CN112884648A - Method and system for multi-class blurred image super-resolution reconstruction - Google Patents
Method and system for multi-class blurred image super-resolution reconstruction Download PDFInfo
- Publication number
- CN112884648A CN112884648A CN202110109414.8A CN202110109414A CN112884648A CN 112884648 A CN112884648 A CN 112884648A CN 202110109414 A CN202110109414 A CN 202110109414A CN 112884648 A CN112884648 A CN 112884648A
- Authority
- CN
- China
- Prior art keywords
- image
- super
- fuzzy
- resolution
- network
- 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 62
- 238000012549 training Methods 0.000 claims abstract description 86
- 238000012795 verification Methods 0.000 claims abstract description 24
- 238000004590 computer program Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 17
- 238000005259 measurement Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000008485 antagonism Effects 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 2
- 239000010410 layer Substances 0.000 description 24
- 238000010586 diagram Methods 0.000 description 19
- 238000013527 convolutional neural network Methods 0.000 description 15
- 230000008569 process Effects 0.000 description 11
- 238000003384 imaging method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 206010016322 Feeling abnormal Diseases 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000006835 compression Effects 0.000 description 4
- 238000007906 compression Methods 0.000 description 4
- 238000005286 illumination Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 238000005304 joining Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4046—Scaling of whole images or parts thereof, e.g. expanding or contracting using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The application relates to a method and a system for multi-class blurred image super-resolution reconstruction, wherein the method for multi-class blurred image super-resolution reconstruction comprises the following steps: acquiring a fuzzy image data set, and dividing the fuzzy image data set into a training set and a verification set; then training a fuzzy image classification network through a training set and a verification set to obtain an optimal classification network model, and inputting the fuzzy image into the model to obtain a fuzzy category of the corresponding classification fuzzy image; then training a super-resolution network through the labeled training data to obtain an optimal super-resolution network model; and finally, correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system. By the method and the device, the problem that corresponding high-definition images cannot be matched in a classified mode when super-resolution reconstruction is carried out on multi-type fuzzy images in the prior art is solved, and the practicability of the model is improved.
Description
Technical Field
The application relates to the field of image processing, in particular to a method and a system for multi-class blurred image super-resolution reconstruction.
Background
In the process of image acquisition, limited by factors such as imaging conditions and imaging modes, an imaging system cannot acquire all information in an original scene generally, and the quality of acquired images is reduced due to the influence of many factors such as deformation, blurring, down-sampling and noise in the imaging process. Therefore, on the premise of not changing imaging system hardware equipment, the quality of the acquired image is effectively improved, which is always an endeavoured and core problem in the imaging technology field, and the research of the technology has important significance. In the field of computer vision, Convolutional Neural Networks (CNNs) are widely used due to their outstanding image classification accuracy, and Super Resolution (SR) can effectively process detailed information and image quality of an image, enhance spatial Resolution, and obtain a high-precision image.
In the related art, there are currently several methods based on deep learning: dong et al introduces a convolutional neural network into the problem of image Super-resolution, designs an image Super-resolution restoration method (SRCNN for short) based on a deep convolutional neural network; kim et al propose an image super-resolution restoration method (VDSR for short) of an extremely deep network by referring to a VGG network structure for image classification on the basis of the SRCNN, VDSR can model a mapping relationship between a low-resolution image and a high-resolution image with a deeper network, however, since the VDSR method has 20 layers of deep networks and there is no interlayer information feedback and context information association therebetween, Kim et al propose a super-resolution method (DRCN for short) of a deep recursive convolutional neural network in order to solve the problem; christian et al uses a Generative countermeasure Network in a Super-Resolution technology, and proposes a Single Image Super-Resolution method (SRGAN for short) based on the Generative countermeasure Network to truly restore detailed information in an Image; furthermore, Bee Lim et al propose an Enhanced depth Residual network (EDSR) for Single Image Super Resolution. Although the multi-class blurred image super-resolution reconstruction method is available, the method still has some problems in practical application: images obtained in real multimedia application often have various degradation factors, such as low resolution, defocus blur, motion blur, low illumination, compression distortion, noise and the like, however, in practical application, a super-resolution reconstruction model for various blurred images is lacked, and the images cannot be matched with corresponding high-definition images in a classified manner.
At present, no effective solution is provided for the problem that when super-resolution reconstruction is performed on multiple types of blurred images in the related art, corresponding high-definition images cannot be matched in a classified manner.
Disclosure of Invention
The embodiment of the application provides a method and a system for super-resolution reconstruction of multiple types of blurred images, which at least solve the problem that corresponding high-definition images cannot be matched in a classified manner when super-resolution reconstruction is performed on the multiple types of blurred images in the related technology.
In a first aspect, an embodiment of the present application provides a method for multi-class blurred image super-resolution reconstruction, where the method includes:
acquiring a fuzzy image data set, and dividing the fuzzy image data set into a training set and a verification set;
training a fuzzy image classification network through the fuzzy image training set and the verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image;
high-definition images corresponding to the fuzzy image data set correspond to the classified fuzzy images one by one to obtain labeled training data, and a super-resolution network is trained through the labeled training data to obtain an optimal super-resolution network model;
and correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system.
In some of these embodiments, said obtaining the blurred image data set comprises:
acquiring a head shadow measurement data set, and enabling high-definition images and fuzzy images in the head shadow measurement data set to correspond one to obtain a head shadow measurement fuzzy image with a label;
processing the high-definition image through a function to obtain a fuzzy image;
and processing the unpaired high-definition images through a reactive network to obtain fuzzy images.
In some embodiments, the training of the fuzzy image classification network through the fuzzy image training set and the validation set, and the optimizing of the network structure and parameters to obtain the optimal classification network model includes:
optimizing the network through an iterative optimization algorithm, and updating the network weight by a weight trial and error method to obtain a global optimal solution;
deepening the classification network through a residual network.
In some embodiments, the one-to-one correspondence between the high-definition images corresponding to the blurred image data set and the classified blurred images to obtain the labeled training data includes:
labeling a high-definition image corresponding to the fuzzy image with the label, a high-definition image corresponding to the fuzzy image generated through a antagonism network, a high-definition image corresponding to the fuzzy image generated through a function and the classified fuzzy image in a one-to-one correspondence manner to obtain the labeled training data;
the labeled training data is saved as an HDF5 file.
In some embodiments, training the super-resolution network with the labeled training data to obtain an optimal super-resolution network model comprises:
calculating a loss function by a calculation method of Euclidean distance;
and performing convolution on the classified fuzzy image in the labeled training data, and activating.
In some embodiments, the calculating the loss function by the euclidean distance calculation method includes:
calculating the distance D between the original high-resolution image and the reconstructed high-resolution image:
IH＝(A1，A2，...，AW×H)
IS＝(a1，a2，...，aW×H)
where WXH is the pixel of the image matrix, IHIs an original high resolution image, ISIs a reconstructed high resolution image, a denotes the original high resolution image, and a denotes the reconstructed high resolution image.
In some embodiments, after obtaining the multi-class blurred image super-resolution reconstruction system, the method includes:
and inputting the blurry head shadow measurement image into the multi-class blurry image super-resolution reconstruction system to obtain a reconstructed multi-class super-resolution image.
In a second aspect, the present application provides a system for multi-class blurred image super-resolution reconstruction, where the system includes:
the system comprises an acquisition module, a verification module and a processing module, wherein the acquisition module is used for acquiring a fuzzy image data set and dividing the fuzzy image data set into a training set and a verification set;
the classification module is used for training a fuzzy image classification network through the fuzzy image training set and the verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image;
the super-resolution module is used for enabling high-definition images corresponding to the fuzzy image data set to correspond to the classified fuzzy images one by one to obtain labeled training data, and training a super-resolution network through the labeled training data to obtain an optimal super-resolution network model;
and the bridging module is used for correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and the processor is configured to execute the computer program to perform the method for multi-class blurred image super-resolution reconstruction described in any one of the above.
In a fourth aspect, the present application provides a storage medium having a computer program stored therein, where the computer program is configured to execute the method for multi-class blurred image super-resolution reconstruction described in any one of the above items when the computer program runs.
Compared with the related technology, the method for multi-class fuzzy image super-resolution reconstruction provided by the embodiment of the application acquires the fuzzy image data set, and divides the fuzzy image data set into the training set and the verification set; then, optimizing a network structure and parameters through the obtained fuzzy image training set and the verification set training fuzzy image classification network to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classification fuzzy image; then, high-definition images corresponding to the fuzzy image data set correspond to the classified fuzzy images one by one to obtain labeled training data, and a super-resolution network is trained through the labeled training data to obtain an optimal super-resolution network model; and finally, correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain a multi-class fuzzy image super-resolution reconstruction system, and inputting the fuzzy head shadow measurement image into the multi-class fuzzy image super-resolution reconstruction system to obtain a reconstructed multi-class super-resolution image. Because the image obtained in the real multimedia application is often a complex degraded low-quality image with various degraded factors coexisting, such as low resolution, defocus blur, motion blur, low illumination, compression distortion, noise and the like, and a super-resolution reconstruction model aiming at various blurred images is lacked in practical application, the problem that corresponding high-definition images cannot be matched in a classified manner when the super-resolution reconstruction is carried out on various blurred images in the prior art is solved by adopting the scheme, and the practicability of the model is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a schematic diagram of an application environment of a method for multi-class blurred image super-resolution reconstruction according to an embodiment of the application;
FIG. 2 is a flowchart of a multi-class blurred image super-resolution reconstruction method according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a classification network model according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a classification network joining a residual network according to an embodiment of the present application;
FIG. 5 is a flow diagram of blurred image classification according to an embodiment of the present application;
FIG. 6 is a schematic structural diagram of a super-resolution network model according to an embodiment of the present application;
FIG. 7 is a schematic flow chart of multi-class super-resolution images obtained by the multi-class blurred image super-resolution reconstruction system according to the embodiment of the application;
FIG. 8 is a block diagram of a multi-class blurred image super-resolution reconstruction system according to an embodiment of the present application;
fig. 9 is an internal structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments provided in the present application without any inventive step are within the scope of protection of the present application. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of ordinary skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms referred to herein shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar words throughout this application are not to be construed as limiting in number, and may refer to the singular or the plural. The present application is directed to the use of the terms "including," "comprising," "having," and any variations thereof, which are intended to cover non-exclusive inclusions; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or elements, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Reference to "connected," "coupled," and the like in this application is not intended to be limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. Reference herein to "a plurality" means greater than or equal to two. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. Reference herein to the terms "first," "second," "third," and the like, are merely to distinguish similar objects and do not denote a particular ordering for the objects.
The method for multi-class blurred image super-resolution reconstruction provided by the present application can be applied to an application environment shown in fig. 1, where fig. 1 is an application environment schematic diagram of the method for multi-class blurred image super-resolution reconstruction according to an embodiment of the present application, as shown in fig. 1, where a system of the application environment includes a server 10 and a device terminal 11, and the specific implementation process is as follows: acquiring a fuzzy image data set, and dividing the fuzzy image data set into a training set and a verification set; then, optimizing a network structure and parameters through the obtained fuzzy image training set and the verification set training fuzzy image classification network to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classification fuzzy image; then, high-definition images corresponding to the fuzzy image data set correspond to the classified fuzzy images one by one to obtain labeled training data, and a super-resolution network is trained through the labeled training data to obtain an optimal super-resolution network model; and finally, correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain a multi-class fuzzy image super-resolution reconstruction system, and inputting the fuzzy head shadow measurement image into the multi-class fuzzy image super-resolution reconstruction system to obtain a reconstructed multi-class super-resolution image. Because the image obtained in the real multimedia application is often a complex degraded low-quality image with various degraded factors coexisting, such as low resolution, defocus blur, motion blur, low illumination, compression distortion, noise and the like, and a super-resolution reconstruction model aiming at various blurred images is lacked in practical application, the problem that corresponding high-definition images cannot be matched in a classified manner when the super-resolution reconstruction is carried out on various blurred images in the prior art is solved by adopting the scheme, and the practicability of the model is improved.
The embodiment provides a method for multi-class blurred image super-resolution reconstruction, and fig. 2 is a flowchart of a multi-class blurred image super-resolution reconstruction method according to an embodiment of the present application, and as shown in fig. 2, the flowchart includes the following steps:
step S201, acquiring a fuzzy image data set, dividing the fuzzy image data set into a training set and a verification set, optionally acquiring a cephalometric image data set from a dental clinic, labeling the data set, and corresponding high-definition images and fuzzy images in the cephalometric image data set one by one to obtain a labeled cephalometric fuzzy image; then, processing the high-definition image through a function to obtain a fuzzy image, such as salt and pepper noise, Gaussian noise, smooth image and the like; in addition, a Cycle-Consistent adaptive network (CycleGAN) is used for processing unpaired high-definition images to obtain fuzzy images, wherein unpaired images are obtained by performing two-step transformation on source domain images: first trying to map it to the target domain and then returning the source domain results in a secondary generated image, eliminating the requirement for image pairing in the target domain, then using a generator network to map the image to the target domain and improving the quality of the generated image by matching the generator with the discriminator. The adversarial network refers to using a generator network and a discriminator network to perform mutual antagonism, wherein the generator tries to generate samples from the expected distribution, the discriminator tries to predict whether the samples are original images or generated images, and the generator and the discriminator are used for joint training, so that the final generator completely approaches to the actual distribution after learning.
In order to increase the model training speed subsequently, the obtained fuzzy image data set is stored in a Database by using a self-contained program convert _ image set of the cafe, optionally, the obtained image data set is stored in a Lightning Memory-Mapped Database (LMDB for short) in the embodiment, wherein the LMDB mainly serves to provide data management in the cafe, and converts the original data of shapes, colors, such as GPEG pictures and binary data, into a unified Key-Value for storage, so that the DataLayer of the cafe can conveniently obtain the data, and in addition, the network structure and the hyper-parameters in the training process can be configured, and a reasonable learning rate, a tchba _ size and the like can be selected;
and S202, training a fuzzy image classification network through a fuzzy image training set and a verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image.
Optionally, in this embodiment, the multi-path classification convolutional neural network is trained through the acquired fuzzy image training set and the validation set, and the network structure and the network parameters are adjusted through experiments, so that the classification performance of the model is improved. Fig. 3 is a schematic structural diagram of a classification network model according to an embodiment of the present application, and as shown in fig. 3, this embodiment makes full use of the capability of a convolutional neural network to extract image features and the property of a soft-max loss function to classify multiple classes of blurred images in the convolutional neural network, and adopts a deep convolutional neural network structure that classifies image data sets according to different blur classes of images. The deep convolutional neural network structure comprises N convolutional-recognition layer microstructures, after sample features are extracted from each convolutional layer (or convolutional layers) and a maximum pooling layer (max-pooling layer), a recognition full connection layer (identification layer) and a soft-max output layer (soft-max layer) are connected, and when weight parameters of the deep neural network are trained, the N recognition layers and the N soft-max layers are trained simultaneously, wherein the N convolutional-recognition layers can recognize N fuzzy images and calculate errors between the N fuzzy images and labels.
In some embodiments, the Loss function is minimized by optimizing the network through an iterative optimization algorithm, that is, alternately calling a forward algorithm and a backward algorithm to update the network parameters, wherein Caffe provides six optimization algorithms to solve the optimal parameters, and in the solution configuration file, the type is set to select. In addition, in the embodiment, the network weight is updated by a weight trial and error method to obtain a global optimal solution, that is, in the training process, the weight is updated by randomly selecting epsilon 100% (0< epsilon <1) from the weights obtained by each back propagation according to a mode different from the previous mode, the possibility that the deep neural network training jumps out of the local optimal predicament is increased, the global optimal solution is reached, the image features are extracted by using a deep convolutional neural network model, the image features are clustered by using a clustering method, so that the optimized classified blurred image deep convolutional neural network is trained, and the steps are repeated until the model classification effect is not improved any more.
In addition, in some embodiments, a classification network is further deepened by adding a residual network, fig. 4 is a schematic structural diagram of the classification network added with the residual network according to the embodiment of the present application, and as shown in fig. 4, for an optimization problem of random gradient descent (SGD for short) caused by deep network gradient diffusion, a classification network structure is deepened by adding a residual network microstructure in the classification network, so that a classification capability of a blurred image is improved.
Repeatedly executing the steps until the error of the neural network output layer reaches the preset precision requirement or the training frequency reaches the maximum iteration frequency, ending the training to obtain an optimal classification network model, and inputting the fuzzy image to be classified in the step S201 into the optimal classification network model to obtain corresponding fuzzy classes of each classified fuzzy image, such as motion blur, defocus blur, low illumination, compression distortion and other fuzzy classes, wherein fig. 5 is a flow schematic diagram of fuzzy image classification according to the embodiment of the application, and is shown in fig. 5;
and S203, enabling the high-definition images corresponding to the fuzzy image data set to correspond to the classified fuzzy images one by one to obtain labeled training data, and training the super-resolution network through the labeled training data to obtain an optimal super-resolution network model.
Optionally, in this embodiment, a high-definition image corresponding to the blurred head image measurement image with a label, a high-definition image corresponding to the blurred image generated by the cyclic consistency countermeasure network, and a high-definition image corresponding to the blurred image generated by the function, and high-definition images corresponding to other blurred images are labeled in one-to-one correspondence with the classified blurred images obtained in step S202, so as to obtain training data with labels, and store the training data as an HDF5 file, where the HDF5 file is a file format capable of storing different types of images and data. Designing a network structure of the super-resolution network, determining the number of nodes of an input layer, the number of nodes of an output layer, the number of hidden layers and the number of nodes of a hidden layer of the network, randomly initializing a connection weight W and a bias b of each layer, giving a learning rate eta, selecting an activation function RELU, and selecting a Loss function Loss. The Loss function Loss is generally selected by using the peak-to-noise ratio PSNR, and the calculation formula is shown in the following formulas 1 and 2:
where MSE is the mean square error between the original image and the processed image, MAXIIs the maximum value of the image color, W, H denotes the width and height of the image, i, j denotes the position of the image pixel, a denotes the original high resolution image, a denotes the reconstructed high resolution image.
In some embodiments, the loss function is calculated by a Euclidean distance calculation method, and in this embodiment, the image matrix has W × H elements, i.e., pixels, and the W × H element values (A) are used1，A2，...，AW×H) One-dimensional vector constituting the original high-resolution image, using (a)1，a2，...，aW×H) One-dimensional vector of reconstructed high-resolution image is formed, and original high-resolution image I is calculated by using Euclidean distance calculation methodHAnd reconstructing a high resolution image ISThe distance D therebetween, as shown in the following formulas 3 to 5:
IH＝(A1，A2，...，AW×H) (3)
IS＝(a1，a2，...，aW×H) (4)
where, wxh is a pixel point of an image matrix, a represents an original high resolution image, and a represents a reconstructed high resolution image.
Further, in the present embodiment, the original high resolution image IHAnd reconstructing a high resolution image ISThe distance D therebetween is shown in the following formula 6:
wherein: i.e. i1＝0，1，...，Wj1＝0，1，...，H。
The smaller the calculated D value is, the smaller the distance between the original high-resolution image and the reconstructed high-resolution image is, which shows that the two images are more similar, and the authenticity of the reconstructed high-resolution image can be effectively improved.
After the super-resolution network design is completed, in some embodiments, the classified images are convolved, and the process is activated, wherein the selected convolution kernel is 3 × 3. Since the size of the image is reduced after each convolution, the image is subjected to a 0 complementing operation before the next convolution, and the original size is restored, wherein the 0 complementing operation means that a pad value is set to be 1, and then four edges of the image are extended by 1 pixel, namely, the width and the height are extended by 2 pixels, and the pad is a variable in a depth learning framework caffe and is used for obtaining a related command for extending the edges of the image. In addition, the network is deepened through a depth residual error network, the gradient is prevented from disappearing, the high-order features and the low-order features are fused through the residual error network, and a reconstructed high-resolution image is obtained after the high-resolution image passes through a plurality of residual error blocks, so that the depth of the super-resolution network is guaranteed, and the network can be trained easily.
Repeatedly executing the super-resolution network training step until the error of the neural network output layer reaches the preset precision requirement or the training frequency reaches the maximum iteration frequency, ending the training, and storing the optimized network structure and parameters to obtain a trained optimal super-resolution network model, wherein fig. 6 is a structural schematic diagram of the super-resolution network model according to the embodiment of the application, as shown in fig. 6, the super-resolution network model comprises a convolutional layer (convolutional layer), a resblock network structure layer and a custom resolution layer (upscale), and when a fuzzy image is input into the super-resolution network model, the output image is a high-definition image;
and S204, correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system. Optionally, in this embodiment, the trained optimal multi-channel fuzzy head shadow measurement image convolutional neural network classification model and the optimal super-resolution network model are correspondingly combined, that is, each class of fuzzy classification network channel is bridged with the corresponding matched super-resolution model to obtain a multi-class fuzzy image super-resolution reconstruction system, the multi-class fuzzy images can be directly subjected to super-resolution reconstruction, the corresponding high-definition images are matched in classes, and the model is applied to actual life, so that the practicability of the model is improved.
Through the steps S201 to S204, compared with the prior art, in practical application, when performing super-resolution reconstruction on multiple types of blurred images, the problem that the corresponding high-definition images cannot be matched in a classified manner is solved. The embodiment acquires a fuzzy image data set, and divides the fuzzy image data set into a training set and a verification set; then, optimizing a network structure and parameters through the obtained fuzzy image training set and the verification set training fuzzy image classification network to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classification fuzzy image; then, high-definition images corresponding to the fuzzy image data set correspond to the classified fuzzy images one by one to obtain labeled training data, and a super-resolution network is trained through the labeled training data to obtain an optimal super-resolution network model; and finally, correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain a multi-class fuzzy image super-resolution reconstruction system, and inputting the fuzzy head shadow measurement image into the multi-class fuzzy image super-resolution reconstruction system to obtain a reconstructed multi-class super-resolution image. The problem that corresponding high-definition images cannot be matched in a classified mode when super-resolution reconstruction is conducted on multi-type fuzzy images in the prior art is solved, and the practicability of the model is improved.
In some embodiments, after obtaining the multi-class blurred image super-resolution reconstruction system, the blurry head shadow measurement image is input into the multi-class blurred image super-resolution reconstruction system, and a reconstructed multi-class super-resolution image is obtained. Fig. 7 is a schematic flow chart of multi-class super-resolution images obtained by the multi-class blurred image super-resolution reconstruction system according to the embodiment of the present application, and as shown in fig. 7, in this embodiment, the total data of blurred images to be classified are input, and after various types of blurred images are separated by the classification network model, a blurred image I is obtained1 BObtaining a corresponding high-definition image I through a super-resolution Model11 HBlurred image I2 BObtaining a corresponding high-definition image I through a super-resolution Model22 HBlurred image I3 BObtaining a corresponding high-definition image I through a super-resolution Model33 HBlurred image I4 BObtaining a corresponding high-definition image I through a super-resolution Model44 HOther blurred images I5 BObtaining a corresponding high-definition image I through a super-resolution Model55 H。
It should be noted that the steps illustrated in the above-described flow diagrams or in the flow diagrams of the figures may be performed in a computer system, such as a set of computer-executable instructions, and that, although a logical order is illustrated in the flow diagrams, in some cases, the steps illustrated or described may be performed in an order different than here.
The embodiment also provides a system for multi-class blurred image super-resolution reconstruction, and the system is used for implementing the above embodiments and preferred embodiments, and the description of the system is omitted for brevity. As used hereinafter, the terms "module," "unit," "subunit," and the like may implement a combination of software and/or hardware for a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
Fig. 8 is a block diagram of a multi-class blurred image super-resolution reconstruction system according to an embodiment of the present application, and as shown in fig. 8, the system includes an acquisition module 81, a classification module 82, a super-resolution module 83, and a bridge module 84:
an obtaining module 81, configured to obtain a fuzzy image data set, and divide the fuzzy image data set into a training set and a verification set; the classification module 82 is used for training a fuzzy image classification network through a fuzzy image training set and a verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image; the super-resolution module 83 is configured to correspond the high-definition images corresponding to the fuzzy image data set to the classified fuzzy images one by one to obtain labeled training data, and train a super-resolution network through the labeled training data to obtain an optimal super-resolution network model; and the bridging module 84 is configured to perform corresponding bridging on the optimal classification network model and the optimal super-resolution network model to obtain a multi-class blurred image super-resolution reconstruction system.
Through the system, the acquisition module 81 acquires the fuzzy image data set, divides the fuzzy image data set into a training set and a verification set, and is used for subsequently training the fuzzy image classification network and testing the classification effect of the classification network model; the classification module 82 trains the fuzzy image classification network through the fuzzy image training set and the verification set, optimizes the network structure and parameters to obtain an optimal classification network model, inputs the fuzzy image into the optimal classification network model to obtain a corresponding fuzzy class of the classified fuzzy image, facilitates subsequent super-resolution reconstruction of the classified fuzzy image, and obtains a corresponding matched high-definition image; the super-resolution module 83 corresponds the high-definition images corresponding to the fuzzy image data set with the classified fuzzy images one by one to obtain labeled training data, and trains a super-resolution network through the labeled training data to obtain an optimal super-resolution network model, so that the authenticity of the reconstructed high-resolution images is improved; the bridge module 84 correspondingly bridges the optimal classification network model and the optimal super-resolution network model to obtain a multi-class blurred image super-resolution reconstruction system, the multi-class blurred images can be directly subjected to super-resolution reconstruction through the multi-class blurred image super-resolution reconstruction system, the corresponding high-definition images are matched in classes, the multi-class blurred image super-resolution reconstruction system is applied to actual life, and the practicability of the models is improved. The whole system solves the problem that corresponding high-definition images cannot be matched in a classified mode when super-resolution reconstruction is carried out on multi-type fuzzy images in the prior art, and the practicability of the model is improved.
The present invention will be described in detail with reference to the following application scenarios.
The invention aims to provide a method and a system for multi-class blurred image super-resolution reconstruction, and the flow steps of the technical scheme of the multi-class blurred image super-resolution reconstruction method in the embodiment comprise the following steps:
s1, designing a classification convolutional neural network model, and classifying the acquired fuzzy image data sets in different categories;
s2, designing a super-resolution network model corresponding to the head shadow measurement blurred image, and corresponding the classified blurred images to the respective super-resolution network models;
s3, bridging the classification convolutional neural network model and the super-resolution network model for classifying the head shadow measurement fuzzy images to obtain a multi-class fuzzy image super-resolution reconstruction system;
and S4, inputting the blurry head shadow measurement image into a multi-class blurry image super-resolution reconstruction system to obtain a plurality of reconstructed super-resolution images.
The present embodiment also provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor is configured to execute the computer program to perform the steps in any of the above method embodiments, where, since the application relates to an image algorithm, a hardware environment in the electronic device is a graphics card, and preferably, a high-performance graphics card is used to effectively increase an operation speed of the algorithm.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, wherein the transmission device is connected to the processor, and the input/output device is connected to the processor.
In addition, in combination with the method for multi-class blurred image super-resolution reconstruction in the foregoing embodiments, the embodiments of the present application may provide a storage medium to implement. The storage medium having stored thereon a computer program; the computer program, when executed by a processor, implements any one of the above-described embodiments of the method for multi-class blurred image super-resolution reconstruction.
In an embodiment, fig. 9 is a schematic internal structure diagram of an electronic device according to an embodiment of the present application, and as shown in fig. 9, there is provided an electronic device, which may be a server, and its internal structure diagram may be as shown in fig. 9. The electronic device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic equipment comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the electronic device is used for storing data. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement a method for multi-class blurred image super-resolution reconstruction.
Those skilled in the art will appreciate that the configuration shown in fig. 9 is a block diagram of only a portion of the configuration relevant to the present application, and does not constitute a limitation on the electronic device to which the present application is applied, and a particular electronic device may include more or less components than those shown in the drawings, or combine certain components, or have a different arrangement of components.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
It should be understood by those skilled in the art that various features of the above-described embodiments can be combined in any combination, and for the sake of brevity, all possible combinations of features in the above-described embodiments are not described in detail, but rather, all combinations of features which are not inconsistent with each other should be construed as being within the scope of the present disclosure.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for multi-class blurred image super-resolution reconstruction is characterized by comprising the following steps:
acquiring a fuzzy image data set, and dividing the fuzzy image data set into a training set and a verification set;
training a fuzzy image classification network through the fuzzy image training set and the verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image;
high-definition images corresponding to the fuzzy image data set correspond to the classified fuzzy images one by one to obtain labeled training data, and a super-resolution network is trained through the labeled training data to obtain an optimal super-resolution network model;
and correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system.
2. The method of claim 1, wherein said acquiring a blurred image data set comprises:
acquiring a head shadow measurement data set, and enabling high-definition images and fuzzy images in the head shadow measurement data set to correspond one to obtain a head shadow measurement fuzzy image with a label;
processing the high-definition image through a function to obtain a fuzzy image;
and processing the unpaired high-definition images through a reactive network to obtain fuzzy images.
3. The method of claim 1, wherein training the fuzzy image classification network through the fuzzy image training set and the validation set, and optimizing network structure and parameters to obtain an optimal classification network model comprises:
optimizing the network through an iterative optimization algorithm, and updating the network weight by a weight trial and error method to obtain a global optimal solution;
deepening the classification network through a residual network.
4. The method according to claim 1 or 2, wherein the one-to-one correspondence of the high definition images corresponding to the blurred image data set with the classified blurred images is obtained by:
labeling a high-definition image corresponding to the fuzzy image with the label, a high-definition image corresponding to the fuzzy image generated through a antagonism network, a high-definition image corresponding to the fuzzy image generated through a function and the classified fuzzy image in a one-to-one correspondence manner to obtain the labeled training data;
the labeled training data is saved as an HDF5 file.
5. The method of claim 1, wherein training a super-resolution network with the labeled training data to obtain an optimal super-resolution network model comprises:
calculating a loss function by a calculation method of Euclidean distance;
and performing convolution on the classified fuzzy image in the labeled training data, and activating.
6. The method of claim 5, wherein the calculating a loss function by the Euclidean distance calculation method comprises:
calculating the distance D between the original high-resolution image and the reconstructed high-resolution image:
IH＝(A1，A2，...，AW×H)
IS＝(a1，a2，...，aW×H)
where WXH is the pixel of the image matrix, IHIs an original high resolution image, ISIs a reconstructed high resolution image, a denotes the original high resolution image, and a denotes the reconstructed high resolution image.
7. The method of claim 1, wherein after obtaining the multi-class blurred image super-resolution reconstruction system, the method comprises:
and inputting the blurry head shadow measurement image into the multi-class blurry image super-resolution reconstruction system to obtain a reconstructed multi-class super-resolution image.
8. A system for multi-class blurred image super-resolution reconstruction, the system comprising:
the system comprises an acquisition module, a verification module and a processing module, wherein the acquisition module is used for acquiring a fuzzy image data set and dividing the fuzzy image data set into a training set and a verification set;
the classification module is used for training a fuzzy image classification network through the fuzzy image training set and the verification set, optimizing a network structure and parameters to obtain an optimal classification network model, and inputting the fuzzy image into the optimal classification network model to obtain a fuzzy category of the corresponding classified fuzzy image;
the super-resolution module is used for enabling high-definition images corresponding to the fuzzy image data set to correspond to the classified fuzzy images one by one to obtain labeled training data, and training a super-resolution network through the labeled training data to obtain an optimal super-resolution network model;
and the bridging module is used for correspondingly bridging the optimal classification network model and the optimal super-resolution network model to obtain the multi-class fuzzy image super-resolution reconstruction system.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the computer program to perform the method for multi-class blurred image super-resolution reconstruction according to any of claims 1 to 7.
10. A storage medium having a computer program stored therein, wherein the computer program is configured to execute the method of multi-class blurred image super-resolution reconstruction according to any one of claims 1 to 7 when the computer program runs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110109414.8A CN112884648A (en) | 2021-01-25 | 2021-01-25 | Method and system for multi-class blurred image super-resolution reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110109414.8A CN112884648A (en) | 2021-01-25 | 2021-01-25 | Method and system for multi-class blurred image super-resolution reconstruction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112884648A true CN112884648A (en) | 2021-06-01 |
Family
ID=76052674
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110109414.8A Pending CN112884648A (en) | 2021-01-25 | 2021-01-25 | Method and system for multi-class blurred image super-resolution reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112884648A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780444A (en) * | 2021-09-16 | 2021-12-10 | 平安科技（深圳）有限公司 | Tongue fur image classification model training method based on progressive learning |
| CN114723604A (en) * | 2022-02-21 | 2022-07-08 | 福建八萃网络科技有限公司 | Video super-resolution method based on sample data set optimization |
| CN117745725A (en) * | 2024-02-20 | 2024-03-22 | 阿里巴巴达摩院(杭州)科技有限公司 | Image processing method, image processing model training method, three-dimensional medical image processing method, computing device, and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133948A (en) * | 2017-05-09 | 2017-09-05 | 电子科技大学 | Image blurring and noise evaluating method based on multitask convolutional neural networks |
| CN108122197A (en) * | 2017-10-27 | 2018-06-05 | 江西高创保安服务技术有限公司 | A kind of image super-resolution rebuilding method based on deep learning |
| CN109102462A (en) * | 2018-08-01 | 2018-12-28 | 中国计量大学 | A kind of video super-resolution method for reconstructing based on deep learning |
-
2021
- 2021-01-25 CN CN202110109414.8A patent/CN112884648A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133948A (en) * | 2017-05-09 | 2017-09-05 | 电子科技大学 | Image blurring and noise evaluating method based on multitask convolutional neural networks |
| CN108122197A (en) * | 2017-10-27 | 2018-06-05 | 江西高创保安服务技术有限公司 | A kind of image super-resolution rebuilding method based on deep learning |
| CN109102462A (en) * | 2018-08-01 | 2018-12-28 | 中国计量大学 | A kind of video super-resolution method for reconstructing based on deep learning |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780444A (en) * | 2021-09-16 | 2021-12-10 | 平安科技（深圳）有限公司 | Tongue fur image classification model training method based on progressive learning |
| CN113780444B (en) * | 2021-09-16 | 2023-07-25 | 平安科技（深圳）有限公司 | Training method of tongue fur image classification model based on progressive learning |
| CN114723604A (en) * | 2022-02-21 | 2022-07-08 | 福建八萃网络科技有限公司 | Video super-resolution method based on sample data set optimization |
| CN114723604B (en) * | 2022-02-21 | 2023-02-10 | 福建八萃网络科技有限公司 | Video super-resolution method based on sample data set optimization |
| CN117745725A (en) * | 2024-02-20 | 2024-03-22 | 阿里巴巴达摩院(杭州)科技有限公司 | Image processing method, image processing model training method, three-dimensional medical image processing method, computing device, and storage medium |
| CN117745725B (en) * | 2024-02-20 | 2024-05-14 | 阿里巴巴达摩院(杭州)科技有限公司 | Image processing method, image processing model training method, three-dimensional medical image processing method, computing device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bulat et al. | To learn image super-resolution, use a gan to learn how to do image degradation first | |
| Chang et al. | Spatial-adaptive network for single image denoising | |
| Shen et al. | Deep semantic face deblurring | |
| US10740897B2 (en) | Method and device for three-dimensional feature-embedded image object component-level semantic segmentation | |
| Xu et al. | Learning deep structured multi-scale features using attention-gated crfs for contour prediction | |
| CN112308200B (en) | Searching method and device for neural network | |
| US20190087726A1 (en) | Hypercomplex deep learning methods, architectures, and apparatus for multimodal small, medium, and large-scale data representation, analysis, and applications | |
| CN112884648A (en) | Method and system for multi-class blurred image super-resolution reconstruction | |
| Cheong et al. | Deep CNN-based super-resolution using external and internal examples | |
| Shen et al. | Exploiting semantics for face image deblurring | |
| CN111476719A (en) | Image processing method, image processing device, computer equipment and storage medium | |
| CN110706214B (en) | Three-dimensional U-Net brain tumor segmentation method fusing condition randomness and residual error | |
| JP2023533907A (en) | Image processing using self-attention-based neural networks | |
| CN109961397B (en) | Image reconstruction method and device | |
| JP2024026745A (en) | Using imager with on-purpose controlled distortion for inference or training of artificial intelligence neural network | |
| Wei et al. | A-ESRGAN: Training real-world blind super-resolution with attention U-Net Discriminators | |
| CN116863194A (en) | Foot ulcer image classification method, system, equipment and medium | |
| Liu et al. | Facial image inpainting using multi-level generative network | |
| US20220269906A1 (en) | Neural network-based image-to-image translation | |
| CN107729885B (en) | Face enhancement method based on multiple residual error learning | |
| Wang et al. | Learning to hallucinate face in the dark | |
| Lee et al. | Robust image translation and completion based on dual auto-encoder with bidirectional latent space regression | |
| Alonso-Fernandez et al. | Super-resolution for selfie biometrics: Introduction and application to face and iris | |
| Bricman et al. | CocoNet: A deep neural network for mapping pixel coordinates to color values | |
| CN115512100A (en) | Point cloud segmentation method, device and medium based on multi-scale feature extraction and fusion |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210601 |