GPU model data carries out convolution algorithm, the Y-component of the second resolution after obtaining super-resolution rebuilding
Texture；The UV component Texture of first resolution is converted to the UV component Texture of second resolution；Second is differentiated
The Y-component Texture of rate merges with the UV component Texture of second resolution, obtains the video frame of second resolution；Described
Two high resolutions are in the first resolution.
A kind of super-resolution video method for reconstructing provided in an embodiment of the present invention, device and electronic equipment, first CPU are obtained
The video to be processed for obtaining first resolution, is converted to yuv format for each video frame of the first resolution in video to be processed,
The Y-component Texture of the first resolution of video frame and the UV component Texture of first resolution are obtained, and is sent to GPU；
GPU accesses the GPU model data of Saved Presets Super-resolution reconstruction established model, to Y-component Texture and GPU model data
Convolution is carried out, the Y-component Texture of second resolution is obtained；Then, the UV component Texture of first resolution is converted to
Second resolution, then merge with the Y-component Texture of second resolution, the video frame of the second resolution after being rebuild.By
In the model data of Super-resolution reconstruction established model is loaded as floating type, therefore can be accessed by GPU, and GPU is carrying out figure
Degree of parallelism is high when image procossing, avoids the appearance of Caton phenomenon, and the real-time super-resolution rate for realizing video is rebuild.
Certainly, implement any of the products of the present invention or method it is not absolutely required at the same reach all the above excellent
Point.
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.
Fig. 1 is a kind of flow diagram of super-resolution video method for reconstructing provided in an embodiment of the present invention；
Fig. 2 is a kind of exemplary diagram of model data texture (Texture) provided in an embodiment of the present invention；
Fig. 3 is a kind of instance graph of video frame super-resolution rebuilding provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of super-resolution video reconstructing device provided in an embodiment of the present invention；
Fig. 5 is a kind of electronic equipment schematic diagram 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.
In order to realize that the real-time super-resolution rate of video is rebuild, a kind of super-resolution video reconstruction provided in an embodiment of the present invention
Method is applied to terminal, which includes central processor CPU and graphics processor GPU.Specifically, as shown in Figure 1, oversubscription
Resolution video method for reconstructing includes:
The video to be processed of S101, CPU acquisition first resolution.
S102, CPU obtain each video frame of the first resolution in video to be processed, and the format of video frame is converted to
Yuv format obtains the Y-component texture (Texture) of the first resolution of video frame and the UV component texture of first resolution
(Texture)。
Specifically, Y indicates luminance signal, and U, V respectively indicate two colour difference signal B-Y (i.e. U), R-Y in yuv format
(i.e. V).
The UV component Texture of the Y-component Texture of first resolution and first resolution is sent to by S103, CPU
GPU。
Optionally, the UV component Texture of the Y-component Texture of first resolution and first resolution is sent out in CPU
Before giving GPU, whether the super-resolution rebuilding function of can detecte in video-see interface is in the open state；If not,
It then indicates not needing to carry out super-resolution rebuilding to the video of viewing；If it is, executing the Y-component of first resolution
Texture, first resolution UV component Texture and GPU model data input GPU the step of.
The GPU model data of S104, GPU access Saved Presets Super-resolution reconstruction established model.
S105, GPU carry out convolution algorithm to Y-component Texture and GPU model data, after obtaining super-resolution rebuilding
The Y-component Texture of second resolution.
Super-resolution reconstruction established model in the embodiment of the present invention can be preparatory trained convolutional neural networks model；Its
In, GPU model data is the parameter of the Super-resolution reconstruction established model, and protects after being converted according to the data format requirement of GPU
There are in GPU, for carrying out convolution algorithm with the input data of Super-resolution reconstruction established model.
Specifically, GPU model data can be obtained using following three kinds of modes:
First way is: the model data that CPU loads Super-resolution reconstruction established model is floating type (float), is obtained
The model data of float type, and using the model data of float type as GPU model data, it saves to GPU.
The second way is: the model data that CPU loads Super-resolution reconstruction established model is float type, obtains float
The model data of type；Every 3 model values in the model data of float type are stored to R points of a pixel
In amount, G component and B component, model data Texture can be formed；It is protected model data Texture as GPU model data
It deposits to GPU.
The third mode is: the model data that CPU loads Super-resolution reconstruction established model is float type, obtains float
The model data of type；Then, it by every 3 model values in the model data of float type, stores to pixel
R component, G component and B component；Finally, each pixel duplication n × n is formed the model of the float type of n × n structure
Data Texture.
Fig. 2 is a kind of exemplary diagram of model data Texture provided in an embodiment of the present invention.By taking n=4 as an example, first will
Every 3 model values in the model data of float type, store to the R component of a pixel, G component and B component；So
Afterwards, each pixel is replicated 4 × 4, the model data Texture of the float type of 4 × 4 structures can be formed.Fig. 2
In, 4 × 4 structure of each of number 1-6 is obtained from replicating 4 × 4 parts as corresponding 1 pixel that is, each 4
The R component numerical value of 16 pixels in × 4 structures, G component values and B component numerical value are all identical.As shown, compiling
In numbers 5 model data Texture, R, G, the B component numerical value of 16 pixels be 0.019700, -0.026700 and -
0.030800, these three numerical value be Super-resolution reconstruction established model in filter parameter, for Texture volumes of Y-component
Product operation.
Optionally, in step S104~S105, GPU accesses the GPU mould of Saved Presets Super-resolution reconstruction established model
When type data, according to the index of preset each available pixel point, indexed in access model data Texture corresponding each
R component, G component and the corresponding model value of B component of available pixel point；It is then possible to using ZoomAI, to Y-component
R component, G component and the B component that corresponding each available pixel point is indexed in Texture and model data Texture are corresponding
Model value carries out convolution, the Y-component Texture of the second resolution after obtaining super-resolution rebuilding；Finally, utilizing bilinearity
Interpolation algorithm obtains the UV component Texture of second resolution to the UV component Texture interpolation of first resolution.Wherein,
ZoomAI is the video source modeling scheme of complete set, includes multiple trained convolutional neural networks models in advance in ZoomAI,
Different convolutional neural networks models for carrying out different optimization processings to client video, as super-resolution rebuilding, sharpening,
Denoising, color enhancement etc..Super-resolution reconstruction established model in the present embodiment is exactly trained in advance for surpassing in ZoomAI
The convolutional neural networks model of resolution reconstruction.
In order to make the super-resolution rebuilding result of the Y-component of first resolution seem softer, GPU can also first root
According to bilinear interpolation algorithm, the Y-component of first resolution is amplified to second resolution, then will be after the super-resolution rebuilding of acquisition
Second resolution Y-component Texture weighted sum therewith, the super-resolution rebuilding result as Y-component Texture.
The UV component Texture of first resolution is converted to the UV component Texture of second resolution by S106, GPU.
In step s 106, GPU can use interpolation algorithm, and the UV component Texture of first resolution is converted to the
The UV component Texture of two resolution ratio.Wherein, interpolation algorithm can be bilinear interpolation algorithm, Tri linear interpolation algorithm etc..
S107, GPU merge the Y-component Texture of second resolution with the UV component Texture of second resolution, obtain
To the video frame of second resolution；Second resolution is higher than first resolution.
Fig. 3 is a kind of instance graph of video frame super-resolution rebuilding provided in an embodiment of the present invention.As shown, in mobile phone
It holds and super-resolution rebuilding is carried out to the video frame of first resolution.When the first resolution of video frame is 667 × 375, in hand
Generator terminal A, mobile phone terminal B and mobile phone terminal C are amplified to the video frame of second resolution 1334 × 750, and the processing time is respectively
15ms, 20ms and 30ms；And when the first resolution of video frame is 960 × 540, being amplified to second resolution is 1920
× 1080 video frame, the processing time is respectively 25ms, 30ms and 40ms.Since human eye has eye storage characteristic, work as video
Frame per second when reaching 25 frames/second, user seems will be very smooth, and as shown in figure 3, two video frames are in different mobile phone terminals
The middle duration for carrying out super-resolution rebuilding, which is respectively less than, is equal to 40ms, i.e. the frame per second of video can achieve 25 frames/second.Therefore, will not
Video pictures are made Caton occur because of super-resolution rebuilding.
It is understood that the calculating due to GPU has the characteristics that highly-parallel, has CPU when handling vector data
The unexistent advantage of serial computing, thus it is more suitable for the calculating of the parallel vector in image procossing.Therefore, it is carried out using GPU
The reconstruction of super-resolution video can be with the treatment process of accelerating algorithm.Meanwhile it will be every in the model data of float type in algorithm
3 model values are stored into the R component of a pixel, G component and B component, have also been further speeded up GPU and have been read GPU mould
The speed of type data improves the real-time of algorithm.
A kind of super-resolution video method for reconstructing provided in an embodiment of the present invention, CPU first obtain first resolution to
Video is handled, each video frame of the first resolution in video to be processed is converted into yuv format, obtains the first of video frame
The Y-component Texture of the resolution ratio and UV component Texture of first resolution, and it is sent to GPU；GPU access has saved pre-
If the GPU model data of Super-resolution reconstruction established model, convolution is carried out to Y-component Texture and GPU model data, obtains second
The Y-component Texture of resolution ratio；Then, the UV component Texture of first resolution is converted into second resolution, then with
The Y-component Texture of two resolution ratio merges, the video frame of the second resolution after being rebuild.Due to by super-resolution rebuilding
The model data of model is loaded as floating type, therefore can be accessed by GPU, and GPU degree of parallelism when carrying out graph and image processing
It is high, the appearance of Caton phenomenon is avoided, the real-time super-resolution rate for realizing video is rebuild.
As a kind of embodiment of the embodiment of the present invention, as shown in figure 4, a kind of super-resolution provided in an embodiment of the present invention
Video reconstructing device is applied to terminal, comprising:
Video obtains module 410, for obtaining the video to be processed of first resolution；
Format converting module 420, for obtaining each video frame of the first resolution in video to be processed, by video frame
Format be converted to yuv format, obtain the Y-component Texture of the first resolution of video frame and the UV component of first resolution
Texture；
Sending module 430, for by the UV component of the Y-component Texture of first resolution and first resolution
Texture is sent to GPU；
Access modules 440, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；
Convolution module 450 obtains super-resolution for carrying out convolution algorithm to Y-component Texture and GPU model data
The Y-component Texture of second resolution after reconstruction；
Resolution converting block 460 is converted to the UV of second resolution for the UV component Texture by first resolution
Component Texture；
Merging module 470, for by the UV component of the Y-component Texture of second resolution and second resolution
Texture merges, and obtains the video frame of second resolution；Second resolution is higher than first resolution.
A kind of super-resolution video reconstructing device provided in an embodiment of the present invention, CPU first obtain first resolution to
Video is handled, each video frame of the first resolution in video to be processed is converted into yuv format, obtains the first of video frame
The Y-component Texture of the resolution ratio and UV component Texture of first resolution, and it is sent to GPU；GPU access has saved pre-
If the GPU model data of Super-resolution reconstruction established model, convolution is carried out to Y-component Texture and GPU model data, obtains second
The Y-component Texture of resolution ratio；Then, the UV component Texture of first resolution is converted into second resolution, then with
The Y-component Texture of two resolution ratio merges, the video frame of the second resolution after being rebuild.Due to by super-resolution rebuilding
The model data of model is loaded as floating type, therefore can be accessed by GPU, and GPU degree of parallelism when carrying out graph and image processing
It is high, the appearance of Caton phenomenon is avoided, the real-time super-resolution rate for realizing video is rebuild.
As a kind of embodiment of the embodiment of the present invention, super-resolution video reconstructing device, further includes:
First data storage module, for saving GPU model data to GPU；
First data storage module, comprising:
First load submodule, the model data for loading Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
First saves submodule, for saving the model data of float type as GPU model data to GPU.
As a kind of embodiment of the embodiment of the present invention, super-resolution video reconstructing device, further includes:
Second data storage module, for saving GPU model data to GPU；
Second data storage module, comprising:
Second load submodule, the model data for loading Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
Texture forms submodule, for utilizing tri- components of R, G, B, stores the model data of float type, is formed
Model data Texture；
Second saves submodule, for saving model data Texture as GPU model data to GPU.
As a kind of embodiment of the embodiment of the present invention, Texture forms submodule, may include:
Numerical value sub-module stored, for storing every 3 model values in the model data of float type to one
In the R component of pixel, G component and B component, model data Texture is formed.
As a kind of embodiment of the embodiment of the present invention, Texture forms submodule, may include:
Numerical value sub-module stored, for storing every 3 model values in the model data of float type to one
In the R component of pixel, G component and B component, model data Texture is formed.
As a kind of embodiment of the embodiment of the present invention, Texture forms submodule, further includes:
Data replicate submodule, execute for numerical value sub-module stored by every 3 in the model data of float type
Model value, after storing the step into the R component of a pixel, G component and B component, by each pixel replicate n ×
N, form the model data Texture of the float type of n × n structure；
Second saves submodule, is specifically used for using the model data Texture of the float type of n × n structure as GPU
Model data is saved to GPU.
As a kind of embodiment of the embodiment of the present invention, access modules 440 are specifically used for according to preset each available
The index of pixel indexes the R component of corresponding each available pixel point, G component and B minutes in access model data Texture
Measure corresponding model value；
Convolution module 450, specifically for indexed in Y-component Texture and model data Texture it is corresponding it is each can
Convolution algorithm is carried out with the R component of pixel, G component and the corresponding model value of B component, after obtaining super-resolution rebuilding
The Y-component Texture of two resolution ratio.
As a kind of embodiment of the embodiment of the present invention, resolution converting block 460, specifically for being calculated according to interpolation
Method carries out interpolation to the UV component Texture of first resolution, obtains the UV component Texture of second resolution.
As a kind of embodiment of the embodiment of the present invention, super-resolution video reconstructing device, further includes:
Detection module, for executing in sending module by the Y-component Texture of the first resolution and UV of first resolution
Before the step of component Texture is sent to GPU, whether the super-resolution rebuilding function of detecting in video-see interface is in out
Open state；If it is, triggering sending module is executed the UV of the Y component Texture of first resolution and first resolution points
The step of amount Texture is sent to GPU.
A kind of super-resolution video reconstructing device provided in an embodiment of the present invention, CPU first obtain first resolution to
Video is handled, each video frame of the first resolution in video to be processed is converted into yuv format, obtains the first of video frame
The Y-component Texture of the resolution ratio and UV component Texture of first resolution, and it is sent to GPU；GPU access has saved pre-
If the GPU model data of Super-resolution reconstruction established model, convolution is carried out to Y-component Texture and GPU model data, obtains second
The Y-component Texture of resolution ratio；Then, the UV component Texture of first resolution is converted into second resolution, then with
The Y-component Texture of two resolution ratio merges, the video frame of the second resolution after being rebuild.Due to by super-resolution rebuilding
The model data of model is loaded as floating type, therefore can be accessed by GPU, and GPU degree of parallelism when carrying out graph and image processing
It is high, the appearance of Caton phenomenon is avoided, the real-time super-resolution rate for realizing video is rebuild.
The embodiment of the invention also provides a kind of electronic equipment, as shown in figure 5, including central processor CPU 501 and figure
Shape processor GPU502, wherein
Central processor CPU 501, for obtaining the video to be processed of first resolution；Obtain in video to be processed
The format of video frame is converted to yuv format by each video frame of one resolution ratio, obtains Y points of the first resolution of video frame
Measure the UV component Texture of Texture and first resolution；By the Y-component Texture of first resolution and first resolution
UV component Texture is sent to GPU；
Graphics processor GPU502, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；It is right
Y-component Texture and GPU model data carries out convolution algorithm, the Y-component of the second resolution after obtaining super-resolution rebuilding
Texture；The UV component Texture of first resolution is converted to the UV component Texture of second resolution；Second is differentiated
The Y-component Texture of rate merges with the UV component Texture of second resolution, obtains the video frame of second resolution；Second
High resolution is in the first resolution.
A kind of electronic equipment provided in an embodiment of the present invention, first CPU obtain the video to be processed of first resolution, will be to
Each video frame of first resolution in processing video is converted to yuv format, obtains the Y-component of the first resolution of video frame
The UV component Texture of Texture and first resolution, and it is sent to GPU；GPU accesses Saved Presets Super-resolution reconstruction
The GPU model data of established model carries out convolution to Y-component Texture and GPU model data, obtains Y points of second resolution
Measure Texture；Then, the UV component Texture of first resolution is converted into second resolution, then the Y with second resolution
Component Texture merges, the video frame of the second resolution after being rebuild.Due to by the pattern number of Super-resolution reconstruction established model
It according to being loaded as floating type, therefore can be accessed by GPU, and GPU degree of parallelism when carrying out graph and image processing is high, avoids card
The appearance for phenomenon of pausing, the real-time super-resolution rate for realizing video are rebuild.
The communication bus that above-mentioned electronic equipment is mentioned can be Peripheral Component Interconnect standard (Peripheral Component
Interconnect, PCI) bus or expanding the industrial standard structure (Extended Industry Standard
Architecture, EISA) bus etc..The communication bus can be divided into address bus, data/address bus, control bus etc..For just
It is only indicated with a thick line in expression, figure, it is not intended that an only bus or a type of bus.
Communication interface is for the communication between above-mentioned electronic equipment and other equipment.
Memory may include random access memory (Random Access Memory, RAM), also may include non-easy
The property lost memory (Non-Volatile Memory, NVM), for example, at least a magnetic disk storage.Optionally, memory may be used also
To be storage device that at least one is located remotely from aforementioned processor.
Above-mentioned processor can be general processor, including central processing unit (Central Processing Unit,
CPU), network processing unit (Network Processor, NP) etc.；It can also be digital signal processor (Digital
Signal Processing, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic
Device, discrete gate or transistor logic, discrete hardware components.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Each embodiment in this specification is all made of relevant mode and describes, 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 device reality
For applying example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method
Part explanation.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (17)
1. a kind of super-resolution video method for reconstructing, which is characterized in that be applied to terminal, the terminal includes central processing unit
CPU and graphics processor GPU, which comprises
The CPU obtains the video to be processed of first resolution；
The CPU obtains each video frame of the first resolution in the video to be processed, and the format of the video frame is turned
It is changed to yuv format, obtains the Y-component texture Texture of the first resolution of the video frame and the UV component of first resolution
Texture Texture；
The UV component Texture of the Y-component Texture of first resolution and first resolution is sent to GPU by the CPU；
The GPU model data of the GPU access Saved Presets Super-resolution reconstruction established model；
The GPU carries out convolution algorithm to Y-component Texture and GPU model data, and second point after obtaining super-resolution rebuilding
The Y-component Texture of resolution；
The UV component Texture of first resolution is converted to the UV component Texture of second resolution by the GPU；
The GPU merges the Y-component Texture of second resolution with the UV component Texture of second resolution, obtains second
The video frame of resolution ratio；The second resolution is higher than the first resolution.
2. the method according to claim 1, wherein the GPU model data, using following steps save to
GPU:
The model data that the CPU loads the Super-resolution reconstruction established model is floating type float, obtains the mould of float type
Type data；
The model data of the float type is saved as GPU model data to GPU.
3. the method according to claim 1, wherein the GPU model data, using following steps save to
GPU:
The model data that the CPU loads the Super-resolution reconstruction established model is float type, obtains the model of float type
Data；
Using tri- components of R, G, B, the model data of the float type is stored, forms model data Texture；
It saves model data Texture as GPU model data to GPU.
4. according to the method described in claim 3, it is characterized in that, described utilization tri- components of R, G, B, store the float
The step of model data of type, formation model data Texture, comprising:
Every 3 model values in the model data of the float type are stored to the R component of a pixel, G component
In B component, model data Texture is formed.
5. according to the method described in claim 4, it is characterized in that, by every 3 moulds in the model data of the float type
Type numerical value, after storing the step into the R component of a pixel, G component and B component, further includes:
By each pixel duplication n × n, the model data Texture of the float type of n × n structure is formed；
Described the step of saving model data Texture to GPU as GPU model data, comprising:
The model data Texture of the float type of n × n structure is saved as GPU model data to GPU.
6. according to the method described in claim 5, it is characterized in that, the GPU accesses the Super-resolution reconstruction established model saved
GPU model data the step of, comprising:
The GPU according to preset each available pixel point index, access indexed in the model data Texture it is corresponding
R component, G component and the corresponding model value of B component of each available pixel point；
The GPU carries out convolution algorithm to Y-component Texture and GPU model data, and second point after obtaining super-resolution rebuilding
The step of Y-component Texture of resolution, comprising:
The GPU is to indexing corresponding each available pixel point in the Y-component Texture and the model data Texture
R component, G component and the corresponding model value of B component carry out convolution algorithm, the second resolution after obtaining super-resolution rebuilding
Y-component Texture.
7. the method according to claim 1, wherein the GPU turns the UV component Texture of first resolution
The step of being changed to the UV component Texture of second resolution, comprising:
The GPU carries out interpolation according to interpolation algorithm, to the UV component Texture of the first resolution, obtains second and differentiates
The UV component Texture of rate.
8. the method according to claim 1, wherein the CPU by the Y-component Texture of first resolution and
Before the step of UV component Texture of first resolution is sent to GPU, further includes:
Whether the super-resolution rebuilding function in CPU detection video-see interface is in the open state；If it is, executing
The step of UV component Texture by the Y-component Texture of first resolution and first resolution is sent to GPU.
9. a kind of super-resolution video reconstructing device, which is characterized in that be applied to terminal, described device includes:
Video obtains module, for obtaining the video to be processed of first resolution；
Format converting module, for obtaining each video frame of the first resolution in the video to be processed, by the video
The format of frame is converted to yuv format, obtains the Y-component Texture and first resolution of the first resolution of the video frame
UV component Texture；
Sending module, for the UV component Texture of the Y-component Texture of first resolution and first resolution to be sent to
GPU；
Access modules, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；
Convolution module, for carrying out convolution algorithm to Y-component Texture and GPU model data, after obtaining super-resolution rebuilding
The Y-component Texture of second resolution；
Resolution converting block is converted to the UV component of second resolution for the UV component Texture by first resolution
Texture；
Merging module is obtained for merging the Y-component Texture of second resolution with the UV component Texture of second resolution
To the video frame of second resolution；The second resolution is higher than the first resolution.
10. device according to claim 9, which is characterized in that described device, further includes:
First data storage module, for saving the GPU model data to GPU；
First data storage module, comprising:
First load submodule, the model data for loading the Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
First saves submodule, for saving the model data of the float type as GPU model data to GPU.
11. device according to claim 9, which is characterized in that described device, further includes:
Second data storage module, for saving the GPU model data to GPU；
Second data storage module, comprising:
Second load submodule, the model data for loading the Super-resolution reconstruction established model is float type, is obtained
The model data of float type；
Texture forms submodule, for utilizing tri- components of R, G, B, stores the model data of the float type, is formed
Model data Texture；
Second saves submodule, for saving model data Texture as GPU model data to GPU.
12. device according to claim 11, which is characterized in that the Texture forms submodule, comprising:
Numerical value sub-module stored, for storing every 3 model values in the model data of the float type to one
In the R component of pixel, G component and B component, model data Texture is formed.
13. device according to claim 12, which is characterized in that the Texture forms submodule, further includes:
Data replicate submodule, execute for the numerical value sub-module stored by every 3 in the model data of the float type
Each pixel after storing the step into the R component of a pixel, G component and B component, is replicated n by a model value
× n, form the model data Texture of the float type of n × n structure；
Described second saves submodule, is specifically used for using the model data Texture of the float type of n × n structure as GPU
Model data is saved to GPU.
14. device according to claim 13, which is characterized in that
The access modules access the model data specifically for the index according to preset each available pixel point
R component, G component and the corresponding model value of B component of corresponding each available pixel point are indexed in Texture；
The convolution module, specifically for corresponding to being indexed in the Y-component Texture and the model data Texture
R component, G component and the corresponding model value of B component of each available pixel point carry out convolution algorithm, obtain super-resolution rebuilding
The Y-component Texture of second resolution afterwards.
15. device according to claim 9, which is characterized in that the resolution converting block is specifically used for according to interpolation
Algorithm carries out interpolation to the UV component Texture of the first resolution, obtains the UV component Texture of second resolution.
16. device according to claim 9, which is characterized in that described device, further includes:
Detection module, for executing the Y-component Texture and first resolution by first resolution in the sending module
UV component Texture the step of being sent to GPU before, whether the super-resolution rebuilding function of detecting in video-see interface is located
In open state；If it is, trigger the sending module execute it is described by the Y-component Texture of first resolution and first
The step of UV component Texture of resolution ratio is sent to GPU.
17. a kind of terminal, which is characterized in that including CPU and GPU；
The CPU, for obtaining the video to be processed of first resolution；Obtain the first resolution in the video to be processed
The format of the video frame is converted to yuv format, obtains the Y-component of the first resolution of the video frame by each video frame
The UV component Texture of Texture and first resolution；By the Y-component Texture of the first resolution and UV of first resolution
Component Texture is sent to GPU；
The GPU, for accessing the GPU model data of Saved Presets Super-resolution reconstruction established model；To Y-component Texture
Convolution algorithm, the Y-component Texture of the second resolution after obtaining super-resolution rebuilding are carried out with GPU model data；By first
The UV component Texture of resolution ratio is converted to the UV component Texture of second resolution；By the Y-component of second resolution
Texture merges with the UV component Texture of second resolution, obtains the video frame of second resolution；The second resolution
Higher than the first resolution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910204955.1A CN110062282A (en) | 2019-03-18 | 2019-03-18 | A kind of super-resolution video method for reconstructing, device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910204955.1A CN110062282A (en) | 2019-03-18 | 2019-03-18 | A kind of super-resolution video method for reconstructing, device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110062282A true CN110062282A (en) | 2019-07-26 |
Family
ID=67317055
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910204955.1A Pending CN110062282A (en) | 2019-03-18 | 2019-03-18 | A kind of super-resolution video method for reconstructing, device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110062282A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110868625A (en) * | 2019-11-22 | 2020-03-06 | 北京金山云网络技术有限公司 | Video playing method and device, electronic equipment and storage medium |
| CN111314741A (en) * | 2020-05-15 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Video super-resolution processing method and device, electronic equipment and storage medium |
| CN112801879A (en) * | 2021-02-09 | 2021-05-14 | 咪咕视讯科技有限公司 | Image super-resolution reconstruction method and device, electronic equipment and storage medium |
| CN113538211A (en) * | 2020-04-22 | 2021-10-22 | 华为技术有限公司 | Image quality enhancement device and related method |
| CN114339412A (en) * | 2021-12-30 | 2022-04-12 | 咪咕文化科技有限公司 | Video quality enhancement method, mobile terminal, storage medium and device |
| CN114501141A (en) * | 2022-01-04 | 2022-05-13 | 杭州网易智企科技有限公司 | Video data processing method, device, equipment and medium |
| CN115460461A (en) * | 2022-09-07 | 2022-12-09 | 北京奇艺世纪科技有限公司 | Video processing method and device, terminal equipment and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103295192A (en) * | 2013-05-08 | 2013-09-11 | 西安电子科技大学 | Image real-time super-resolution reconstruction method based on acceleration of GPU |
| WO2014114635A1 (en) * | 2013-01-24 | 2014-07-31 | Thomson Licensing | Method and apparatus for performing super-resolution of single images |
| CN108259997A (en) * | 2018-04-02 | 2018-07-06 | 腾讯科技（深圳）有限公司 | Image correlation process method and device, intelligent terminal, server, storage medium |
-
2019
- 2019-03-18 CN CN201910204955.1A patent/CN110062282A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014114635A1 (en) * | 2013-01-24 | 2014-07-31 | Thomson Licensing | Method and apparatus for performing super-resolution of single images |
| CN103295192A (en) * | 2013-05-08 | 2013-09-11 | 西安电子科技大学 | Image real-time super-resolution reconstruction method based on acceleration of GPU |
| CN108259997A (en) * | 2018-04-02 | 2018-07-06 | 腾讯科技（深圳）有限公司 | Image correlation process method and device, intelligent terminal, server, storage medium |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110868625A (en) * | 2019-11-22 | 2020-03-06 | 北京金山云网络技术有限公司 | Video playing method and device, electronic equipment and storage medium |
| CN113538211A (en) * | 2020-04-22 | 2021-10-22 | 华为技术有限公司 | Image quality enhancement device and related method |
| WO2021213336A1 (en) * | 2020-04-22 | 2021-10-28 | 华为技术有限公司 | Image quality enhancement device and related method |
| CN113538211B (en) * | 2020-04-22 | 2024-07-19 | 华为技术有限公司 | Image quality enhancement device and related method |
| CN111314741A (en) * | 2020-05-15 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Video super-resolution processing method and device, electronic equipment and storage medium |
| CN112801879A (en) * | 2021-02-09 | 2021-05-14 | 咪咕视讯科技有限公司 | Image super-resolution reconstruction method and device, electronic equipment and storage medium |
| CN112801879B (en) * | 2021-02-09 | 2023-12-08 | 咪咕视讯科技有限公司 | Image super-resolution reconstruction method and device, electronic equipment and storage medium |
| CN114339412B (en) * | 2021-12-30 | 2024-02-27 | 咪咕文化科技有限公司 | Video quality enhancement method, mobile terminal, storage medium and device |
| CN114339412A (en) * | 2021-12-30 | 2022-04-12 | 咪咕文化科技有限公司 | Video quality enhancement method, mobile terminal, storage medium and device |
| CN114501141A (en) * | 2022-01-04 | 2022-05-13 | 杭州网易智企科技有限公司 | Video data processing method, device, equipment and medium |
| CN114501141B (en) * | 2022-01-04 | 2024-02-02 | 杭州网易智企科技有限公司 | Video data processing method, device, equipment and medium |
| CN115460461B (en) * | 2022-09-07 | 2024-04-26 | 北京奇艺世纪科技有限公司 | Video processing method and device, terminal equipment and computer readable storage medium |
| CN115460461A (en) * | 2022-09-07 | 2022-12-09 | 北京奇艺世纪科技有限公司 | Video processing method and device, terminal equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110062282A (en) | A kind of super-resolution video method for reconstructing, device and electronic equipment | |
| CN110136066B (en) | Video-oriented super-resolution method, device, equipment and storage medium | |
| CN108022212B (en) | High-resolution picture generation method, generation device and storage medium | |
| Zhang et al. | Residual dense network for image super-resolution | |
| CN109410123B (en) | Deep learning-based mosaic removing method and device and electronic equipment | |
| RU2697928C1 (en) | Superresolution of an image imitating high detail based on an optical system, performed on a mobile device having limited resources, and a mobile device which implements | |
| CN111951167B (en) | Super-resolution image reconstruction method, super-resolution image reconstruction device, computer equipment and storage medium | |
| CN111510739B (en) | Video transmission method and device | |
| CN110223256A (en) | A kind of inverse tone mapping (ITM) method, apparatus and electronic equipment | |
| US11983848B2 (en) | AI frame engine for mobile edge | |
| CN111739150B (en) | Noble metal three-dimensional model construction method and device | |
| CN109102463B (en) | Super-resolution image reconstruction method and device | |
| CN112929672A (en) | Video compression method, device, equipment and computer readable storage medium | |
| CN110838088A (en) | Multi-frame noise reduction method and device based on deep learning and terminal equipment | |
| CN116071279A (en) | Image processing method, device, computer equipment and storage medium | |
| CN107547773A (en) | A kind of image processing method, device and equipment | |
| CN111724448A (en) | Image super-resolution reconstruction method and device and terminal equipment | |
| CN109996085B (en) | Model training method, image processing method and device and electronic equipment | |
| CN113822803A (en) | Image super-resolution processing method, device, equipment and computer readable storage medium | |
| US20180096488A1 (en) | Method of texture synthesis and image processing apparatus using the same | |
| Dilip et al. | Bilinear interpolation image scaling processor for VLSI architecure | |
| CN115049572A (en) | Image processing method, image processing device, electronic equipment and computer readable storage medium | |
| CN113096010A (en) | Image reconstruction method and apparatus, and storage medium | |
| CN114119377A (en) | Image processing method and device | |
| CN103559692B (en) | Method and device for processing image |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190726 |
|
| RJ01 | Rejection of invention patent application after publication |