CN114677684A - Distorted image correction method, device and equipment and computer readable storage medium - Google Patents
Distorted image correction method, device and equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN114677684A CN114677684A CN202210293242.9A CN202210293242A CN114677684A CN 114677684 A CN114677684 A CN 114677684A CN 202210293242 A CN202210293242 A CN 202210293242A CN 114677684 A CN114677684 A CN 114677684A
- Authority
- CN
- China
- Prior art keywords
- image
- preset
- initial
- deformation
- neural 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 91
- 238000003702 image correction Methods 0.000 title claims description 40
- 238000003062 neural network model Methods 0.000 claims abstract description 106
- 230000008569 process Effects 0.000 claims description 39
- 239000011159 matrix material Substances 0.000 claims description 34
- 238000012549 training Methods 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 18
- 238000013139 quantization Methods 0.000 claims description 15
- 230000008439 repair process Effects 0.000 claims description 6
- 230000037303 wrinkles Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 abstract description 7
- 238000012545 processing Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 7
- 238000005452 bending Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- 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
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Processing (AREA)
Abstract
The application belongs to the technical field of image processing, and provides a method and a device for correcting a distorted image, computer equipment and a computer readable storage medium. In order to solve the problem that correction is inaccurate when a distorted image is subjected to flattening correction, an acquired initial distorted image is input to a preset first U-shaped neural network model contained in a preset superposed U-shaped neural network model to obtain an image distortion disturbance field, the image distortion disturbance field is input to a preset second U-shaped neural network model contained in the preset superposed U-shaped neural network model to obtain an image restoration disturbance field, the initial distorted image is restored by adopting the image restoration disturbance field to obtain a corrected image, the image restoration disturbance field corresponding to the disturbance field for restoring the distorted image is directly predicted through the superposed U-shaped neural network model, the distorted image is restored by utilizing the image restoration disturbance field to obtain a restored corrected image, and the precision and accuracy of correction of the distorted image can be improved.
Description
Technical Field
The present application relates to the field of image processing technologies, and in particular, to a method and an apparatus for correcting a distorted image, a computer device, and a computer-readable storage medium.
Background
Due to the popularity of mobile cameras, capturing document images is a common method of digitizing and recording physical documents. To make text recognition easier, it is often desirable to digitally flatten the document image as the physical document page is folded or bent. In the conventional technology, a folded or bent image is generally directly flattened to obtain a corrected image, and because the folded or bent original image is directly flattened, the problem of inaccurate corrected image exists.
Disclosure of Invention
The application provides a distorted image correction method, a distorted image correction device, computer equipment and a computer readable storage medium, which can solve the technical problem that correction is inaccurate when a distorted image is subjected to flattening correction in the prior art.
In a first aspect, the present application provides a method for warped image correction, comprising: acquiring an initial distorted image, inputting the initial distorted image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, and obtaining an image distortion disturbing field contained in the initial distorted image, wherein the image distortion disturbing field is used for describing the deformation quantization characteristics of a flattened image in the deformation process of the initial distorted image; inputting the image distortion disturbing field into a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbing field, wherein the image restoration disturbing field is used for describing deformation quantization characteristics in the deformation process of deforming the initial distortion image into a flattened image; and repairing the initial distorted image by adopting the image repairing disturbance field to obtain a corrected image corresponding to the initial distorted image.
In a second aspect, the present application also provides a distorted image correction apparatus, comprising: the image warping method comprises a first input unit, a second input unit and a third input unit, wherein the first input unit is used for acquiring an initial warped image and inputting the initial warped image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model to obtain an image warping disturbance field contained in the initial warped image, and the image warping disturbance field is used for describing a deformation quantization characteristic in a deformation process of a flattened image deformed into the initial warped image; the second input unit is used for inputting the image distortion disturbing field to a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbing field, wherein the image restoration disturbing field is used for describing deformation quantization characteristics in a deformation process of deforming the initial distortion image into a flattened image; and the restoration unit is used for restoring the initial distorted image by adopting the image restoration disturbance field to obtain a corrected image corresponding to the initial distorted image.
In a third aspect, the present application further provides a computer device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the warped image correction method when executing the computer program.
In a fourth aspect, the present application further provides a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the warped image correction method.
The application provides a method and a device for correcting a distorted image, computer equipment and a computer readable storage medium. The method comprises the steps of obtaining an initial distorted image, inputting the initial distorted image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, obtaining an image distortion disturbing field contained in the initial distorted image, inputting the image distortion disturbing field to a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model, obtaining an image restoration disturbing field, adopting the image restoration disturbing field to restore the initial distorted image, obtaining a corrected image corresponding to the initial distorted image, directly predicting the image restoration disturbing field corresponding to the disturbance field for restoring the distorted image through the superposed U-shaped neural network model, restoring the distorted image by utilizing the image restoration disturbing field, obtaining the restored corrected image, and considering the distorted deformation result characteristics such as the strength and the direction of image deformation due to the image restoration disturbing field, and moreover, a distortion mode is taken as a repairing operation, compared with most methods for directly predicting the corrected image, the method and the device for correcting the distorted image restore the image distortion inverse process as real as possible through the image repairing disturbance field, and can improve the accuracy and precision of the distorted image correction.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flowchart of a warped image correction method according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a first sub-flow of a warped image correction method according to an embodiment of the present application;
FIG. 3 is a second sub-flowchart of a warped image correction method according to an embodiment of the present application;
fig. 4 is a schematic diagram of a third sub-flow of a warped image correction method according to an embodiment of the present application;
FIG. 5 is a schematic block diagram of a warped image correction apparatus provided by an embodiment of the present application;
fig. 6 is a schematic block diagram of a computer device provided in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. 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 application.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
Referring to fig. 1, fig. 1 is a schematic flow chart of a distorted image correction method according to an embodiment of the present disclosure. As shown in FIG. 1, the method includes the following steps S11-S13:
s11, obtaining an initial distorted image, inputting the initial distorted image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, and obtaining an image distortion disturbing field contained in the initial distorted image, wherein the image distortion disturbing field is used for describing the deformation quantization characteristics of the flattened image in the deformation process of the initial distorted image.
The U-shaped neural network model is a U-Net neural network model, is a deformation of a convolution neural network model, is mainly structurally drawn to be similar to a letter U, and therefore the name of the U-Net is obtained, and the whole neural network model mainly comprises two parts: a contracting path (contracting path) and an expanding path (expanding path). The superposed U-shaped neural network model is a combined neural network model constructed by superposing at least two U-shaped neural network models.
Specifically, a preset training sample image is prepared in advance, the preset training sample image includes an initial warped training image and a corrected training image corresponding to the initial warped training image (i.e., a flattened training image corresponding to the initial warped training image), the initial warped training image may include shapes formed by various warped deformations such as folding (i.e., folding) or bending (i.e., curving), a preset first U-shaped neural network model included in a preset superimposed U-shaped neural network model is trained in advance using the preset training sample image, the trained preset first U-shaped neural network model may predict an image warping process of a warped image, that is, a warping process that a predicted image may take when the flattened image is changed into the warped image, the warping process is described by using an image warping perturbation field for describing a deformation amount of the flattened image in the deformation process of the initial warped image The warping process including warping results such as warping direction and warping strength, and the warping mode corresponding to the warping is a shape change mode such as wrinkle or bending, and the image warping disturbance field is a vector field (i.e. vector field) including the direction, strength and change mode of the image shape change. For example, different twisted images can be formed by folding or bending a sheet of paper in different directions, and all the twisted images correspond to one piece of flat paper, but the direction, strength and manner of twisting may be different, or the direction and strength of twisting may be the same, but the manner of twisting may be different.
After the training of a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model is completed, an initial distorted image can be obtained, the initial distorted image is input into the preset first U-shaped neural network model, and therefore the distortion processes such as the distortion direction, the distortion intensity and the distortion mode corresponding to the initial distorted image are predicted through the preset first U-shaped neural network model, namely the distortion process is quantitatively described by adopting the distortion characteristics. For the same initial distorted image, predicting and obtaining an image distortion process corresponding to the initial distorted image according to a deformed shape of the initial distorted image, thereby obtaining an image distortion disturbance field contained in the initial distorted image, wherein the image distortion disturbance field contains a distortion direction, a distortion strength and a distortion mode of each image distortion, for example, a piece of paper is distorted into a shape with an included angle of 30 degrees, and the paper can be folded from top to bottom or folded from bottom to top, and the folding angle and the folding direction are different, so that different distortion processes may exist for the same initial distorted image, thereby corresponding to the image distortion disturbance field.
Further, the image distortion disturbance field comprises an image distortion vector field and an image deformation mode characteristic. The image distortion vector field comprises the direction and the strength of image distortion, and can be understood as parameters of distortion action, and the image deformation mode characteristics are used for describing the characteristics of different image deformation modes such as folds and curves of an image, and can be understood as distortion action.
Specifically, since the image distortion disturbance field includes the direction, intensity and manner of each image distortion, the direction, intensity and manner of distortion can be described in a quantitative manner, respectively, so as to obtain the image distortion disturbance field including the image distortion vector field and the image deformation manner feature, thereby predicting the distortion process of the distorted image, i.e. the formation process of the initial distorted image, as perfectly and accurately as possible through the image distortion vector field and the image deformation manner feature included in the image distortion disturbance field.
And S12, inputting the image distortion disturbance field to a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbance field, wherein the image restoration disturbance field is used for describing deformation quantification characteristics in a deformation process of deforming the initial distortion image into a flattened image.
Specifically, the model frame structure (i.e. the main component structure) of the preset second U-shaped neural network model included in the preset superimposed U-shaped neural network model is the same as the model frame structure (i.e. the main component structure) of the preset first U-shaped neural network model included in the preset superimposed U-shaped neural network model, but the weights and assignments of the respective parameters are different, and the uses (i.e. the respective functions) of the preset first U-shaped neural network model and the preset first U-shaped neural network model are different, so that the preset second U-shaped neural network model can be trained in a manner similar to the training of the preset first U-shaped neural network model, so that the preset second U-shaped neural network model can be learned according to the corresponding relationship between the training sample image distortion perturbation field of the preset training sample image and the corresponding training sample image restoration perturbation field, the method comprises the steps of learning the corresponding relation between a training sample image distortion disturbance field and a corresponding training sample image restoration disturbance field, and then predicting the image restoration disturbance field corresponding to the image distortion disturbance field according to a predicted image distortion disturbance field, wherein the image restoration disturbance field is used for describing deformation quantization characteristics in a deformation process of deforming an initial distorted image into a flattened image, and the image distortion disturbance field and the image restoration disturbance field are in a mutual inverse relation. And inputting the image distortion disturbance field into the preset second U-shaped neural network model, so as to obtain an image restoration disturbance field. If the image distortion disturbing field is an image distortion vector field y1, and the image deformation mode is characterized by c, the image restoration disturbing field may be y2 ═ f (y1 ×, c), which is used to describe how y1 and c cooperate with each other to perform flattening on the distorted image, so as to obtain a flattened image, i.e. a corrected image, which is an inverse process of warping the flattened image to obtain a warped image, and meanwhile, the image distortion disturbing field is an inverse process of y 2. It should be noted that, in the embodiment of the present application, the "first" in the preset first U-shaped neural network model and the "second" in the preset second U-shaped neural network model are not used to define the U-shaped neural network model, but are used to distinguish two different U-shaped neural network models included in the preset superimposed U-shaped neural network model.
Further, the loss functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein L iseIs the difference in loss, yiIs the content of the output, and,
is yiIn the previously obtained or preset target reference content, i and j are different pixel points, and n is the number of the pixels;
the drift deformation functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are as follows:
wherein L issFor the magnitude or amplitude of the image deformation, the meaning of the other parameter refers to the meaning of the corresponding parameter in equation (1) of the loss function.
Compared with the method that the corrected image is directly generated in one step, the image restoration disturbance field is firstly obtained, the initial distorted image is restored by utilizing the image restoration disturbance field subsequently, the restored corrected image is obtained, the problem that local detail is easy to cause local blurring and distortion due to loss of features when the corrected image is directly generated can be avoided, and particularly, the problem that the subsequent character recognition is difficult due to difficulty in controlling the local character font under the condition that the application scene is the character image can be avoided.
And S13, repairing the initial distorted image by adopting the image repairing disturbance field to obtain a corrected image corresponding to the initial distorted image.
Specifically, since the image restoration is the reverse process of the image distortion, after the image restoration disturbance field is obtained according to the image distortion disturbance field, the image restoration disturbance field can be used to restore the initial distorted image, and the image of the initial distorted image before distortion deformation, that is, the corrected image, can be obtained.
Further, referring to fig. 2, fig. 2 is a schematic view of a first sub-flow of a warped image correction method according to an embodiment of the present application, as shown in fig. 2, in this embodiment, the restoring the initial warped image by using the image restoring perturbed field to obtain a corrected image corresponding to the initial warped image includes:
s131, acquiring an initial distortion matrix corresponding to the initial distortion image and a repairing disturbance matrix corresponding to the image repairing disturbance place;
s132, calculating the initial distortion matrix and the repair disturbance matrix to obtain a calculation matrix;
and S133, converting the operation matrix into a corresponding image to obtain a corrected image corresponding to the initial distorted image.
Specifically, the initial warped image may be described by using a corresponding initial warped matrix, and the image restoration disturbance field may also be described by using a corresponding restoration disturbance matrix, and the initial warped image is restored by using the image restoration disturbance field, that is, the initial warped matrix and the restoration disturbance matrix are operated. Therefore, an image restoration disturbance field is obtained, and then an initial distorted image is obtained, wherein both the image restoration disturbance field and the initial distorted image can be in a matrix form, and then the initial distorted image is restored by using the image restoration disturbance field, that is, a product operation is performed on an image restoration disturbance matrix corresponding to the image restoration disturbance field and an initial distorted image matrix corresponding to the initial distorted image, so as to obtain an operated matrix, that is, a correction matrix of a corrected image corresponding to the initial distorted image, and the corrected image can be obtained according to the correction matrix.
In the embodiment of the application, an initial distorted image is obtained, the initial distorted image is input into a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, an image distortion disturbing field contained in the initial distorted image is obtained, the image distortion disturbing field is input into a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model, an image restoration disturbing field is obtained, the initial distorted image is restored by adopting the image restoration disturbing field, a corrected image corresponding to the initial distorted image is obtained, the image restoration disturbing field corresponding to the disturbed field restored to the distorted image is directly predicted by the superposed U-shaped neural network model, and the distorted image is restored by utilizing the image restoration disturbing field, so that the restored corrected image is obtained, compared with most methods for directly predicting corrected images, the image restoration disturbance field restores the inverse process of image distortion as real as possible to correct the distorted images, and accuracy and precision of distorted image correction can be improved.
In an embodiment, please refer to fig. 3, fig. 3 is a second sub-flowchart of the method for correcting a warped image according to the embodiment of the present application, as shown in fig. 3, in which before inputting the initial warped image to a preset first U-shaped neural network model included in a preset superimposed U-shaped neural network model, the method further includes:
s14, acquiring a preset plane image, and generating a disturbance grid on the preset plane image;
and S15, with the disturbance grid as a distortion control point, distorting the preset plane image to obtain a distorted image, and taking the distorted image as a preset training sample image used by the preset superposition U-shaped neural network model.
In particular, since the document image deformation can be expressed in the forms of three-dimensional meshes, surface normals, two-dimensional streams and the like, in the actual business, it is difficult to accurately capture various forms of document image deformation in any form, and it is almost impossible to manually fold or distort various deformations to cover all documents of real deformation situations in the actual business, the embodiment of the present application trains the preset superposition U-shaped neural network model by using synthetic distorted image data.
Since the purpose of image correction is to map the distorted image into the corrected flattened image, the synthesized distorted image data is the inverse process of flattening the distorted image, that is, the flattened image is deformed to obtain the distorted image, and the flattened image can be distorted in different shapes, so as to obtain different synthesized distorted images.
Acquiring a preset plane image, wherein the preset plane image can be a corrected image obtained by flattening and correcting a distorted image, and can also be a plane digital image of newspapers, books, magazine pages and the like, because the deformation of the image can be understood as the propagation of pixel points on the image in space, and the basic deformation of the image is the generation of creases and curls, a disturbance grid is generated on the preset plane image in advance, an M × n grid M can be generated on the preset plane image to provide control points for distorting the preset plane image, then the disturbance grid is used as a distortion control point to distort the preset plane image in a set direction, strength, distance and path, so that a synthesized distorted image is obtained on the basis that the disturbance grid provides a sparse deformation field, and the distorted image is the generated disturbance image, and taking the distorted image as a preset training sample image used by the preset superposed U-shaped neural network model to train a preset first U-shaped neural network model and a preset second U-shaped neural network model contained in the preset superposed U-shaped neural network model.
Further, referring to fig. 4, fig. 4 is a schematic view of a third sub-flow of a distorted image correction method according to an embodiment of the present application, as shown in fig. 4, in this embodiment, the distorting the preset plane image by using the disturbance grid as a distortion control point to obtain a distorted image includes:
s151, acquiring an initial deformation vertex, and determining the direction and the strength of deformation of the initial deformation vertex;
s152, according to the direction and the strength, transmitting the initial deformation vertex to vertexes at other positions by adopting a preset formula to obtain a distorted image, wherein the preset formula is as follows:
pT＝pI+wv；
wherein p isTTo distort the coordinates of the vertex T, pIThe coordinates of the initial deformation vertex P of the preset plane image I are set, w is the deformation strength of the deformation corresponding to the distortion, and v is the deformation direction of the deformation corresponding to the distortion.
Specifically, for a perturbation grid M on a preset planar image I, an initial deformation vertex p may be randomly obtained on M by using the perturbation grid M as a distortion control point, and a direction v and an intensity w of deformation are determined, and the direction v may also be randomly generated, and then a dense texture is constructed at a pixel level according to the deformation direction v and the deformation intensity w (i.e., a deformation weight of a quantization feature)And performing linear interpolation on the disturbance grid M to apply the distortion map to a preset plane image I, so that the initial deformation vertex P is propagated to distortion vertices P corresponding to other position vertices on the disturbance grid M based on the disturbance grid MTFrom the initial deformed vertex pITwisted vertex pTAnd the distortion intensity can be determined, so that the distorted disturbance image can be generated, wherein the linear interpolation refers to an interpolation mode of which the interpolation function is a first-order polynomial, the interpolation error of the linear interpolation on an interpolation node is zero, and the coordinates of the vertexes at other positions are as follows:
pT＝pI+ wv equation (3);
wherein p isTTo distort the coordinates of the vertex T, pIAnd presetting the coordinates of the initial deformation vertex P on the plane image I, thereby distorting the plane image to obtain a synthesized distorted image.
Further, it is important to define w, if p and v define a straight line, if w and v need to be predetermined to warp the preset planar image, and v can be randomly determined, so as to warp the vertex pTIt can also be determined that the straight lines defined by p and v are the initial deformation vertex pI and the distortion vertex pTStraight line therebetween, vertex p due to initial deformationIAnd a warped vertex pTThe maximum value of the straight-line distance between the two points can be the initial deformation vertex pIDistance from preset plane image edge, i.e. initial deformation vertex pIAnd a warped vertex pTThe distance of the straight-line distance therebetween ranges from 0 to the maximum value, and therefore, the initial deformation vertex p can be setIAnd a warped vertex pTThe linear distance between the p and v definition lines is normalized by using the maximum value as a reference and mapping the linear distance between the p and v definition lines to 0,1]Or [ -1,1 [)]Within the interval, a normalized description of the linear distance between p and v defining a straight line is obtained, so that first the normalized distance d between p and v defining a straight line is calculated, and w is defined as a function of d, and for the distortion (i.e. distortion) of the wrinkle type, w is defined as follows:
for curve-type distortion (i.e., distortion), w is defined as follows:
w＝1-dαformula (5);
in the formula (4) and the formula (5), α controls a propagation range of the initial deformation vertex for deformation, α is determined according to the size of a preset plane image, the preset plane image is large and can be correspondingly large, the preset plane image is small and can be correspondingly small, α cannot exceed the maximum range of the preset plane image, and can be randomly generated within a certain range and can be used as an α random generation range according to an empirical value. A larger a leads w to 1, which means that all other vertices share the same deformation as p, making the deformation more global, while a smaller a limits the deformation to a local area around p.
Through the distorted disturbance images generated, 100K images can be synthesized on a single CPU, each image contains up to 19 synthesized distortions (30% is bending distortion, and 70% is folding), wherein when the folding is arbitrary, the curve needs to keep Gaussian curvature to be 0 everywhere, so that various distorted images and flattened normal preset plane images matched with the distorted images can be constructed. Compared with other interpolation modes, such as parabolic interpolation, the linear interpolation has the characteristics of simplicity and convenience.
It should be noted that, the distorted image correction method described in each of the above embodiments may recombine the technical features included in different embodiments as needed to obtain a combined implementation, but all of them are within the protection scope claimed in the present application.
Referring to fig. 5, fig. 5 is a schematic block diagram of a distorted image correction apparatus according to an embodiment of the present disclosure. Corresponding to the above distorted image correction method, the embodiment of the application also provides a distorted image correction device. As shown in fig. 5, the warped image correction apparatus includes a unit for performing the above-described warped image correction method, and the warped image correction apparatus may be configured in a computer device. Specifically, referring to fig. 5, the warped image correction device 50 includes a first input unit 51, a second input unit 52 and a repairing unit 53.
The first input unit 51 is configured to acquire an initial warped image, and input the initial warped image to a preset first U-shaped neural network model included in a preset superposition U-shaped neural network model to obtain an image warping disturbance field included in the initial warped image, where the image warping disturbance field is used to describe a deformation quantization characteristic in a deformation process of a flattened image into the initial warped image;
a second input unit 52, configured to input the image distortion perturbation field to a preset second U-shaped neural network model included in the preset superimposed U-shaped neural network model to obtain an image restoration perturbation field, where the image restoration perturbation field is used to describe a deformation quantization feature in a deformation process of deforming the initial distorted image into a flattened image;
and the repairing unit 53 is configured to repair the initial distorted image by using the image repairing disturbance field to obtain a corrected image corresponding to the initial distorted image.
In one embodiment, the image distortion perturbation field comprises an image distortion vector field and an image deformation mode feature.
In an embodiment, the loss functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein the content of the first and second substances,Leis the difference in loss, yiIs the content of the output, and,
is yiIn the previously obtained or preset target reference content, i and j are different pixel points, and n is the number of the pixels;
the drift deformation functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein L issThe meaning of the other parameter is referred to the meaning of the corresponding parameter in the formula of the loss function, which is the magnitude or amplitude of the image deformation.
In one embodiment, the repair unit 53 includes:
the first obtaining subunit is used for obtaining an initial warping matrix corresponding to the initial warped image and a repairing disturbance matrix corresponding to the image repairing disturbance place;
the operation subunit is used for operating the initial distortion matrix and the repair disturbance matrix to obtain an operation matrix;
and the conversion subunit is used for converting the operation matrix into a corresponding image to obtain a corrected image corresponding to the initial distorted image.
In an embodiment, the warped image correction device 50 further comprises:
the device comprises an acquisition unit, a processing unit and a display unit, wherein the acquisition unit is used for acquiring a preset plane image and generating a disturbance grid on the preset plane image;
and the distortion unit is used for distorting the preset plane image by taking the disturbance grid as a distortion control point to obtain a distorted image, and taking the distorted image as a preset training sample image used by the preset superposition U-shaped neural network model.
In one embodiment, the twisting unit includes:
the acquisition subunit is used for acquiring an initial deformation vertex and determining the direction and the strength of deformation of the initial deformation vertex;
a propagation subunit, configured to propagate, according to the direction and the intensity, the initial deformation vertex to distortion vertices corresponding to vertices at other positions by using a preset formula, so as to obtain a distortion image, where the preset formula is as follows:
pT＝pI+wv；
wherein p isTTo distort the coordinates of the vertex T, pIThe coordinates of the initial deformation vertex P of the preset plane image I are set, w is the deformation strength of the deformation corresponding to the distortion, and v is the deformation direction of the deformation corresponding to the distortion.
In an embodiment, the propagation subunit, in particular for a twist of the wrinkle type, defines w as follows:
and/or for curve type warping, define w as follows:
w＝1-dα；
and d is the normalized distance of a straight line determined based on p and v in the warping process of warping the preset plane image, and alpha is used for controlling the propagation range of the initial deformation vertex for deformation.
It should be noted that, as can be clearly understood by those skilled in the art, the detailed implementation process of the above-mentioned distorted image correction apparatus and each unit may refer to the corresponding description in the foregoing method embodiments, and for convenience and brevity of description, no further description is provided herein.
Meanwhile, the division and connection manner of each unit in the above-mentioned distorted image correction apparatus are only used for illustration, in other embodiments, the distorted image correction apparatus may be divided into different units as required, or each unit in the distorted image correction apparatus may adopt different connection order and manner to complete all or part of the functions of the above-mentioned distorted image correction apparatus.
The above-described distorted image correction apparatus may be implemented in the form of a computer program that can be run on a computer device as shown in fig. 6.
Referring to fig. 6, fig. 6 is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device 500 may be a computer device such as a desktop computer or a server, or may be a component or part of another device.
Referring to fig. 6, the computer device 500 includes a processor 502, a memory, which may include a non-volatile storage medium 503 and an internal memory 504, which may also be a volatile storage medium, and a network interface 505 connected by a system bus 501.
The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032, when executed, causes the processor 502 to perform one of the above-described warped image correction methods.
The processor 502 is used to provide computing and control capabilities to support the operation of the overall computer device 500.
The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503, and when the computer program 5032 is executed by the processor 502, the processor 502 may be enabled to perform a method for correcting a distorted image as described above.
The network interface 505 is used for network communication with other devices. Those skilled in the art will appreciate that the configuration shown in fig. 6 is a block diagram of only a portion of the configuration associated with the present application and does not constitute a limitation of the computer device 500 to which the present application may be applied, and that a particular computer device 500 may include more or less components than those shown, or may combine certain components, or have a different arrangement of components. For example, in some embodiments, the computer device may only include a memory and a processor, and in such embodiments, the structures and functions of the memory and the processor are consistent with those of the embodiment shown in fig. 6, and are not described herein again.
Wherein the processor 502 is configured to run the computer program 5032 stored in the memory to implement the following steps: acquiring an initial distorted image, inputting the initial distorted image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, and obtaining an image distortion disturbing field contained in the initial distorted image, wherein the image distortion disturbing field is used for describing the deformation quantization characteristics of a flattened image in the deformation process of the initial distorted image; inputting the image distortion disturbing field into a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbing field, wherein the image restoration disturbing field is used for describing deformation quantization characteristics in a deformation process of deforming the initial distortion image into a flattened image; and repairing the initial distorted image by adopting the image repairing disturbance field to obtain a corrected image corresponding to the initial distorted image.
In an embodiment, when the processor 502 implements the inputting of the initial warped image into the preset first U-shaped neural network model included in the preset superposition U-shaped neural network model to obtain the image warped perturbation field included in the initial warped image, the image warped perturbation field includes an image warped vector field and an image warped mode feature.
In an embodiment, when the processor 502 implements the preset superimposed U-shaped neural network model, the loss functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein L iseIs the difference in loss, yiIs the content of the output, and,
is yiI and j are different pixel points, and n is the number of pixels;
the drift deformation functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein L issFor the magnitude or amplitude of the image deformation, the meaning of the other parameter refers to the meaning of the corresponding parameter in the formula of the loss function.
In an embodiment, when the processor 502 implements the repairing of the initial warped image by using the image repairing disturbing field to obtain a corrected image corresponding to the initial warped image, the following steps are specifically implemented:
acquiring an initial warping matrix corresponding to the initial warping image and a repairing disturbing matrix corresponding to the image repairing disturbing place;
calculating the initial distortion matrix and the repair disturbance matrix to obtain a calculation matrix;
and converting the operation matrix into a corresponding image to obtain a corrected image corresponding to the initial distorted image.
In an embodiment, before implementing the inputting of the initial warped image to the preset first U-shaped neural network model included in the preset superimposed U-shaped neural network model, the processor 502 further implements the following steps:
acquiring a preset plane image, and generating a disturbance grid on the preset plane image;
and with the disturbance grid as a distortion control point, distorting the preset plane image to obtain a distorted image, and taking the distorted image as a preset training sample image used by the preset superposition U-shaped neural network model.
In an embodiment, when the processor 502 implements that the perturbation grid is used as a warping control point, and the preset planar image is warped to obtain a warped image, the following steps are specifically implemented:
acquiring an initial deformation vertex, and determining the direction and the strength of the deformation of the initial deformation vertex;
according to the direction and the strength, a preset formula is adopted to transmit the initial deformation vertex to a distortion vertex corresponding to a vertex at other positions to obtain a distortion image, wherein the preset formula is as follows:
pT＝pI+wv；
wherein p isTTo distort the coordinates of the vertex T, pIThe coordinates of the initial deformation vertex P of the preset plane image I are set, w is the deformation strength of the deformation corresponding to the distortion, and v is the deformation direction of the deformation corresponding to the distortion.
In one embodiment, the processor 502 defines w as follows for the distortion of the wrinkle type when implementing the preset formula:
and/or for curve type warping, define w as follows:
w＝1-dα；
and d is the normalized distance of a straight line determined based on p and v in the warping process of warping the preset plane image, and alpha is used for controlling the propagation range of the initial deformation vertex for deformation.
It should be understood that in the embodiment of the present Application, the Processor 502 may be a Central Processing Unit (CPU), and the Processor 502 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
It will be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiments may be implemented by a computer program, and the computer program may be stored in a computer readable storage medium. The computer program is executed by at least one processor in the computer system to implement the flow steps of the embodiments of the method described above.
Accordingly, the present application also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, the computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of:
a computer program product which, when run on a computer, causes the computer to perform the steps of the warped image correction method described in the embodiments above.
The computer readable storage medium may be an internal storage unit of the aforementioned device, such as a hard disk or a memory of the device. The computer readable storage medium may also be an external storage device of the apparatus, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), etc. provided on the apparatus. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of the apparatus.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses, devices and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The storage medium is an entity and non-transitory storage medium, and may be various entity storage media capable of storing computer programs, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative. For example, the division of each unit is only one logic function division, and there may be another division manner in actual implementation. For example, various elements or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented.
The steps in the method of the embodiment of the application can be sequentially adjusted, combined and deleted according to actual needs. The units in the device of the embodiment of the application can be combined, divided and deleted according to actual needs. In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in a storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing an electronic device (which may be a personal computer, a terminal, or a network device) to perform all or part of the steps of the method according to the embodiments of the present application.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the present application, and these modifications or substitutions should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A method of warped image correction, the method comprising:
acquiring an initial distorted image, inputting the initial distorted image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, and obtaining an image distortion disturbing field contained in the initial distorted image, wherein the image distortion disturbing field is used for describing the deformation quantization characteristics of a flattened image in the deformation process of the initial distorted image;
inputting the image distortion disturbing field into a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbing field, wherein the image restoration disturbing field is used for describing deformation quantization characteristics in a deformation process of deforming the initial distortion image into a flattened image;
and repairing the initial distorted image by adopting the image repairing disturbance field to obtain a corrected image corresponding to the initial distorted image.
2. A warped image correction method according to claim 1, wherein the image warping disturbance field comprises an image warping vector field and an image warping mode characteristic.
3. The method according to claim 1 or 2, wherein the loss functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are:
wherein L iseIs the difference in loss, yiIs the content of the output, and,
is yiIn the previously obtained or preset target reference content, i and j are different pixel points, and n is the number of the pixels;
the drift deformation functions adopted by the preset first U-shaped neural network model and the preset second U-shaped neural network model are both:
wherein L issFor the magnitude or amplitude of the image deformation, the meaning of the other parameter refers to the meaning of the corresponding parameter in the formula of the loss function.
4. The method for correcting a warped image according to claim 1, wherein the restoring the initial warped image using the image restoration perturbation field to obtain a corrected image corresponding to the initial warped image comprises:
acquiring an initial distortion matrix corresponding to the initial distortion image and a repairing disturbance matrix corresponding to the image repairing disturbance place;
calculating the initial distortion matrix and the repair disturbance matrix to obtain a calculation matrix;
and converting the operation matrix into a corresponding image to obtain a corrected image corresponding to the initial distorted image.
5. The method for correcting a warped image according to claim 1, wherein before inputting the initial warped image into a preset first U-shaped neural network model included in a preset U-shaped neural network model, the method further comprises:
acquiring a preset plane image, and generating a disturbance grid on the preset plane image;
and with the disturbance grid as a distortion control point, distorting the preset plane image to obtain a distorted image, and taking the distorted image as a preset training sample image used by the preset superposition U-shaped neural network model.
6. The method according to claim 5, wherein the warping the preset planar image with the perturbation grid as a warping control point to obtain a warped image comprises:
acquiring an initial deformation vertex, and determining the direction and the strength of the deformation of the initial deformation vertex;
according to the direction and the strength, a preset formula is adopted to transmit the initial deformation vertex to a distortion vertex corresponding to a vertex at other positions to obtain a distortion image, wherein the preset formula is as follows:
pT＝pI+wv；
wherein p isTFor warping the coordinates of the vertex T, pIThe coordinates of the initial deformation vertex P of the preset plane image I are set, w is the deformation strength of the deformation corresponding to the distortion, and v is the deformation direction of the deformation corresponding to the distortion.
7. A warped image correction method according to claim 6, wherein for a warping of the wrinkle type, w is defined as follows:
and/or for curve type warping, define w as follows:
w＝1-dα；
and d is the normalized distance of a straight line determined based on p and v in the warping process of warping the preset plane image, and alpha is used for controlling the propagation range of the initial deformation vertex for deformation.
8. A distorted image correction apparatus, characterized in that the apparatus comprises:
the image warping method comprises the steps that a first input unit is used for obtaining an initial warped image, inputting the initial warped image to a preset first U-shaped neural network model contained in a preset superposition U-shaped neural network model, and obtaining an image warping disturbance field contained in the initial warped image, wherein the image warping disturbance field is used for describing the deformation quantization characteristics of a flattened image in the deformation process of the initial warped image;
the second input unit is used for inputting the image distortion disturbing field to a preset second U-shaped neural network model contained in the preset superposition U-shaped neural network model to obtain an image restoration disturbing field, wherein the image restoration disturbing field is used for describing deformation quantization characteristics in a deformation process of deforming the initial distortion image into a flattened image;
and the restoration unit is used for restoring the initial distorted image by adopting the image restoration disturbance field to obtain a corrected image corresponding to the initial distorted image.
9. A computer device, comprising a memory and a processor coupled to the memory; the memory is used for storing a computer program; the processor is adapted to run the computer program to perform the steps of the method according to any of claims 1-7.
10. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when being executed by a processor, carries out the steps of the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210293242.9A CN114677684A (en) | 2022-03-23 | 2022-03-23 | Distorted image correction method, device and equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210293242.9A CN114677684A (en) | 2022-03-23 | 2022-03-23 | Distorted image correction method, device and equipment and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114677684A true CN114677684A (en) | 2022-06-28 |
Family
ID=82073986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210293242.9A Pending CN114677684A (en) | 2022-03-23 | 2022-03-23 | Distorted image correction method, device and equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114677684A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114882116A (en) * | 2022-07-12 | 2022-08-09 | 南通市恒瑞精密机械制造有限公司 | Lead screw bending positioning and straightening method and system |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110322416A (en) * | 2019-07-09 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Image processing method, device and computer readable storage medium |
| US20200134789A1 (en) * | 2018-10-31 | 2020-04-30 | Chung Ang University Industry Academic Cooperation Foundation | System and method for correcting image through estimation of distortion parameter |
| US20200286212A1 (en) * | 2019-03-07 | 2020-09-10 | Ricoh Company, Ltd. | Image processing method, apparatus, and computer-readable recording medium |
| US20200327639A1 (en) * | 2019-04-10 | 2020-10-15 | Eagle Technology, Llc | Hierarchical Neural Network Image Registration |
| CN111898701A (en) * | 2020-08-13 | 2020-11-06 | 网易（杭州）网络有限公司 | Model training, frame image generation, frame interpolation method, device, equipment and medium |
| CN112561807A (en) * | 2020-11-26 | 2021-03-26 | 北京大学 | End-to-end radial distortion correction method based on convolutional neural network |
| US20210211739A1 (en) * | 2020-01-05 | 2021-07-08 | Isize Limited | Processing image data |
| US20210334934A1 (en) * | 2020-04-24 | 2021-10-28 | Apple Inc. | Local image warping in image processor using homography transform function |
| CN113591528A (en) * | 2021-02-05 | 2021-11-02 | 腾讯科技（深圳）有限公司 | Document correction method, device, computer equipment and storage medium |
| CN113808033A (en) * | 2021-08-06 | 2021-12-17 | 上海深杳智能科技有限公司 | Image document correction method, system, terminal and medium |
| CN113971629A (en) * | 2020-07-24 | 2022-01-25 | 三星电子株式会社 | Image restoration method and device |
-
2022
- 2022-03-23 CN CN202210293242.9A patent/CN114677684A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200134789A1 (en) * | 2018-10-31 | 2020-04-30 | Chung Ang University Industry Academic Cooperation Foundation | System and method for correcting image through estimation of distortion parameter |
| US20200286212A1 (en) * | 2019-03-07 | 2020-09-10 | Ricoh Company, Ltd. | Image processing method, apparatus, and computer-readable recording medium |
| US20200327639A1 (en) * | 2019-04-10 | 2020-10-15 | Eagle Technology, Llc | Hierarchical Neural Network Image Registration |
| CN110322416A (en) * | 2019-07-09 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Image processing method, device and computer readable storage medium |
| US20210211739A1 (en) * | 2020-01-05 | 2021-07-08 | Isize Limited | Processing image data |
| US20210334934A1 (en) * | 2020-04-24 | 2021-10-28 | Apple Inc. | Local image warping in image processor using homography transform function |
| CN113971629A (en) * | 2020-07-24 | 2022-01-25 | 三星电子株式会社 | Image restoration method and device |
| CN111898701A (en) * | 2020-08-13 | 2020-11-06 | 网易（杭州）网络有限公司 | Model training, frame image generation, frame interpolation method, device, equipment and medium |
| CN112561807A (en) * | 2020-11-26 | 2021-03-26 | 北京大学 | End-to-end radial distortion correction method based on convolutional neural network |
| CN113591528A (en) * | 2021-02-05 | 2021-11-02 | 腾讯科技（深圳）有限公司 | Document correction method, device, computer equipment and storage medium |
| CN113808033A (en) * | 2021-08-06 | 2021-12-17 | 上海深杳智能科技有限公司 | Image document correction method, system, terminal and medium |
Non-Patent Citations (1)
| Title |
|---|
| 张峰峰等: "基于BP神经网络的X光图像畸变校正技术的研究", 计算机应用研究, vol. 25, no. 10, 15 October 2008 (2008-10-15), pages 3050 - 3052 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114882116A (en) * | 2022-07-12 | 2022-08-09 | 南通市恒瑞精密机械制造有限公司 | Lead screw bending positioning and straightening method and system |
| CN114882116B (en) * | 2022-07-12 | 2022-09-06 | 南通市恒瑞精密机械制造有限公司 | Lead screw bending positioning and straightening method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3839875B1 (en) | Method and apparatus for correcting distorted document image | |
| JP5722482B2 (en) | Image processing system and image processing method | |
| US8285077B2 (en) | Automatic correction of digital image distortion | |
| EP2536122B1 (en) | Image processing method, image processing device and scanner | |
| JP6147273B2 (en) | Image registration method and system robust against noise | |
| You et al. | Multiview rectification of folded documents | |
| Dong et al. | Image reconstruction with locally adaptive sparsity and nonlocal robust regularization | |
| US11830170B2 (en) | Systems and methods for image data processing to correct document deformations using machine learning system | |
| CN110322400B (en) | Image processing method and device, image processing system and training method thereof | |
| CN108846385B (en) | Image identification and correction method and device based on convolution-deconvolution neural network | |
| CN112560861A (en) | Bill processing method, device, equipment and storage medium | |
| CN114677684A (en) | Distorted image correction method, device and equipment and computer readable storage medium | |
| CN114612289A (en) | Stylized image generation method and device and image processing equipment | |
| CN109697442B (en) | Training method and device of character recognition model | |
| CN114742722A (en) | Document correction method, device, electronic equipment and storage medium | |
| WO2022147285A4 (en) | Book scanning using machine-trained model | |
| US20040156556A1 (en) | Image processing method | |
| JP6645442B2 (en) | Information processing apparatus, information processing method, and program | |
| CN102843479A (en) | File scanning method, file scanning device and portable electronic device | |
| US11145037B1 (en) | Book scanning using machine-trained model | |
| CN115187834A (en) | Bill identification method and device | |
| CN113591528A (en) | Document correction method, device, computer equipment and storage medium | |
| CN112241740B (en) | Feature extraction method and device | |
| CN113496468A (en) | Method and device for restoring depth image and storage medium | |
| Sasao et al. | A study on effect of automatic perspective correction on exemplar-based image inpainting |
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 |