CN110766635A - Identification card digital portrait restoration method and device, storage medium and processor - Google Patents
Identification card digital portrait restoration method and device, storage medium and processor Download PDFInfo
- Publication number
- CN110766635A CN110766635A CN201911030852.4A CN201911030852A CN110766635A CN 110766635 A CN110766635 A CN 110766635A CN 201911030852 A CN201911030852 A CN 201911030852A CN 110766635 A CN110766635 A CN 110766635A
- Authority
- CN
- China
- Prior art keywords
- digital portrait
- portrait
- digital
- residual error
- card
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 64
- 238000013528 artificial neural network Methods 0.000 claims abstract description 22
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000011084 recovery Methods 0.000 claims abstract description 6
- 238000007906 compression Methods 0.000 claims description 18
- 230000006835 compression Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 10
- 230000006837 decompression Effects 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 230000000694 effects Effects 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000006731 degradation reaction Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000015556 catabolic process Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the technical field of image processing, and particularly discloses a method for recovering a digital portrait of a certificate card, which comprises the following steps: collecting digital portrait photos; establishing a digital portrait training database according to the digital portrait photos; establishing a depth residual error convolution neural network according to the digital portrait photo; performing depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model; and inputting the card digital portrait to be restored to the deep residual error network restoration model to obtain the restored card digital portrait. The invention also discloses a card digital portrait restoration device, a storage medium and a processor. The method for recovering the digital portrait of the identification card provided by the invention does not need a default specific image prior method adopted in the traditional digital image repairing method, has more pertinence to the recovery of the portrait of the identification card, and can obviously improve the recovery effect.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method for recovering a certificate card digital portrait, a device for recovering the certificate card digital portrait, a storage medium and a processor.
Background
In the existing image and video compression technology, block-based compression coding has been widely adopted, such as JPEG (discrete cosine transform), JPEG2000 (wavelet transform), h.264(AVC), H.265(HEVC), etc., which are currently the most common. In the quantization compression process, when the sampling rate is low, block degradation effects such as block blurring and the like occur in the compressed image, the visual quality is affected, and the possibility of further reducing the compression capacity is also prevented. In order to solve this problem, a number of post-processing techniques have been developed recently, that is, a certain transformation and decompression are performed, and then a certain degree of restoration is performed on the decompressed picture, so as to enhance the image definition and retain the image details.
These post-treatment methods can be broadly divided into two broad categories: an image enhancement method and an image restoration method. The image enhancement method mainly adopts some filtering algorithms to smooth the image in a space domain and a frequency domain, so as to weaken noise. The image restoration-based method usually uses some priori knowledge to formalize the whole restoration model into a ill-conditioned image restoration inverse problem, such as sparse prior, low rank prior, clustering similar prior, non-local self-similar prior and the like, wherein the non-local self-similar characteristic of the image is the prior information form which is adopted more at present. The image restoration algorithms are not very effective in restoring the certificate photo mainly based on the portrait, because the existing restoration algorithms are all performed on common natural images, and more, the existing restoration algorithms are processed under the condition of defaulting a certain prior model, and the method cannot be directly applied to the portrait photo.
In recent years, the second generation certificate system, i.e. the certificate card, has gained general attention at home and abroad and has begun to be applied and popularized in various fields. Different from the traditional paper certificate, the certificate card can be added with an electronic chip in a medium for storing data and can encrypt and protect the chip data by utilizing some encryption algorithms, and the safety and the practicability of the certificate card are obviously higher than those of the paper certificate. Due to the limitation of the identification card on the storage capacity, the image needs to be compressed and stored at a very high compression ratio, which causes a great loss of image information, and the authentication of the identification card portrait is seriously affected after the image is decompressed. At present, more identification card portrait compression and restoration technologies are still focused on compression and decompression links (identification cards and the like), research on post-processed identification card portrait restoration technologies is less, and at present, no optimal solution exists.
Disclosure of Invention
The invention provides a method for recovering a certificate card digital portrait, a device for recovering the certificate card digital portrait, a storage medium and a processor, which solve the problem of post-processing of the certificate card recovered portrait in the related technology.
As a first aspect of the present invention, there is provided an identification card digital portrait restoration method, including:
collecting digital portrait photos;
establishing a digital portrait training database according to the digital portrait photos;
establishing a depth residual error convolution neural network according to the digital portrait photo;
performing depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
and inputting the card digital portrait to be restored to the deep residual error network restoration model to obtain the restored card digital portrait.
Further, the establishing a digital portrait training database according to the digital portrait photos includes:
carrying out portrait positioning on the digital portrait photo to obtain a standardized portrait photo;
and performing wavelet transformation and decompression processing on the normalized portrait photo to obtain a digital portrait training database.
Further, the establishing a digital portrait training database according to the digital portrait photos further includes:
and carrying out mirror image transformation on the images in the digital portrait training database.
Further, the portrait positioning the digital portrait photo to obtain a normalized portrait photo includes:
carrying out portrait positioning on the digital portrait photo according to a multitask cascade convolution neural network and a skin color model algorithm;
and adjusting the resolution of the digital portrait photo to obtain the normalized portrait photo.
Further, the establishing a depth residual convolutional neural network according to the digital portrait photo includes:
establishing an image restoration model according to the digital portrait photo;
carrying out residual error processing on the image restoration model;
determining input and output data of the image restoration model to obtain a target function;
and performing expected learning according to the target function to obtain the depth residual convolution neural network.
Further, the depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model, including:
repeatedly training the depth residual convolution neural network according to the residual paired data set;
counting the peak signal-to-noise ratio of the restored image and the original image after each training, and recording the training times and parameters used when the target peak signal-to-noise ratio is reached;
and checking and network debugging the repeated training result according to the recorded training times and parameters when the target peak signal-to-noise ratio value is reached to obtain a deep residual error network restoration model.
Further, the inputting the license card digital portrait to be restored to the deep residual error network restoration model to obtain the restored license card digital portrait includes:
after the digital signature information in the identification card is verified, extracting image compression data of the identification card digital portrait to be restored;
performing wavelet decompression on the image compressed data to obtain a low-quality image;
performing residual prediction according to the depth residual network restoration model to obtain a prediction residual;
and calculating according to the prediction residual error and the low-quality image to obtain a recovered certificate card digital portrait.
As another aspect of the present invention, there is provided an identification card digital portrait restoration apparatus, including:
the acquisition module is used for acquiring digital portrait photos;
the training database establishing module is used for establishing a digital portrait training database according to the digital portrait photos;
the residual error convolution neural network establishing module is used for establishing a depth residual error convolution neural network according to the digital portrait photo;
the model acquisition module is used for carrying out depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
and the restoration module is used for inputting the input value of the card digital portrait to be restored into the deep residual error network restoration model to obtain the restored card digital portrait.
As another aspect of the present invention, a storage medium is provided, wherein the storage medium stores a computer program, wherein the computer program is configured to execute the identification card digital portrait restoration method as described above when running.
As another aspect of the present invention, a processor is provided, wherein the processor is configured to run a computer program, wherein the computer program is run to execute the identification card digital portrait restoration method as described above.
According to the certificate card digital portrait restoration method, firstly, a digital portrait photo is collected, then a portrait training database is established for the digital portrait photo, a depth residual error convolutional neural network is established, depth residual error network learning is carried out according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model, and finally a certificate card portrait needing restoration is restored according to the restoration model.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of the method for recovering the digital portrait of the identification card provided by the invention.
Fig. 2 is a flowchart of an embodiment of the method for recovering a digital portrait of an identification card according to the present invention.
FIG. 3 is a schematic diagram of a human image degradation model provided by the present invention.
Fig. 4 is a schematic diagram of a depth residual error network structure for restoring a human image according to the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged under appropriate circumstances in order to facilitate the description of the embodiments of the invention herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In this embodiment, an identification card digital portrait restoration method is provided, and fig. 1 is a flowchart of an identification card digital portrait restoration method according to an embodiment of the present invention, as shown in fig. 1, including:
s110, collecting digital portrait photos;
s120, establishing a digital portrait training database according to the digital portrait photos;
s130, establishing a depth residual error convolution neural network according to the digital portrait photo;
s140, performing depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
s150, inputting the input value of the card digital portrait to be restored and the deep residual error network restoration model to obtain the restored card digital portrait.
According to the certificate card digital portrait restoration method, firstly, a digital portrait photo is collected, then a portrait training database is established for the digital portrait photo, a depth residual error convolutional neural network is established, depth residual error network learning is carried out according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model, and finally a certificate card portrait needing restoration is restored according to the restoration model.
The following describes in detail a specific implementation process of the identification card digital portrait restoration method according to the embodiment of the present invention with reference to fig. 2.
Specifically, the establishing of the digital portrait training database according to the digital portrait photos includes:
carrying out portrait positioning on the digital portrait photo to obtain a standardized portrait photo;
and performing wavelet transformation and decompression processing on the normalized portrait photo to obtain a digital portrait training database.
It can be understood that the certificate portrait standardized picture data are collected by using public platforms such as the internet, the mobile phone app and the like, and collected portrait pictures should preliminarily meet the driving certificate portrait shooting standard specification, namely, basic conditions of single background, moderate size proportion, reasonable head portrait position and the like.
The normalized portrait is subjected to wavelet transformation, then quantization processing is carried out, zero-tree coding compression is carried out by high compression ratio, and the compressed data is added with a digital signature and stored in a chip for certificate card anti-counterfeiting. Meanwhile, the compressed picture is decoded, and the decoded picture and the original picture form a data point pair together and are used as training data for training the late deep learning model.
Further specifically, the establishing a digital portrait training database according to the digital portrait photos further includes:
and carrying out mirror image transformation on the images in the digital portrait training database.
It will be appreciated that by making mirror changes to the images in the training dataset, the dataset size can be increased and the risk of over-fitting mitigated.
Preferably, the portrait positioning the digital portrait photo to obtain a normalized portrait photo includes:
carrying out portrait positioning on the digital portrait photo according to a multitask cascade convolution neural network and a skin color model algorithm;
and adjusting the resolution of the digital portrait photo to obtain the normalized portrait photo.
It should be noted that, a multi-task cascade convolution neural network (MTCNN) and a skin color model algorithm may be used to perform portrait positioning on the picture acquired in a1, remove part of the boundary, only retain the final head region, and simultaneously normalize the size and adjust the resolution to about 270 × 420 pixels. The size of the portrait picture is normalized, and good model precision can be obtained for a small portrait data set.
Specifically, the establishing of the depth residual convolutional neural network according to the digital portrait photo includes:
establishing an image restoration model according to the digital portrait photo;
for example, in the case of a low sampling rate, the wavelet-decompressed image has significant blocking and blurring effects with respect to the original image, mainly because the wavelet transform is also a block-based transform coding scheme in nature. In order to recover the original image from the decompressed image as much as possible, it is necessary to efficiently model the degradation process. By taking the image denoising theory as reference, the block blurring effect formed by wavelet compression is regarded as an extra noise interference v, the original image is x, the decompressed image y is a low-quality image mixed with specific noise (as shown in fig. 3), and the relationship of the three satisfies y being x + v.
Carrying out residual error processing on the image restoration model;
for example, the depth residual convolution network model can extract stable and efficient image features from complex image data, and can automatically mine potential image high-level information in an image through the learning of a large number of samples. The convolutional neural network architecture for constructing deep residual learning comprises three basic structures, as shown in fig. 4, specifically: convolution (Conv) + linear rectification function (ReLU), convolution + Batch Normalization) + linear rectification function, convolution; the first structure uses 64 convolution filters of 3 × 3 × c, which can generate 64 feature maps, and then is non-linearized by the ReLU; in the second structure, besides the adoption of 3 multiplied by c convolution filtering, a BN layer is additionally added into the re-convolution layer and the ReLU layer; the last structure uses only 3 × 3 × c convolution filtering for the design of the last layer.
Determining input and output data of the image restoration model to obtain a target function;
for example, in order to achieve efficient restoration using a depth residual convolution network model, the decoded image y is input as a model, and the difference from the original image x is output as a desired model output, that is, the difference between the model and the original image x is output
As a training data point pair; the mean square error of the objective function is:
wherein, theta represents a model parameter expected to be obtained by learning, F represents an F-mode, and R represents mapping.
And performing expected learning according to the target function to obtain the depth residual convolution neural network.
Specifically, the depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model, including:
repeatedly training the depth residual convolution neural network according to the residual paired data set;
counting the peak signal-to-noise ratio of the restored image and the original image after each training, and recording the training times and parameters used when the target peak signal-to-noise ratio is reached;
and checking and network debugging the repeated training result according to the recorded training times and parameters when the target peak signal-to-noise ratio value is reached to obtain a deep residual error network restoration model.
It can be understood that the performance of the deep learning neural network is mastered by counting the peak signal-to-noise ratio (PSNR) values of the restored image and the original image after each training and recording the training times and parameters used when the target PSNR is reached. And (4) checking and debugging the network according to the feedback information, so that a high-efficiency and practical deep residual error network restoration model can be formed.
Specifically, the obtaining of the recovered license card digital portrait by the input value of the license card digital portrait to be recovered and the deep residual error network recovery model comprises:
after the digital signature information in the identification card is verified, extracting image compression data of the identification card digital portrait to be restored;
performing wavelet decompression on the image compressed data to obtain a low-quality image;
performing residual prediction according to the depth residual network restoration model to obtain a prediction residual;
and calculating according to the prediction residual error and the low-quality image to obtain a recovered certificate card digital portrait.
To sum up, the method for restoring a digital portrait of a certificate card according to the embodiments of the present invention includes collecting digital portrait photos on a plurality of public platforms, removing photos that do not meet the basic standards of portrait shooting, performing normalization processing on the photos, performing wavelet compression and decompression to obtain a training data set, performing residual error processing on the images to obtain standard input and output data of a model, training a depth residual error network, and continuously repeating the training and modifying the network, so that the method can be used for restoring the decompressed images of the certificate card.
Different from the existing image restoration technology, the depth residual convolution network adopted by the embodiment of the invention can autonomously learn the image degradation priori knowledge by updating the convolution kernel, so that a default specific image prior method adopted in the traditional digital image restoration method is not needed, the restoration of the image of the card is more targeted, and the restoration effect can be obviously improved.
In order to further improve the compression efficiency, the original portrait needs to be cut to a certain degree, and the portrait area is removed as much as possible. The multi-task cascade convolution neural network (MTCNN) and the skin color model algorithm adopted by the embodiment of the invention can automatically judge the face boundary and cut out the portrait photos meeting the requirements, thereby greatly reducing the manual workload.
In addition, in the embodiment of the invention, the parallel computation of the CPU and the GPU can be carried out in the model training process of the depth residual error network, so that the computation efficiency is improved, and the computation resources are fully utilized.
Finally, the embodiment of the invention has wider picture collection and larger collected data volume, and can adapt to more identification photo environments, so the recovery effect has better generalization performance.
The method for restoring the digital portrait of the certificate card provided by the embodiment of the invention can restore the blocking effect formed by wavelet compression, can adapt to other compression algorithms, only needs to replace training data sets formed by other compression algorithms, and has good universality.
As another embodiment of the present invention, an identification card digital portrait restoration device is provided, which is used to implement the above embodiments, and the description of the above embodiments is omitted. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the system or apparatus described in the embodiments below is implemented in software, implementations in hardware, or a combination of software and hardware are also possible and contemplated.
Specifically, the identification card digital portrait restoration device comprises:
the acquisition module is used for acquiring digital portrait photos;
the training database establishing module is used for establishing a digital portrait training database according to the digital portrait photos;
the residual error convolution neural network establishing module is used for establishing a depth residual error convolution neural network according to the digital portrait photo;
the model acquisition module is used for carrying out depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
and the restoration module is used for inputting the input value of the card digital portrait to be restored into the deep residual error network restoration model to obtain the restored card digital portrait.
According to the certificate card digital portrait restoration device, a digital portrait photo is collected firstly, a portrait training database is established for the digital portrait photo, a depth residual error convolutional neural network is established, depth residual error network learning is carried out according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model, and finally a certificate card portrait needing restoration is restored according to the restoration model.
An embodiment of the present invention further provides a storage medium, wherein the storage medium stores therein a computer program, and the computer program is configured to execute the identification card digital portrait restoration method as described above when running.
Embodiments of the present invention further provide a processor, wherein the processor is configured to run a computer program, and the computer program is executed to execute the identification card digital portrait restoration method as described above.
It will be understood that the above embodiments are merely exemplary embodiments taken to illustrate the principles of the present invention, which is not limited thereto. It will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit and substance of the invention, and these modifications and improvements are also considered to be within the scope of the invention.
Claims (10)
1. A method for recovering a digital portrait of an identification card is characterized by comprising the following steps:
collecting digital portrait photos;
establishing a digital portrait training database according to the digital portrait photos;
establishing a depth residual error convolution neural network according to the digital portrait photo;
performing depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
and inputting the card digital portrait to be restored to the deep residual error network restoration model to obtain the restored card digital portrait.
2. The method for recovering the digital portrait of the identification card according to claim 1, wherein the establishing of the digital portrait training database according to the digital portrait photo comprises:
carrying out portrait positioning on the digital portrait photo to obtain a standardized portrait photo;
and performing wavelet transformation and decompression processing on the normalized portrait photo to obtain a digital portrait training database.
3. The identification card digital portrait restoration method according to claim 2, wherein the establishing of a digital portrait training database according to the digital portrait photos further comprises:
and carrying out mirror image transformation on the images in the digital portrait training database.
4. The identification card digital portrait restoration method according to claim 2, wherein the portrait positioning of the digital portrait photo to obtain a normalized portrait photo comprises:
carrying out portrait positioning on the digital portrait photo according to a multitask cascade convolution neural network and a skin color model algorithm;
and adjusting the resolution of the digital portrait photo to obtain the normalized portrait photo.
5. The card digital portrait restoration method according to claim 1, wherein the establishing of the depth residual convolutional neural network according to the digital portrait photo comprises:
establishing an image restoration model according to the digital portrait photo;
carrying out residual error processing on the image restoration model;
determining input and output data of the image restoration model to obtain a target function;
and performing expected learning according to the target function to obtain the depth residual convolution neural network.
6. The method for recovering the digital portrait of the certificate card according to claim 1, wherein the obtaining of the recovery model of the depth residual error network by performing the deep residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network comprises:
repeatedly training the depth residual convolution neural network according to the residual paired data set;
counting the peak signal-to-noise ratio of the restored image and the original image after each training, and recording the training times and parameters used when the target peak signal-to-noise ratio is reached;
and checking and network debugging the repeated training result according to the recorded training times and parameters when the target peak signal-to-noise ratio value is reached to obtain a deep residual error network restoration model.
7. The method for recovering the card digital portrait according to claim 1, wherein the step of inputting the card digital portrait to be recovered to the deep residual error network recovery model to obtain the recovered card digital portrait comprises the following steps:
after the digital signature information in the identification card is verified, extracting image compression data of the identification card digital portrait to be restored;
performing wavelet decompression on the image compressed data to obtain a low-quality image;
performing residual prediction according to the depth residual network restoration model to obtain a prediction residual;
and calculating according to the prediction residual error and the low-quality image to obtain a recovered certificate card digital portrait.
8. An identification card digital portrait restoration device, comprising:
the acquisition module is used for acquiring digital portrait photos;
the training database establishing module is used for establishing a digital portrait training database according to the digital portrait photos;
the residual error convolution neural network establishing module is used for establishing a depth residual error convolution neural network according to the digital portrait photo;
the model acquisition module is used for carrying out depth residual error network learning according to the digital portrait training database and the depth residual error convolutional neural network to obtain a depth residual error network restoration model;
and the restoration module is used for inputting the input value of the card digital portrait to be restored into the deep residual error network restoration model to obtain the restored card digital portrait.
9. A storage medium having stored thereon a computer program, wherein the computer program is configured to execute the method for recovering a digital image of an identification card according to any one of claims 1 to 7 when the computer program is executed.
10. A processor adapted to run a computer program, wherein the computer program is operable to perform the method of recovering a digital image of an identification card of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030852.4A CN110766635A (en) | 2019-10-28 | 2019-10-28 | Identification card digital portrait restoration method and device, storage medium and processor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030852.4A CN110766635A (en) | 2019-10-28 | 2019-10-28 | Identification card digital portrait restoration method and device, storage medium and processor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110766635A true CN110766635A (en) | 2020-02-07 |
Family
ID=69334225
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911030852.4A Withdrawn CN110766635A (en) | 2019-10-28 | 2019-10-28 | Identification card digital portrait restoration method and device, storage medium and processor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110766635A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111970513A (en) * | 2020-08-14 | 2020-11-20 | 成都数字天空科技有限公司 | Image processing method and device, electronic equipment and storage medium |
-
2019
- 2019-10-28 CN CN201911030852.4A patent/CN110766635A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111970513A (en) * | 2020-08-14 | 2020-11-20 | 成都数字天空科技有限公司 | Image processing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fu et al. | Jpeg artifacts reduction via deep convolutional sparse coding | |
| Liu et al. | Multi-level wavelet-CNN for image restoration | |
| Li et al. | An efficient deep convolutional neural networks model for compressed image deblocking | |
| Dong et al. | Compression artifacts reduction by a deep convolutional network | |
| Chen et al. | DPW-SDNet: Dual pixel-wavelet domain deep CNNs for soft decoding of JPEG-compressed images | |
| Ono et al. | Cartoon-texture image decomposition using blockwise low-rank texture characterization | |
| CN102088606B (en) | Sparse representation-based deblocking method | |
| Zhao et al. | Invertible image decolorization | |
| Li et al. | An iterative framework of cascaded deblocking and superresolution for compressed images | |
| Salmon et al. | From patches to pixels in non-local methods: Weighted-average reprojection | |
| Fu et al. | Learning dual priors for jpeg compression artifacts removal | |
| Huang et al. | Two-step approach for the restoration of images corrupted by multiplicative noise | |
| Hang et al. | Attention cube network for image restoration | |
| US9230161B2 (en) | Multiple layer block matching method and system for image denoising | |
| CN103914816A (en) | Video super-resolution method based on non-local regularization | |
| CN110782406A (en) | Image denoising method and device based on information distillation network | |
| CN110503608B (en) | Image denoising method based on multi-view convolutional neural network | |
| CN110766635A (en) | Identification card digital portrait restoration method and device, storage medium and processor | |
| CN102254305A (en) | Image restoring method based on three dictionary block matching | |
| Li et al. | Deep scale-aware image smoothing | |
| Sadreazami et al. | Data-adaptive color image denoising and enhancement using graph-based filtering | |
| Nejati et al. | Low-rank regularized collaborative filtering for image denoising | |
| CN114862711B (en) | Low-illumination image enhancement and denoising method based on dual complementary prior constraints | |
| Zhang | The SURE-LET approach using hybrid thresholding function for image denoising | |
| Luo et al. | Super-resolving compressed images via parallel and series integration of artifact reduction and resolution enhancement |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200207 |