CN115512367A - Paper image standardization method and device, computer equipment and storage medium - Google Patents
Paper image standardization method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN115512367A CN115512367A CN202211220060.5A CN202211220060A CN115512367A CN 115512367 A CN115512367 A CN 115512367A CN 202211220060 A CN202211220060 A CN 202211220060A CN 115512367 A CN115512367 A CN 115512367A
- Authority
- CN
- China
- Prior art keywords
- contour
- vertex
- coordinates
- pixel point
- paper
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/15—Cutting or merging image elements, e.g. region growing, watershed or clustering-based techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/18—Extraction of features or characteristics of the image
- G06V30/1801—Detecting partial patterns, e.g. edges or contours, or configurations, e.g. loops, corners, strokes or intersections
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of computer vision, and provides a paper image standardization method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list; screening contour vertex coordinates from the contour pixel point coordinate list based on the vertex distance maximum extreme characteristic and the vertex right-angle characteristic; and standardizing the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image. The method can improve the accuracy of text reconstruction.
Description
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a paper image standardization method and device, computer equipment and a storage medium.
Background
Paper media are the main carriers for transmitting information in daily life of people at present, and are popular among users due to the convenience and carrying simplicity of paper media in reading. However, because paper media are not easy to store for a long time, paper information is inconvenient to edit for the second time, and other problems, the text content of paper needs to be manually edited on electronic media such as Word, PDF, and the like, and the working efficiency is seriously influenced. Therefore, in order to solve the tedious work, a research work of document reconstruction has been developed in academic and industrial circles, and the core task thereof is to convert a paper document into an electronic document by taking a paper image based on the image.
However, in practical applications, the user can shoot the paper document freely and uncontrollably, which causes a series of problems of distortion, occlusion, deformation and the like of the paper image, and the shape of the paper image is greatly different from that of a standard paper document, thereby greatly increasing the difficulty of document reconstruction and reducing the accuracy of document reconstruction.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a paper image normalization method, apparatus, computer device, and storage medium that can improve the accuracy of text reconstruction.
The invention provides a paper image standardization method, which comprises the following steps:
extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
screening contour vertex coordinates from the contour pixel point coordinate list based on the vertex distance maximum extreme value characteristics and the vertex right angle characteristics;
and standardizing the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
In one embodiment, the screening of the vertex coordinates of the contour from the contour pixel point coordinate list based on the vertex distance maximum extremum feature and the vertex right angle feature includes:
screening contour pixel point coordinates from the contour pixel point coordinate list as candidate contour vertex coordinates based on the maximum extreme value characteristics of the vertex distance;
and determining the coordinates of the vertexes of the contours from the candidate vertexes of the contours based on the vertex rectangular features.
In one embodiment, the screening, from the contour pixel point coordinate list, the contour pixel point coordinates as candidate contour vertex coordinates based on the maximum extreme feature of vertex distance includes:
determining the coordinates of the contour central points according to the coordinates of each contour pixel point in the contour pixel point coordinate list;
the contour pixel point coordinate list is regarded as an end-to-end ring, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set;
respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set;
and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
In one embodiment, the determining the vertex coordinates of the contour from the candidate vertex coordinates of the contour based on the vertex rectangular feature includes:
traversing each candidate contour vertex coordinate, acquiring a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinate based on the position of the current candidate contour vertex coordinate in the contour pixel point coordinate list, performing straight line fitting to obtain a first straight line function parameter, and acquiring a second preset number of contour pixel point coordinates after the current candidate contour vertex coordinate, and performing straight line fitting to obtain a second straight line function parameter;
calculating the straight line included angle of the vertex coordinate of the current candidate contour by using the first straight line function parameter and the second straight line function parameter to obtain a list of the straight line included angles corresponding to the vertex coordinate of each candidate contour;
and screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates based on the distance between each straight line included angle and the right-angle included angle in the straight line included angle list.
In one embodiment, the normalizing the to-be-processed paper image according to the contour vertex coordinates to obtain a normalized paper image includes:
determining standard vertex coordinates of the paper image to be processed according to the industry standard size of paper contained in the paper image to be processed;
computing a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates;
and standardizing the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
In one embodiment, the extracting the paper content area of the paper image to be processed includes:
acquiring a paper image to be processed, and calling a semantic segmentation network trained in a knowledge distillation mode;
and inputting the paper image to be processed into the semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
In one embodiment, the extracting the pixel coordinates of the contour of the paper content area to obtain a contour pixel coordinate list includes:
amplifying the paper content area, and then carrying out edge detection to obtain a pixel coordinate point of the outline of the paper content area;
and sequentially arranging the coordinates of the pixel points clockwise according to the outline to obtain an outline pixel point coordinate list.
A sheet image normalization apparatus, comprising:
the extraction module is used for extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
the vertex screening module is used for screening the vertex coordinates of the contour from the contour pixel point coordinate list based on the maximum extreme value characteristics of the vertex distance and the vertex right angle characteristics;
and the standardization processing module is used for carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
The present invention also provides a computer apparatus comprising a processor and a memory, the memory storing a computer program, the processor implementing the steps of the paper image normalization method according to any one of the above when executing the computer program.
The present invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the paper image normalization method of any one of the above.
According to the paper image standardization method, the paper image standardization device, the computer equipment and the storage medium, the area of the paper is extracted from the complex image in a positioning mode, the outline vertex of the paper area is found out based on the outline of the paper and the inherent invariant feature of the right angle, and finally the standardization processing of the paper image at any angle is realized based on the outline vertex to obtain the standardized paper image, so that the standard paper image can be provided for a series of subsequent steps such as text layout analysis, text detection and text recognition in the subsequent document reconstruction process, the difficulty of the subsequent steps is greatly reduced, and the accuracy of a text reconstruction task is improved.
Drawings
FIG. 1 is a diagram of an application environment of a method for normalizing a sheet image in one embodiment.
FIG. 2 is a flowchart illustrating a method for normalizing a sheet image in one embodiment.
FIG. 3 is a schematic illustration of an image of a sheet to be processed in one embodiment.
FIG. 4 is an image of a paper content area in one embodiment.
FIG. 5 is a block diagram showing the structure of a sheet image normalization apparatus according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The paper image normalization method provided by the application can be applied to the application environment shown in fig. 1, wherein the application environment relates to the terminal 102 and the server 104. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 may be, but is not limited to, various personal computers, laptops, smartphones, tablets and portable wearable devices, and the server 104 may be implemented as a stand-alone server or a server cluster composed of a plurality of servers. The paper image normalization method may be implemented by the terminal 102 alone or by the server 104. Taking the server 104 as an example, specifically, the server 104 extracts a paper content area of a paper image to be processed and extracts pixel point coordinates of a contour of the paper content area to obtain a contour pixel point coordinate list; screening contour vertex coordinates from a contour pixel point coordinate list based on the vertex distance maximum extreme value characteristics and the vertex right angle characteristics; and carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
In one embodiment, as shown in fig. 2, a method for standardizing a paper image is provided, which is described by taking the method as an example for a server, and includes the following steps:
step S201, extracting a paper content area of the paper image to be processed and extracting pixel coordinates of a contour of the paper content area to obtain a contour pixel coordinate list.
The image of the paper to be processed refers to an image containing paper content, and the image of the paper to be processed may be an image containing only the paper content or an image containing both the paper content and other object content. As shown in fig. 3, a schematic diagram of an image of a sheet to be processed is provided, and the image shown in fig. 3 has other objects such as a keyboard, a table-ware, and the like in addition to the content of the sheet. The paper content area refers to an area including only paper content, and as shown in fig. 4, a schematic diagram of the paper content area is provided, and a central gray rectangular area in fig. 4 is the area including the paper content in fig. 3.
Specifically, when a server receives a paper image standardization task, firstly, a paper image to be processed is obtained, then a paper content area in the paper image to be processed is extracted, then, pixel points on the outline of the paper content area are extracted, and an outline pixel point coordinate list is obtained.
And S202, screening the vertex coordinates of the contour from the contour pixel point coordinate list based on the maximum extreme point characteristic of the vertex distance and the vertex right-angle characteristic.
The distance between the paper vertex and the paper center point is the farthest relative to the adjacent peripheral pixel points, so that the method is also suitable for the image distortion scene. In addition, since the vertex angle of the paper is a characteristic of a right angle, although the vertex angle is transformed to a certain degree in the case of distortion or deformation of an image, the vertex angle is still the point closest to the right angle, and thus, the invalid point can be filtered to obtain the vertex by utilizing this property.
In one embodiment, step S202 includes: screening contour pixel point coordinates from a contour pixel point coordinate list based on the maximum extreme value characteristics of the vertex distance to serve as candidate contour vertex coordinates; and determining the coordinates of the vertex of the contour from the candidate coordinates of the vertex of the contour based on the vertex rectangular feature.
Specifically, since the contour coordinate points that can be screened out by the maximum extreme feature of the distance from the central point do not necessarily satisfy the rectangular feature, when the contour vertex coordinates are screened out by using the maximum extreme feature of the vertex distance and the vertex rectangular feature, a part of candidate vertex coordinates are screened out from the contour pixel point coordinate list by using the vertex of the maximum extreme feature of the vertex distance, and then invalid candidate vertices are filtered out from the candidate vertex coordinates by using the vertex rectangular feature to obtain the determined vertex coordinates.
And step S203, carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
Specifically, after determining the contour vertex coordinates of the paper image to be processed, the paper image to be processed is transformed based on the contour vertex coordinates and the standard vertex coordinates of the paper to complete the standardization processing, and a standardized paper image is obtained.
According to the paper image standardization method, the area of the paper is positioned and extracted from the complex image, the outline vertex of the paper area is found out based on the outline of the paper and the inherent invariant feature of a right angle, and finally standardization processing of the paper image at any angle is realized based on the outline vertex to obtain the standardized paper image, so that the standard paper image can be provided for a series of subsequent steps such as text layout analysis, text detection, text recognition and the like in the subsequent document reconstruction process, the difficulty of the subsequent steps is greatly reduced, and the accuracy of a text reconstruction task is improved.
In one embodiment, screening contour pixel point coordinates from a contour pixel point coordinate list as candidate contour vertex coordinates based on the maximum extreme feature of vertex distance comprises: determining the coordinates of the contour central points according to the coordinates of the contour pixel points in the contour pixel point coordinate list; the contour pixel point coordinate list is regarded as a ring which is connected end to end, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set; respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set; and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
The contour center point coordinate refers to a center point coordinate of paper contained in the paper image to be processed. The current contour pixel point coordinate refers to a contour pixel point coordinate currently visited when the coordinates of all contour pixel points in the contour pixel point coordinate list are sequentially traversed and visited. The first preset number is preset and is used for determining the numerical value of the number of the coordinates of the selected pixel points and can be set according to actual requirements.
Specifically, in a complex and distorted scene, the distance from the paper vertex to the central point is farthest relative to other pixel points, so that the candidate contour vertex can be accurately screened out from the contour pixel point coordinate list by using the characteristic. Firstly, the coordinate (x) of the center point of the outline is obtained according to the coordinate list IContour of the pixel points of the outline c ，y y ) Contour center point coordinate (x) c ，y y ) The calculation formula is as follows:
wherein n represents the total number of contour pixel coordinates in the contour pixel coordinate list IContour, (x) i ，y i ) And (3) representing the ith contour pixel point coordinate in the contour pixel point coordinate list IContour, wherein i = {1,2,3 \8230; \ 8230; (n }). Then, regarding the contour pixel point coordinate list IContour as an end-to-end ring, and regarding any one contour pixel point coordinate (x) i ，y i ) The position is got and is located k profile pixel points and the position is located k profile pixel points of its front and forms a set of profile pixel point coordinate set, and k is for predetermineeing first quantity promptly, and profile pixel point coordinate set record is PC, specifically as follows:
PC＝{(x i-k ，y i-k )，(x i-k+1 ，y i-k+1 )，…，(x i ，y i )，…，(x i+k-1 ，y i+k-1 )，(x i+k ，y i+k )}
secondly, calculating the distance (x) between each contour pixel point coordinate in the contour pixel point coordinate set PC and the contour central point coordinate in sequence c ，y y ) A pixel distance group DisPC is obtained, which is specifically as follows:
DisPC＝{d i-k ，d i-k+1 ，…，d i ，…，d i+k-1 ，d i+k }
finally, if the first k elements in the pixel distance group DisPC are incremented and the last k elements are decremented, i.e., with the current contour pixel point coordinate (x) i ，y i ) Corresponding distance d i Judging whether the distance in front of the boundary point is in an increasing state and the distance behind the boundary point is in a decreasing state for the boundary point, if so, determining the coordinate (x) of the contour pixel point i ，y i ) And if the requirement of the candidate contour vertex coordinates is met, the candidate contour vertex coordinates are taken as the candidate contour vertex coordinates. Traversing each contour pixel point coordinate in the contour pixel point coordinate list IContour in the above manner, and finally obtaining all candidate contour vertex coordinates.
In one embodiment, determining the silhouette vertex coordinates from the candidate silhouette vertex coordinates based on the vertex rectangular feature comprises: traversing all candidate contour vertex coordinates, acquiring a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinates based on the position of the current candidate contour vertex coordinates in the contour pixel point coordinate list, performing straight line fitting to obtain a first straight line function parameter, and acquiring a second preset number of contour pixel point coordinates after the current candidate contour vertex coordinates, and performing straight line fitting to obtain a second straight line function parameter; calculating a straight line included angle of the vertex coordinates of the current candidate contour by using the first straight line function parameters and the second straight line function parameters to obtain a list of the straight line included angles corresponding to the vertex coordinates of each candidate contour; and based on the distance between each straight line included angle in the straight line included angle list and the right-angle included angle, screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates.
And the current candidate contour vertex coordinate is the candidate contour vertex coordinate which is currently visited when traversal visit is carried out on each candidate contour vertex coordinate. The second preset number is the same as the first preset number, is a preset numerical value for determining the number of the coordinates of the selected pixel points, and can be set according to actual requirements. The second predetermined number may be equal to or different from the first predetermined number.
Specifically, each candidate contour vertex coordinate is traversed in sequence, and the traversed current candidate contour vertex coordinate (x) is used for the current candidate contour vertex coordinate j ，y j ) And positioning the position of the contour pixel point coordinate set in the contour pixel point coordinate list IContour, and sequentially taking the contour pixel point coordinates t before the contour pixel point coordinate set and the contour pixel point coordinates t after the contour pixel point coordinate set to form a pixel point coordinate set SC.
SC＝{(x j-t ，y j-t )，(x j-t+1 ，y j-t+1 )，…，(x j ，y j )，…，(x j+t-1 ，y j+t-1 )，(x j+t ，y j+t )}
And then, performing linear fitting on the coordinates of the first t contour pixel points in the SC to obtain a first linear function parameter, and similarly, performing linear fitting on the coordinates of the last t contour pixel points in the SC to obtain a second linear function parameter. In this embodiment, the straight line fitting manner may adopt any one of the existing manners, such as a least square method, a gradient descent method, and the like, in this embodiment, the least square method is preferably used for straight line fitting, and the first straight line function and the second straight line function obtained by fitting are schematically shown as follows:
y t1 ＝a 1 *x t1 +b 1
y t2 ＝a 2 *x t2 +b 2
reuse of the first linear function parameter a 1 、b 1 And a second linear function a 2 、b 2 Parameter calculation of current candidate contour vertex coordinates (x) j ，y j ) Angle of straight line theta j . Calculating the straight line included angles of all candidate contour vertex coordinates through the steps to form a straight lineAngle list IAngle = { theta = j-1 ,…θ j ,…,θ j+1 }. Included angle of straight line theta j The calculation formula of (a) is as follows:
and finally, calculating the distance between each straight line included angle and a right angle included angle pi/2 in the straight line included angle list, and recording the distance as IAngle _ Dis = { | theta j-1 -π/2|,…|θ j -π/2|,…,|θ j+1 - π/2| }. Selecting the minimum four elements from the IAngle _ Dis, and selecting four candidate contour vertex coordinates from the candidate contour vertex coordinates according to the corresponding relation as finally determined contour vertex coordinates, wherein the contour vertex coordinates can be recorded as ISV = { (x) s1 ,y s1 )，(x s2 ,y s2 )，(x s3 ,y s3 )，(x s4 ,y s4 ) Wherein (x) s1 ,y s1 ) Is the top left vertex, (x) s2 ,y s2 ) Is the top right vertex, (x) s3 ,y s3 ) Is the lower right vertex, (x) s4 ,y s4 ) The lower left vertex. The top left vertex can be determined by the condition that the vertex with the minimum y coordinate is selected from the points with the x coordinate smaller than the mean value of the x coordinates of the four vertices, and then clockwise sequencing is performed based on the corresponding position relation of the contour pixel point coordinate list IContour to sequentially obtain the top right vertex, the bottom right vertex and the bottom left vertex. In this embodiment, the final contour vertex coordinates are obtained by filtering invalid candidate points based on the feature that the paper vertex is a right angle and tends to be a right angle under the condition of image distortion and deformation, and the accuracy of vertex acquisition is improved.
In one embodiment, step S203 includes: determining standard vertex coordinates of the paper image to be processed according to the industry standard size of paper contained in the paper image to be processed; calculating a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates; and carrying out standardization processing on the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
Specifically, first, according to the industry of the paper contained in the paper image to be processedAnd determining standard vertex coordinates of the paper image to be processed according to the standard size. For example, when A4 sheet is included in the image of the sheet to be processed, the standard width w of the A4 sheet can be determined A4 And a height h A4 Obtaining a standard upper left vertex (0, 0) and a standard upper right vertex (w) A4 0), standard lower right vertex (w) A4 ,h A4 ) And the standard lower left vertex (0, h) A4 ). Then, according to the contour vertex coordinates (x) s1 ,y s1 )，(x s2 ,y s2 )，(x s3 ,y s3 )，(x s4 ,y s4 ) And the standard vertex coordinates (0, 0), (w) A4 ,0)，(w A4 ,h A4 )，(0,h A4 ) The transformation matrix M is calculated using perspective transformation in image processing. And finally, standardizing the paper image to be processed through a transformation matrix M to obtain a standardized paper image. In this embodiment, the paper is subjected to the standardization processing based on the standard vertex coordinates and the contour vertex coordinates of the paper, and the accuracy of the standardization processing can be improved.
In one embodiment, extracting a sheet content area of a sheet image to be processed includes: acquiring a paper image to be processed, and calling a semantic segmentation network trained in a knowledge distillation mode; and inputting the paper image to be processed into a semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
Specifically, the server acquires a paper image I to be processed, and extracts a paper content area in the paper image I to be processed based on a semantic segmentation network. The semantic segmentation network in this embodiment may adopt any existing neural network architecture, and a PIDNet network architecture is preferred in this embodiment. And respectively building a semantic segmentation lightweight network PID _ Small and a heavy-weight network PID _ Large based on a PIDNet network architecture. Collecting a batch of paper images containing paper content areas, labeling the paper content areas in the paper images by using labeling tools such as LabelMe and the like to obtain corresponding image labels, and constructing the labeled paper images into a training data set S based on the ratio of 8: 2. Then, a semantic segmentation heavy-weight network PID _ Large is trained based on a training data set S, wherein training parameters are consistent with an original PIDNet network architecture, and a trained heavy-weight network PID _ Large model is recorded as PID _ Large _ Frozen. And taking PID _ Large _ Frozen as a Teacher model, taking PID _ Small of a semantic segmentation lightweight network as a Student model, training the Student model by adopting a knowledge distillation mode based on a training data set S, and recording the trained model as PID _ Small _ Frozen as the trained semantic segmentation model. And processing the subsequent paper image I to be processed by adopting a PID _ Small _ Frozen model to obtain a paper content area. That is, the paper-based image to be processed shown in fig. 3 is input to the PID _ Small _ Frozen model for processing, and the image shown in fig. 4 is output, where fig. 3 is identical to the image shown in fig. 4 in size, except that the pixel point value of the paper content area portion in fig. 4 is different from the pixel point value of other areas, for example, the pixel point value of the paper content area portion in fig. 4 is 1, and the pixel point value of the other areas is 0. The distillation loss function kd _ loss is defined in this example knowledge as follows:
kd_loss＝loss so +KLDivLoss(log_softmax(so/T)，softmax(to/T))*α*T 2
therein, loss so For the loss function of the PID _ Small network, KLDivloss, log _ softmax and softmax are common functions in a deep learning framework, and so and to respectively represent output results after the PID _ Small network and the PID _ Large _ Frozen model are combined. a is a weight equilibrium coefficient, and T is the temperature of knowledge distillation.
In one embodiment, extracting pixel coordinates of a contour of a paper content area to obtain a contour pixel coordinate list includes: carrying out edge detection after the paper content area is amplified to obtain a pixel coordinate point of the outline of the paper content area; and sequentially arranging the coordinates of all the pixel points clockwise according to the contour to obtain a contour pixel point coordinate list.
Specifically, when the server extracts the outline pixel point coordinate list, the server first multiplies the paper content area by 225 to perform amplification processing, taking a pixel point value 1 as an example, that is, the pixel point value of the paper content area is 225, and the pixel points of other areas are unchanged. Then, an edge detection operation is performed on the image after the paper content area is enlarged, and the edge detection operation is preferably implemented by using Canny operator in this embodiment, so that each pixel coordinate point on the outline of the paper content area is obtained. And subsequently, screening contour vertex coordinates based on the maximum vertex distance characteristic and the vertex right-angle characteristic, and sequentially arranging the pixel point coordinates according to the clockwise direction of the contour to obtain a contour pixel point coordinate list which is sequentially arranged according to the clockwise direction for convenient processing. For example, taking the image shown in fig. 4 as an example, after multiplying the pixel point value of the paper content area shown in the image of fig. 4 by 225, edge detection is performed on the enlarged paper content area in fig. 4 to obtain the coordinates of each pixel point on the outline.
It should be understood that, although the steps in the flowchart of fig. 2 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in fig. 2 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.
In one embodiment, as shown in fig. 5, there is provided a sheet image normalization apparatus including:
the extraction module 501 is configured to extract a paper content area of a paper image to be processed and extract pixel coordinates of a contour of the paper content area to obtain a contour pixel coordinate list;
a vertex screening module 502, configured to screen a contour vertex coordinate from the contour pixel point coordinate list based on the vertex distance maximum extremum feature and the vertex right angle feature;
and a standardization processing module 503, configured to perform standardization processing on the to-be-processed paper image according to the contour vertex coordinates, so as to obtain a standardized paper image.
In one embodiment, the vertex screening module 502 is further configured to screen contour pixel coordinates from the contour pixel coordinate list as candidate contour vertex coordinates based on the maximum extreme feature of the vertex distance; and determining the coordinates of the vertex of the contour from the candidate coordinates of the vertex of the contour based on the vertex rectangular feature.
In one embodiment, the vertex filtering module 502 is further configured to determine the coordinates of the contour center point according to the coordinates of each contour pixel point in the contour pixel point coordinate list; the contour pixel point coordinate list is regarded as an end-to-end ring, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set; respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set; and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
In one embodiment, the vertex screening module 502 is further configured to traverse each candidate contour vertex coordinate, obtain, based on a position of the current candidate contour vertex coordinate in the contour pixel point coordinate list, a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinate for performing straight line fitting to obtain a first straight line function parameter, and obtain, after the current candidate contour vertex coordinate, a second preset number of contour pixel point coordinates for performing straight line fitting to obtain a second straight line function parameter; calculating the straight line included angle of the vertex coordinate of the current candidate contour by using the first straight line function parameter and the second straight line function parameter to obtain a list of the straight line included angles corresponding to the vertex coordinate of each candidate contour; and based on the distance between each straight line included angle in the straight line included angle list and the right-angle included angle, screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates.
In one embodiment, the normalization processing module 503 is further configured to determine standard vertex coordinates of the to-be-processed paper image according to an industry standard size of paper included in the to-be-processed paper image; calculating a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates; and carrying out standardization processing on the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
In one embodiment, the extraction module 501 is further configured to obtain an image of a paper to be processed, and call a semantic segmentation network trained in a knowledge distillation manner; and inputting the paper image to be processed into a semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
In one embodiment, the extracting module 501 is further configured to perform edge detection after the paper content area is enlarged, so as to obtain a pixel coordinate point of the outline of the paper content area; and sequentially arranging the coordinates of all the pixel points clockwise according to the outline to obtain an outline pixel point coordinate list.
For specific definition of the paper image standardization means, reference may be made to the definition of the paper image standardization method above, and details are not repeated here. Each block in the above-described paper image standardizing apparatus may be entirely or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in hardware or independent of a processor in the computer device, or can be stored in a memory in the computer device in software, so that the processor can call and execute operations corresponding to the modules. Based on such understanding, all or part of the flow in the method according to the above embodiments may be implemented by a computer program, which may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the steps of the above embodiments of the method for standardizing a paper image may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc.
In one embodiment, a computer device, which may be a server, is provided that includes a processor, a memory, and a network interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program to run on the non-volatile storage medium. The database of the computer device is used for storing data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a paper image normalization method. Illustratively, a computer program may be partitioned into one or more modules, which are stored in a memory and executed by a processor to implement the present invention. One or more of the modules may be a sequence of computer program instruction segments for describing the execution of a computer program in a computer device that is capable of performing certain functions.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like which is the control center for the computer device and which connects the various parts of the overall computer device using various interfaces and lines.
The memory may be used to store the computer programs and/or modules, and the processor may implement various functions of the computer apparatus by executing or executing the computer programs and/or modules stored in the memory and calling data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, etc. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
It will be understood by those skilled in the art that the computer device structure shown in the embodiment is only a partial structure related to the solution of the present invention, and does not constitute a limitation to the computer device to which the present invention is applied, and a specific computer device may include more or less components, or combine some components, or have different component arrangements.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
screening contour vertex coordinates from a contour pixel point coordinate list based on the vertex distance maximum extreme value characteristics and the vertex right angle characteristics;
and carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
In one embodiment, the processor when executing the computer program further performs the steps of: screening contour pixel point coordinates from a contour pixel point coordinate list based on the maximum extreme value characteristics of the vertex distance to serve as candidate contour vertex coordinates; and determining the coordinates of the vertex of the contour from the candidate coordinates of the vertex of the contour based on the vertex rectangular feature.
In one embodiment, the processor, when executing the computer program, further performs the steps of: determining the coordinates of the contour central points according to the coordinates of each contour pixel point in the contour pixel point coordinate list; the contour pixel point coordinate list is regarded as an end-to-end ring, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set; respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set; and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
In one embodiment, the processor, when executing the computer program, further performs the steps of: traversing each candidate contour vertex coordinate, acquiring a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinate to perform straight line fitting to obtain a first straight line function parameter based on the position of the current candidate contour vertex coordinate in the contour pixel point coordinate list, and acquiring a second preset number of contour pixel point coordinates after the current candidate contour vertex coordinate to perform straight line fitting to obtain a second straight line function parameter; calculating the straight line included angle of the vertex coordinate of the current candidate contour by using the first straight line function parameter and the second straight line function parameter to obtain a list of the straight line included angles corresponding to the vertex coordinate of each candidate contour; and based on the distance between each straight line included angle in the straight line included angle list and the right-angle included angle, screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates.
In one embodiment, the processor, when executing the computer program, further performs the steps of: determining standard vertex coordinates of the paper image to be processed according to the industry standard size of paper contained in the paper image to be processed; calculating a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates; and carrying out standardization processing on the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
In one embodiment, the processor when executing the computer program further performs the steps of: acquiring a paper image to be processed, and calling a semantic segmentation network trained in a knowledge distillation mode; and inputting the paper image to be processed into a semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
In one embodiment, the processor when executing the computer program further performs the steps of: amplifying the paper content area, and then carrying out edge detection to obtain a pixel coordinate point of the outline of the paper content area; and sequentially arranging the coordinates of all the pixel points clockwise according to the contour to obtain a contour pixel point coordinate list.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, performs the steps of:
extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
screening contour vertex coordinates from a contour pixel point coordinate list based on the vertex distance maximum extreme characteristic and the vertex right-angle characteristic;
and carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image. In one embodiment, the computer program when executed by the processor further performs the steps of: screening contour pixel point coordinates from a contour pixel point coordinate list based on the maximum extreme value characteristics of the vertex distance to serve as candidate contour vertex coordinates; and determining the coordinates of the vertexes of the contours from the coordinates of the vertexes of the candidate contours based on the vertex rectangular features.
In one embodiment, the computer program when executed by the processor further performs the steps of: determining the coordinates of the contour central points according to the coordinates of the contour pixel points in the contour pixel point coordinate list; the contour pixel point coordinate list is regarded as a ring which is connected end to end, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set; respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set; and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
In one embodiment, the computer program when executed by the processor further performs the steps of: traversing all candidate contour vertex coordinates, acquiring a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinates based on the position of the current candidate contour vertex coordinates in the contour pixel point coordinate list, performing straight line fitting to obtain a first straight line function parameter, and acquiring a second preset number of contour pixel point coordinates after the current candidate contour vertex coordinates, and performing straight line fitting to obtain a second straight line function parameter; calculating the straight line included angle of the vertex coordinate of the current candidate contour by using the first straight line function parameter and the second straight line function parameter to obtain a list of the straight line included angles corresponding to the vertex coordinate of each candidate contour; and screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates based on the distance between each straight line included angle and the right-angle included angle in the straight line included angle list.
In one embodiment, the computer program when executed by the processor further performs the steps of: determining standard vertex coordinates of the paper image to be processed according to the industry standard size of paper contained in the paper image to be processed; calculating a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates; and carrying out standardization processing on the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring a paper image to be processed, and calling a semantic segmentation network trained in a knowledge distillation mode; and inputting the paper image to be processed into a semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
In one embodiment, the computer program when executed by the processor further performs the steps of: amplifying the paper content area, and then carrying out edge detection to obtain a pixel coordinate point of the outline of the paper content area; and sequentially arranging the coordinates of all the pixel points clockwise according to the outline to obtain an outline pixel point coordinate list.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is specific and detailed, but not to be understood as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method of normalizing a sheet image, comprising:
extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
screening contour vertex coordinates from the contour pixel point coordinate list based on the vertex distance maximum extreme value characteristics and the vertex right angle characteristics;
and standardizing the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
2. The method of claim 1, wherein the screening contour vertex coordinates from the contour pixel point coordinate list based on vertex distance maximum extremum features and vertex right angle features comprises:
screening contour pixel point coordinates from the contour pixel point coordinate list based on the maximum extreme characteristic of the vertex distance to serve as candidate contour vertex coordinates;
and determining the coordinates of the vertexes of the contours from the candidate vertexes of the contours based on the vertex rectangular features.
3. The method of claim 2, wherein the screening contour pixel coordinates from the contour pixel coordinate list as candidate contour vertex coordinates based on the vertex distance maximum extremum feature comprises:
determining the coordinates of the contour central points according to the coordinates of the contour pixel points in the contour pixel point coordinate list;
the contour pixel point coordinate list is regarded as an end-to-end ring, all contour pixel point coordinates in the contour pixel point coordinate list are traversed, and a first preset number of contour pixel point coordinates before the current contour pixel point coordinates and a first preset number of contour pixel point coordinates after the current contour pixel point coordinates form a contour pixel point coordinate set;
respectively calculating the distance between each contour pixel point coordinate in the contour pixel point coordinate set and the contour central point coordinate to obtain a pixel distance set;
and taking the distance corresponding to the current contour pixel point coordinate as a boundary point, and taking the current contour pixel point coordinate as a candidate contour vertex coordinate when the distance in the pixel distance group before the boundary point is in an increasing state and the distance after the boundary point is in a decreasing state.
4. The method of claim 2, wherein determining contour vertex coordinates from the candidate contour vertex coordinates based on vertex rectangular features comprises:
traversing each candidate contour vertex coordinate, acquiring a second preset number of contour pixel point coordinates before the current candidate contour vertex coordinate based on the position of the current candidate contour vertex coordinate in the contour pixel point coordinate list, performing straight line fitting to obtain a first straight line function parameter, and acquiring a second preset number of contour pixel point coordinates after the current candidate contour vertex coordinate, and performing straight line fitting to obtain a second straight line function parameter;
calculating a straight line included angle of the vertex coordinates of the current candidate contour by using the first straight line function parameters and the second straight line function parameters to obtain a list of straight line included angles corresponding to the vertex coordinates of each candidate contour;
and screening four candidate contour vertex coordinates with the minimum distance as contour vertex coordinates based on the distance between each straight line included angle and the right-angle included angle in the straight line included angle list.
5. The method according to claim 1, wherein the normalizing the image of the paper to be processed according to the contour vertex coordinates to obtain a normalized paper image comprises:
determining standard vertex coordinates of the paper image to be processed according to the industry standard size of paper contained in the paper image to be processed;
computing a transformation matrix based on the standard vertex coordinates and the contour vertex coordinates;
and standardizing the paper image to be processed by utilizing the transformation matrix to obtain a standardized paper image.
6. The method according to claim 1, wherein the extracting of the paper content area of the paper image to be processed comprises:
acquiring a paper image to be processed, and calling a semantic segmentation network trained in a knowledge distillation mode;
and inputting the paper image to be processed into the semantic segmentation network, and segmenting the paper image to be processed by the semantic segmentation network to obtain a paper content area.
7. The method of claim 1, wherein the extracting pixel coordinates of the contour of the paper content region to obtain a contour pixel coordinate list comprises:
carrying out edge detection after the paper content area is amplified to obtain a pixel coordinate point of the outline of the paper content area;
and sequentially arranging the coordinates of the pixel points clockwise according to the outline to obtain an outline pixel point coordinate list.
8. A sheet image normalization apparatus, comprising:
the extraction module is used for extracting a paper content area of a paper image to be processed and extracting pixel point coordinates of the outline of the paper content area to obtain an outline pixel point coordinate list;
the vertex screening module is used for screening the vertex coordinates of the contour from the contour pixel point coordinate list based on the maximum extreme point characteristics of the vertex distance and the vertex right-angle characteristics;
and the standardization processing module is used for carrying out standardization processing on the paper image to be processed according to the contour vertex coordinates to obtain a standardized paper image.
9. A computer device comprising a processor and a memory, the memory storing a computer program, wherein the processor is configured to implement the paper image normalization method of any one of claims 1-7 when the computer program is executed.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the paper image normalization method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211220060.5A CN115512367A (en) | 2022-10-08 | 2022-10-08 | Paper image standardization method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211220060.5A CN115512367A (en) | 2022-10-08 | 2022-10-08 | Paper image standardization method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115512367A true CN115512367A (en) | 2022-12-23 |
Family
ID=84507868
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211220060.5A Pending CN115512367A (en) | 2022-10-08 | 2022-10-08 | Paper image standardization method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115512367A (en) |
-
2022
- 2022-10-08 CN CN202211220060.5A patent/CN115512367A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2699687C1 (en) | Detecting text fields using neural networks | |
| EP3437019B1 (en) | Optical character recognition in structured documents | |
| WO2018233055A1 (en) | Method and apparatus for entering policy information, computer device and storage medium | |
| KR102399508B1 (en) | Layout analysis method, reading assisting device, circuit and medium | |
| CN113837151B (en) | Table image processing method and device, computer equipment and readable storage medium | |
| CN110502694B (en) | Lawyer recommendation method based on big data analysis and related equipment | |
| CN111061933A (en) | Picture sample library construction method and device, readable storage medium and terminal equipment | |
| CN113850060A (en) | Civil aviation document data identification and entry method and system | |
| CN110503051B (en) | Precious wood identification system and method based on image identification technology | |
| CN110796145B (en) | Multi-certificate segmentation association method and related equipment based on intelligent decision | |
| CN113326887B (en) | Text detection method, device and computer equipment | |
| CN113780116A (en) | Invoice classification method and device, computer equipment and storage medium | |
| Kim et al. | Improving the search accuracy of the VLAD through weighted aggregation of local descriptors | |
| CN111062388B (en) | Advertisement character recognition method, system, medium and equipment based on deep learning | |
| CN113806472A (en) | Method and equipment for realizing full-text retrieval of character, picture and image type scanning piece | |
| CN115512367A (en) | Paper image standardization method and device, computer equipment and storage medium | |
| CN115512369A (en) | Document image layout analysis method and device, computer equipment and storage medium | |
| JP4518212B2 (en) | Image processing apparatus and program | |
| CN111325194B (en) | Character recognition method, device and equipment and storage medium | |
| CN108334884B (en) | Handwritten document retrieval method based on machine learning | |
| CN113705157A (en) | Method for correcting paper work by photographing | |
| JP4517822B2 (en) | Image processing apparatus and program | |
| CN112287763A (en) | Image processing method, apparatus, device and medium | |
| CN110807452A (en) | Prediction model construction method, device and system and bank card number identification method | |
| CN117058699B (en) | Resume layout dividing method, system and storage medium based on LayoutLMv model |
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 |