CN110390642B - Method for geometrically correcting woodcut Tibetan image - Google Patents
Method for geometrically correcting woodcut Tibetan image Download PDFInfo
- Publication number
- CN110390642B CN110390642B CN201810358858.3A CN201810358858A CN110390642B CN 110390642 B CN110390642 B CN 110390642B CN 201810358858 A CN201810358858 A CN 201810358858A CN 110390642 B CN110390642 B CN 110390642B
- Authority
- CN
- China
- Prior art keywords
- points
- image
- sub
- plate frame
- tibetan
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012937 correction Methods 0.000 claims abstract description 43
- 230000009466 transformation Effects 0.000 claims abstract description 21
- 230000011218 segmentation Effects 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000012216 screening Methods 0.000 claims description 4
- 239000013589 supplement Substances 0.000 claims description 3
- 238000011426 transformation method Methods 0.000 abstract description 13
- 230000000694 effects Effects 0.000 abstract description 6
- 239000002023 wood Substances 0.000 abstract description 5
- 238000005452 bending Methods 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 4
- 101100278332 Dictyostelium discoideum dotA gene Proteins 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 101150116173 ver-1 gene Proteins 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000003702 image correction Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000000528 statistical test Methods 0.000 description 1
Classifications
-
- G06T5/80—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Abstract
A method for geometrically correcting a woodcut Tibetan image. A large number of woodcut Tibetan document images are generated in the information age, and geometric correction is the basis of document analysis and character recognition, however, the traditional geometric correction method has difficulty. In this regard, the invention adopts the thinking of short length and simple complexity according to deformation characteristics, proposes the segmentation strategy of geometric correction of the Tibetan document image of the woodcarving plate, divides the image into a plurality of sub-images in the long side direction, and corrects each sub-image by adopting projective transformation. The method solves the problems that the existing projection method, hough transformation method, fourier transformation method, neighbor method and the like which take the detection inclination angle as the main ideas cannot be suitable due to the random bending of the Tibetan image of the wood carving plate, and the existing correction method which takes mathematical modeling as the main ideas is difficult due to the complexity of deformation of each page and the difference between pages, and improves the effect of geometric correction of the Tibetan document image of the wood carving plate.
Description
Technical Field
The invention belongs to the field of image processing, and particularly relates to document image correction.
Background
Aiming at the geometric deformation problem of the image, researchers propose more correction methods. These methods can be classified into rigid deformation correction methods and non-rigid deformation correction methods. The first method considers that the image only has direction change, and the image can be corrected by solving the inclination angle and rotating the image, and the methods comprise a projection method, a Hough transformation method, a Fourier transformation method, a neighbor method and the like. The second method considers that the image has geometric distortion, and maps each pixel position of the deformed image to a non-deformed space by constructing a space function model between the deformed image and the non-deformed image, such as: when the barrel distortion of the electronic endoscope image is processed by the literature (Haneishi H, yagihashi Y, miyake Y A New Method for Distortion Correction of Electronic Endoscope images IEEE Transactions on Medical Imaging, 1995, 14 (3): 548-555), a grid-shaped correction template is adopted, and the relationship between the distorted image and an ideal image is determined by examining the change of the positions of grid nodes, so that the distortion correction of the image is realized; the literature (Ai S, wang X, ma M, et al A method for Correcting Non-linear Geometric Distortion in Ultra-wide-angle Imaging system, optik-International Journal for Light and Electron Optics, 2013, 124 (24): 7014-7021) corrects radial distortion using the cross-ratio invariance of the center point and collinear points of a planar grid template image and the obtained distortion parameters when dealing with nonlinear geometric distortion of panoramic field images obtained in large-tilt ultra-wide angle systems, and corrects perspective distortion using perspective transformation.
In the first type of method, the principle is relatively simple, but is very effective only when the entire image has the same rotation fulcrum and rotation angle. The woodcut Tibetan document image is in a strip shape, and the shape is not changed only in direction, but also has geometric distortion, so that the method is not suitable for adopting the first type of method. In the second method, the correction accuracy directly depends on the accuracy of the mathematical model, however, tibetan document images have defects of difficult modeling due to complex deformation and different deformation of each page.
Disclosure of Invention
Aiming at the problems, the invention adopts the thinking of short length and simple complexity according to the deformation characteristics of the woodcut Tibetan document image, and provides a segmentation method for geometric correction of the woodcut Tibetan document image, which is used for segmenting the image into a plurality of sub-images in the long side direction and correcting each sub-image by adopting projective transformation.
The technical scheme of the invention is as follows:
step 1: candidate points on the upper and lower frames are acquired. Dividing the image into a plurality of sub-images by using vertical lines according to a set number, setting an observation window with a certain height and the same width as the sub-images in each sub-image, enabling the bottom edge of the observation window to move upwards row by row from the middle row of each sub-image, counting the proportion of background pixels in the window, and judging that the row above the observation window and the column at the center of the observation window are the positions of candidate points on the upper frame line at the moment once the proportion is higher than the set threshold. After all the sub-images are processed in this way, all the candidate points on the upper frame line can be obtained. Similarly, all candidate points on the lower plate frame line can be determined.
Step 2: and screening candidate points on the upper and lower frames. A preferred method is to calculate the maximum and minimum values of the ordinate of all candidate points on the upper plate frame line, and remove candidate points below the mean of the maximum and minimum values; and calculating the maximum value and the minimum value of the ordinate of all candidate points on the lower plate frame line, and removing the candidate points higher than the average value of the maximum value and the minimum value.
Step 3: and (5) filling up points on the upper and lower plate frame lines. Because the proportion of background pixels in the viewing window may not reach the set threshold and some candidate points are removed, the points on the upper (lower) plate frame line obtained do not necessarily have points on the lower (upper) plate frame line corresponding thereto, which requires replenishment of the points on the upper and lower plate frame lines. In this regard, a preferred method is to supplement the missing points using a linear interpolation method.
Step 4: segmentation is based on points on the upper and lower plate frame lines. The original image is segmented into a plurality of sub-images by taking the connecting lines of adjacent 2 pairs of points as left and right boundaries, wherein the sub-images at the two end edges also comprise the outer images of the points at the end edges.
Step 5: a projective transformation matrix of each sub-image is calculated. The calculation of the projective transformation matrix requires the positions of 4 points before transformation and 4 points after transformation, and the points used to calculate the projective transformation matrix in the present invention are determined as follows: in each sub-image, the 2 points used to segment the image are the 4 points before correction required to calculate the projective transformation matrix; the 4 points are in the same positions after correction as the corresponding points before correction on the abscissa and the preferable method is that the points on the upper (lower) plate frame line are all the same as the first points on the upper (lower) plate frame line after correctionround(Num2) the same points, whereinround(.) means rounding to an integer,Numrepresenting the number of points acquired on the upper (lower) plate frame line.
Step 6: based on the matrix of step 5, each sub-image is geometrically corrected by projective transformation.
Step 7: pixel values for the positions of points on the geometrically corrected image are calculated.
The invention has the following technical effects or advantages: the invention solves the problems that the existing projection method, hough transformation method, fourier transformation method, neighbor method and the like which take the detection inclination angle as the main idea cannot be suitable due to the irregular bending of the Tibetan image of the wood carving plate, and the existing correction method which takes mathematical modeling as the main idea is difficult to be powerful due to the complexity of deformation of each page and the difference between pages, and improves the effect of geometric correction of the Tibetan document image of the wood carving plate.
Drawings
Fig. 1 is a schematic view showing a line-by-line movement of an observation window for obtaining candidate points on an upper frame line, fig. 4 is a schematic view showing candidate points determined in the obtaining of candidate points on an upper frame line, fig. 5 is a schematic view showing the screening of candidate points on an upper frame and a lower frame line, fig. 6 is a schematic view showing the addition of a part of points, fig. 7 is a schematic view showing the positions of points required for segmentation and calculation of a projective transformation matrix, fig. 8 is a correction result (wherein a dotted rectangular frame is used for reference) of a Hough transformation method for an example of a woodcut Tibetan document image, fig. 9 is a correction result (wherein a dotted rectangular frame is used for reference) of a Fourier transformation method for an example of a woodcut Tibetan document image, fig. 10 is a correction result (wherein a dotted rectangular frame is used for reference) of a method for an example of a woodcut Tibetan document image, fig. 11 is a correction result (wherein a dotted rectangular frame is used for a rectangular frame is used for reference) of a method for an example of a woodcut Tibetan document image, and fig. 8 is a correction result (wherein a dotted rectangular frame is used for a dotted frame is used for a square frame is used for a reference) of a method for a correction result (wherein a dotted frame is used for a two-dimensional transformation method is used for a fig. 13).
Detailed Description
The invention is described below in connection with specific embodiments.
Taking a certain two-page image of Tarsque woodcut in the book of Buddha's second edition as an example, the deformation diversity is shown: in fig. 1, on the basis of the oblique deformation, the left end is slightly narrow and sags in the long side direction, and the right end is slightly wide and warps upwards; in fig. 2, the deformation is in the long and short directions, particularly the long side, and the deformation is mainly in a bent shape in the long side direction. These are also common phenomena in the woodcut Tibetan literature. In this regard, the correction procedure using the method of the invention is as follows:
step 1: candidate points on the upper and lower frames are acquired. The image is equally divided into 10 sub-images by using vertical lines, in each sub-image, an observation window with a certain height and the same width as the sub-image is arranged, the height of the observation window is 0.1 times of the height of the sub-image, the bottom edge of the observation window moves upwards from the middle line of each sub-image line by line, and the proportion of background pixels in the window is counted, once the proportion is higher than 0.99, the position of the candidate points on the upper frame line, where the line above the observation window and the center of the observation window are located, is judged, as shown in fig. 3 and 4, wherein a black rectangular frame is an observation frame. After all the sub-images are processed in this way, all the candidate points on the upper frame line can be obtained. Similarly, all candidate points on the lower plate frame line can be determined.
Step 2: and screening candidate points on the upper and lower frames. Calculating maximum values and minimum values of the ordinate of all candidate points on the upper frame line, and removing the candidate points lower than the average value of the maximum values and the minimum values; and calculating the maximum value and the minimum value of the ordinate of all candidate points on the lower plate frame line, and removing the candidate points higher than the average value of the maximum value and the minimum value. In fig. 5, filled circles and triangles represent points on the upper and lower frame lines obtained in step 1, respectively, open circles and triangles represent some candidate points removed on the upper and lower frame lines in the process, which are located at leftmost and rightmost ends on the upper and lower frame lines, and 7 th candidate point positions of the upper frame line from left to right, respectively, which are removed because a large number of background pixels outside the left and right frame lines or excessive blank areas in the frame lines are contained in the observation window where they are located.
Step 3: and (5) filling up points on the upper and lower plate frame lines. Because the proportion of background pixels in the observation window may not reach the set threshold and some candidate points are removed, in the same sub-image, the upper (lower) plate frame line obtained by the previous section isThe points on the lower (upper) plate frame line are not necessarily corresponding to the points, which affects the positional relationship between the points required for computing the projective transformation matrix in step 5, as shown in FIG. 6DAndErespectively lacks pointsCAndFcorrespondingly, in fig. 6, the filled circles and open circles represent the points on the plate frame line that have been selected and the points to be filled in, respectively. The present embodiment supplements the missing points by using a linear interpolation method. In the example of FIG. 6, dotsCAndFaccording to the positions of (3)AAndE、BandDdetermination by linear interpolation, recordingA、B、DAndEthe positions of (2) are respectively、/>、/>And->Point thenCAndFthe positions of (2) are +.>And->. Thus far, the points on the acquired upper and lower plate frame lines all appear in pairs, each pair being located in the same column.
Step 4: segmentation is based on points on the upper and lower plate frame lines. The original image is segmented into a plurality of sub-images by taking the connecting lines of adjacent 2 pairs of points as left and right boundaries, wherein the sub-images at the two end edges also comprise the outer images of the points at the end edges. As shown in fig. 6, dotsAAndBpoint(s)CAndDpoint(s)EAndFeach being a pair of points, adjacent linesABAnd (3) withCDBetween, lineCDAnd (3) withEFEach of which is a sub-image, the latter being the most edgeEdge sub-images, so also including linesEFThe image on the right end results in an image segmentation result as shown in fig. 7, where each two adjacent vertical dashed lines are followed by a sub-image.
Step 5: a projective transformation matrix of each sub-image is calculated. Wherein the points used to calculate the projective transformation matrix are determined as follows: in each sub-image, the 2 points used to segment the image are the 4 points before correction required to calculate the projective transformation matrix; the 4 points are at the same positions after correction as the corresponding points before correction on the abscissa, and the points on the upper (lower) plate frame line are all the same as the first points before correction on the upper (lower) plate frame line on the ordinateround(Num2) the same points, whereinround(.) means rounding to an integer,Numrepresenting the number of points acquired on the upper (lower) plate frame line. As shown in fig. 7, the black dots are dots before correction, and the black rectangular dots are dots after correction, wherein: on the abscissa, a dotA’、B’、C’、D’、E’AndF’respectively and point toA、B、C、D、E、FThe same; on the ordinate, a dotA’、C’、E’The first position before correction on the upper plate frameround(Num2) the same pointsB’、D’、F’The first position before correction on the upper plate frameround(NumAnd/2) the points are identical.
Step 6: based on the matrix of step 5, each sub-image is geometrically corrected by projective transformation.
Step 7: and obtaining pixel values of the positions of each point on the geometrically corrected image by bilinear interpolation.
Fig. 8-13 show the comparison results of the method of the present invention and the conventional Hough transform method and Fourier transform method after correction, wherein the dotted line is a horizontally placed rectangular frame, the geometric correction effect can be judged by observing the change of the distance between the rectangular frame and the plate frame line, and the more uniform the change of the distance, the better the correction effect is. By comparison, the method of the invention has the optimal effect, because the segmentation strategy is adopted, each sub-image is corrected by using projective transformation and seamless splicing can be realized, and compared with the existing Hough transformation method and Fourier transformation method which adopt integral inclination correction, the method has the defect of 'pressing the head but tilting the head'.
In terms of time complexity, statistical tests are performed on all images (19 volumes in total, 17198 pages) of the second edition of the Buddha's canal (Taerte wood engraving) made by the pectoral karman in the Tibetan Buddha resource center (Tibetan Buddhist Resource Center) website, and the results show that the method has the advantages in terms of time complexity compared with the traditional Hough transformation method and Fourier transformation method although the images are segmented and time resources are consumed in the segmentation process.
The foregoing describes one embodiment of the present invention in detail, but the description is only a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications within the scope of the present invention are intended to be covered by the present invention.
Claims (2)
1. A method for geometrically correcting a woodcut Tibetan image is characterized by comprising the following steps:
step 1, obtaining candidate points on an upper plate frame and a lower plate frame: dividing an image into a plurality of sub-images by using vertical lines according to a set number, setting an observation window with a certain height and the same width as the sub-images in each sub-image, enabling the bottom edge of the observation window to move upwards line by line from the middle line of each sub-image, counting the proportion of background pixels in the window, judging that the line above the observation window and the line at the center of the observation window are the positions of candidate points on the upper frame line at the moment once the proportion is higher than a set threshold, and obtaining all the candidate points on the upper frame line after all the sub-images are processed in the way; similarly, all candidate points on the lower plate frame line are judged;
step 2, screening candidate points on the upper and lower plate frames;
step 3, filling points on the upper and lower plate frame lines;
step 4, segmentation is carried out based on points on upper and lower plate frame lines: segmenting the original image into a plurality of sub-images by taking the connecting lines of adjacent 2 pairs of points as left and right boundaries, wherein the sub-images at two ends also comprise the outer side images of the point at the most edge;
step 5, calculating a projective transformation matrix of each sub-image, wherein the points used for calculating the projective transformation matrix are determined as follows: in each sub-image, the 2 points used to segment the image are the 4 points before correction required to calculate the projective transformation matrix; the 4 points are at the corrected positions, on the abscissa, the positions are the same as the corresponding points before correction, on the ordinate, the points on the upper plate frame line are the same as the first points before correction on the upper plate frame lineround(NumAnd 2) the points are the same, and the points on the lower plate frame line are the same as the first point on the lower plate frame line before correctionround(Num2) the same points, whereinround(.) means rounding to an integer,Numrepresenting the number of points acquired on the upper or lower plate frame line;
step 6, based on the matrix in step 5, performing geometric correction on each sub-image by adopting projective transformation;
and 7, calculating pixel values of positions of each point on the geometrically corrected image.
2. A method of geometrically correcting a woodcut Tibetan image as in claim 1, wherein:
in the step 2, on the upper frame line, calculating the maximum value and the minimum value of the ordinate of all candidate points, and removing the candidate points lower than the average value of the maximum value and the minimum value; calculating maximum values and minimum values of the longitudinal coordinates of all candidate points on a lower plate frame line, and removing the candidate points higher than the average value of the maximum values and the minimum values;
in the step 3, a linear interpolation method is adopted to supplement the missing points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810358858.3A CN110390642B (en) | 2018-04-20 | 2018-04-20 | Method for geometrically correcting woodcut Tibetan image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810358858.3A CN110390642B (en) | 2018-04-20 | 2018-04-20 | Method for geometrically correcting woodcut Tibetan image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110390642A CN110390642A (en) | 2019-10-29 |
| CN110390642B true CN110390642B (en) | 2023-12-08 |
Family
ID=68283666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810358858.3A Active CN110390642B (en) | 2018-04-20 | 2018-04-20 | Method for geometrically correcting woodcut Tibetan image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110390642B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111223065B (en) * | 2020-01-13 | 2023-08-01 | 中国科学院重庆绿色智能技术研究院 | Image correction method, irregular text recognition device, storage medium and apparatus |
| CN111915524B (en) * | 2020-08-04 | 2024-03-29 | 深圳企业云科技股份有限公司 | Full-automatic image perspective correction method based on cross ratio operation |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005267457A (en) * | 2004-03-19 | 2005-09-29 | Casio Comput Co Ltd | Image processing device, imaging apparatus, image processing method and program |
| CN101398895A (en) * | 2007-09-26 | 2009-04-01 | 杨高波 | Image preprocess method based on data matrix two-dimension bar code identification |
| CN101630360A (en) * | 2008-07-14 | 2010-01-20 | 上海分维智能科技有限公司 | Method for identifying license plate in high-definition image |
| JP2010171976A (en) * | 2009-01-22 | 2010-08-05 | Canon Inc | Method and system for correcting distorted document image |
| JP2011054068A (en) * | 2009-09-04 | 2011-03-17 | Fuji Xerox Co Ltd | Image processing system, image processor, and program |
| JP2011059841A (en) * | 2009-09-08 | 2011-03-24 | Nikon Corp | Imaging apparatus |
| JP2011159300A (en) * | 2011-02-24 | 2011-08-18 | Nintendo Co Ltd | Image processing program, image processing device, image processing system, and image processing method |
| JP2012019474A (en) * | 2010-07-09 | 2012-01-26 | Oki Data Corp | Image reading device and image correction program |
| CN103413271A (en) * | 2013-07-18 | 2013-11-27 | 西安交通大学 | Document image rectifying method based on local information |
| CN104079907A (en) * | 2013-03-27 | 2014-10-01 | 精工爱普生株式会社 | Projector, image correction method, and program |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6798541B2 (en) * | 2001-06-29 | 2004-09-28 | Xerox Corporation | Systems and methods for generating binary clustered, irrational halftone dots |
| US7239331B2 (en) * | 2004-02-17 | 2007-07-03 | Corel Corporation | Method and apparatus for correction of perspective distortion |
| CN104657730B (en) * | 2013-11-20 | 2018-01-05 | 富士通株式会社 | Means for correcting, method and the scanner of file and picture |
-
2018
- 2018-04-20 CN CN201810358858.3A patent/CN110390642B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005267457A (en) * | 2004-03-19 | 2005-09-29 | Casio Comput Co Ltd | Image processing device, imaging apparatus, image processing method and program |
| CN101398895A (en) * | 2007-09-26 | 2009-04-01 | 杨高波 | Image preprocess method based on data matrix two-dimension bar code identification |
| CN101630360A (en) * | 2008-07-14 | 2010-01-20 | 上海分维智能科技有限公司 | Method for identifying license plate in high-definition image |
| JP2010171976A (en) * | 2009-01-22 | 2010-08-05 | Canon Inc | Method and system for correcting distorted document image |
| JP2011054068A (en) * | 2009-09-04 | 2011-03-17 | Fuji Xerox Co Ltd | Image processing system, image processor, and program |
| JP2011059841A (en) * | 2009-09-08 | 2011-03-24 | Nikon Corp | Imaging apparatus |
| JP2012019474A (en) * | 2010-07-09 | 2012-01-26 | Oki Data Corp | Image reading device and image correction program |
| JP2011159300A (en) * | 2011-02-24 | 2011-08-18 | Nintendo Co Ltd | Image processing program, image processing device, image processing system, and image processing method |
| CN104079907A (en) * | 2013-03-27 | 2014-10-01 | 精工爱普生株式会社 | Projector, image correction method, and program |
| CN103413271A (en) * | 2013-07-18 | 2013-11-27 | 西安交通大学 | Document image rectifying method based on local information |
Non-Patent Citations (1)
| Title |
|---|
| 段立娟 等.Text extraction method for historical Tibetan document images based on block projections.OPTOELECTRONICS LETTERS.第13卷(第13期),0457-0461. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110390642A (en) | 2019-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109086714B (en) | Form recognition method, recognition system and computer device | |
| CN102790841B (en) | Method of detecting and correcting digital images of books in the book spine area | |
| US8472726B2 (en) | Document comparison and analysis | |
| US8472727B2 (en) | Document comparison and analysis for improved OCR | |
| JP5082637B2 (en) | Image processing program, image processing method, and image processing apparatus | |
| CN114529459B (en) | Method, system and medium for enhancing image edge | |
| CN110647795B (en) | Form identification method | |
| CN115170669B (en) | Identification and positioning method and system based on edge feature point set registration and storage medium | |
| KR20020018936A (en) | Apparatus and method for correcting distortion of input image | |
| CN111353961B (en) | Document curved surface correction method and device | |
| CN110400278B (en) | Full-automatic correction method, device and equipment for image color and geometric distortion | |
| WO2019056346A1 (en) | Method and device for correcting tilted text image using expansion method | |
| CN114529925B (en) | Method for identifying table structure of whole line table | |
| CN110390642B (en) | Method for geometrically correcting woodcut Tibetan image | |
| CN113033558B (en) | Text detection method and device for natural scene and storage medium | |
| CN111145124A (en) | Image tilt correction method and device | |
| Liu et al. | End-to-end trainable video super-resolution based on a new mechanism for implicit motion estimation and compensation | |
| CN102737240A (en) | Method of analyzing digital document images | |
| CN108197624A (en) | The recognition methods of certificate image rectification and device, computer storage media | |
| CN115273114A (en) | Document image optimization method and medium | |
| CN115082935A (en) | Method, apparatus and storage medium for correcting document image | |
| CN113065404A (en) | Method and system for detecting train ticket content based on equal-width character segments | |
| CN117095417A (en) | Screen shot form image text recognition method, device, equipment and storage medium | |
| CN108335266A (en) | A kind of antidote of file and picture distortion | |
| CN115170525A (en) | Image difference detection method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |