CN101520852B - Vanishing point detecting device and detecting method - Google Patents
Vanishing point detecting device and detecting method Download PDFInfo
- Publication number
- CN101520852B CN101520852B CN2008100809519A CN200810080951A CN101520852B CN 101520852 B CN101520852 B CN 101520852B CN 2008100809519 A CN2008100809519 A CN 2008100809519A CN 200810080951 A CN200810080951 A CN 200810080951A CN 101520852 B CN101520852 B CN 101520852B
- Authority
- CN
- China
- Prior art keywords
- candidate
- fladellum
- point
- projection
- unit
- 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.)
- Expired - Fee Related
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a vanishing point detecting device and a detecting method, and the vanishing point detecting device comprises a line segment detecting unit used for obtaining a horizontal line segment or a vertical line segment of an input image, an intersection point set obtaining device used for obtaining the intersection point set of the horizontal line segment or an intersection point set of the vertical line segment, a candidate vanishing point detecting unit used for determining a candidate vanishing point, and a vanishing point determining unit used for determining a final vanishing point according to the candidate vanishing point determined by the candidate vanishing point detecting unit; wherein the candidate vanishing point detecting unit is one or more of the following: acandidate vanishing point detecting unit based on fanning beam projection and density estimation, which carries out fanning beam projection to the input image and takes the intersection point with th e highest point density in a high-density fanning beam area as the candidate vanishing point; a candidate vanishing point detecting unit based on fanning beam projection and clustering, which carries out the fanning beam projection, obtains a plurality of clustering of the intersection points in the high-density fanning beam area and takes the center of each clustering as the candidate vanishing point.
Description
Technical field
The present invention relates to apparatus and method that the perspective distortion of file and picture (image that particularly contains a small amount of text) is proofreaied and correct.More specifically, relate to based on the character horizontal stroke and detect the method and apparatus of end point and utilize the fladellum projection to detect the method and apparatus of end point.
Background technology
It is an important step of document analysis and identification that file and picture based on digital camera is carried out the perspective distortion correction.The implication of perspective distortion correct operation is the image that the image with perspective distortion is changed into no perspective distortion.
The principle that various file and picture perspective distortions are proofreaied and correct can be referring to P.Clark, M.Mirmehdi is at Pattern Recognition 36 (11), 2003 article " Rectifying perspective views oftext in 3D scenes using vanishing points. ", M.Pilu is at the article " Extract of illusory linear clues in perspectively skewed documents " of international conference IEEEConference on Computer Vision and Pattern Recognition 2001, and S.J.Lu, B.M.Chen, C.C.Ko is at Image and Vision Computing 23 (5), 2005 article " Perspective rectification of document images using fuzzy set andmorphological operations. " in addition, bearing calibration also has patent to deliver about the file and picture perspective distortion, as United States Patent (USP) 6,873,732, C.R.Dance, " Method and apparatus forresolving perspective distortion in a document image and for calculating
Generally the file and picture that obtains by digital camera, the first-class shooting of shooting has perspective distortion, can influence follow-up relevant document analysis and identification.In order to proofread and correct these perspective distortions, some different file and picture perspective distortion correcting algorithms have been proposed, these algorithms have certain effect.But, because the different angles the when different model of digital camera and camera and shooting, perspective distortion situation very complicated (it is especially true that mobile phone cam is taken the file and picture that obtains), and a lot of situation hypographs only contain a small amount of text (as signboard image, restaurant menus image etc.), and present correcting algorithm still can not be dealt with problems preferably.
Summary of the invention
The above-mentioned shortcoming that the present invention is directed to prior art is made.The present invention aims to provide candidate's end point (comprising horizontal end point and vertical end point) pick-up unit and the method at a small amount of text.On the other hand, the present invention aims to provide a kind of candidate's vanishing Point Detection Method method that is distributed in a small amount of text of one or several main direction at character stroke (horizontal stroke or vertical stroke).On the other hand, the present invention aims to provide a kind of at a small amount of text, and most of character stroke is distributed in a main direction, and presents candidate's vanishing Point Detection Method method of the main direction of several height in the zone at main direction place.On the other hand, the present invention has been intended to propose a kind of vanishing Point Detection Method method of carrying out vanishing Point Detection Method in conjunction with multiple a small amount of text situation.
For realizing above purpose, the application provides following invention.
Invention 1, a kind of vanishing point detecting device, described vanishing point detecting device detects the end point of file and picture, and described vanishing point detecting device comprises:
The line segment detecting unit is used to the horizontal line section or the vertical line segment of the image that obtains to import;
Intersection point set obtains the unit, is used to obtain the set of the intersection point of the set of intersection point of the horizontal line section that described line segment detecting unit obtained or vertical line segment;
Candidate's end point according to the set that described intersection point is gathered the intersection point that obtains the unit acquisition, is determined in candidate's vanishing Point Detection Method unit;
The end point determining unit according to candidate's end point that candidate's vanishing Point Detection Method unit is determined, is determined final end point,
Wherein, described candidate's vanishing Point Detection Method unit is in one in the following candidate's vanishing Point Detection Method unit or more how:
Based on candidate's vanishing Point Detection Method unit of cluster, set obtains the intersection point that the unit obtains and carries out cluster to described intersection point, obtains a plurality of clusters, with the central point of each cluster as candidate's end point;
Based on candidate's vanishing Point Detection Method unit of fladellum projection and density Estimation, the image of described input is carried out the fladellum projection, the dot density of the intersection point in the high density fan-beam district is estimated, with the intersection point of dot density maximum as candidate's end point;
Based on candidate's vanishing Point Detection Method unit of fladellum projection and cluster, the image of described input is carried out the fladellum projection, the intersection point in the high density fan-beam district is carried out cluster, obtain a plurality of clusters, with the center of each cluster as candidate's end point.
The invention 2, according to the invention 1 described vanishing point detecting device, it is characterized in that, described line segment detecting unit comprises the character stroke detecting unit, and described character stroke detecting unit detects the vertical stroke line segment of described file and picture or horizontal stroke line segment as described vertical line segment or horizontal line section.
The invention 3, according to the invention 1 described vanishing point detecting device, it is characterized in that described character stroke detecting unit comprises:
The connected domain computing unit obtains the connected domain of the image of described input;
The stroke detecting unit, the shape and the size of the connected domain that described connected domain computing unit is obtained are analyzed, and obtain candidate's horizontal stroke or candidate's vertical stroke;
The stroke segmental determining unit according to candidate's horizontal stroke or the candidate's vertical stroke that described stroke detecting unit is obtained, is determined horizontal stroke line segment or vertical stroke line segment.
Invention 4, according to invention 3 described vanishing point detecting devices, it is characterized in that described stroke detecting unit selects width and character duration is close and its direction is close with horizontal direction connected domain as horizontal stroke.
Invention 5, according to invention 3 described vanishing point detecting devices, it is characterized in that described stroke detecting unit selects height is close with character height and its direction is close with vertical direction connected domain as vertical stroke.
The invention 6, according to the invention 1 described vanishing point detecting device, it is characterized in that described candidate's vanishing Point Detection Method unit based on fladellum projection and density Estimation comprises:
Intersection point collection fladellum projecting cell carries out the fladellum projection to the image of described input, obtains the intersection point set that the unit obtained according to described intersection point set, obtains the projection value of each fladellum;
The fladellum selected cell according to the projection value of each fladellum, is selected the fladellum of projection value maximum;
The dot density computing unit calculates the density of the point in the selected fladellum of described fladellum selected cell;
Candidate's end point selected cell, the point of selecting the density maximum that described dot density computing unit calculates is as candidate's end point.
The invention 7, according to the invention 1 described vanishing point detecting device, it is characterized in that described candidate's vanishing Point Detection Method unit based on fladellum projection and cluster comprises:
Intersection point collection fladellum projecting cell carries out the fladellum projection to the image of described input, obtains the intersection point set that the unit obtained according to described intersection point set, obtains the projection value of each fladellum;
The fladellum selected cell according to the projection value of each fladellum, is selected the fladellum of projection value maximum;
Cluster cell, the fladellum selected to described fladellum selected cell carries out cluster, obtains one or more cluster;
Candidate's end point selected cell, the center of selecting each cluster is as candidate's end point.
The invention 8, according to the invention 1 described vanishing point detecting device, it is characterized in that it is characterized in that, described end point determining unit comprises:
First weight-coefficient calculating unit obtains first weight coefficient of each candidate's end point;
Second weight-coefficient calculating unit obtains second weight coefficient of each candidate's end point;
The end point analytic unit according to described first weight coefficient and described second weight coefficient of each described candidate's end point, is selected final end point.
Invention 9, a kind of vanishing Point Detection Method method, described vanishing Point Detection Method method detects the end point of file and picture, and described vanishing Point Detection Method method comprises:
Line segment detects step, is used to the horizontal line section or the vertical line segment of the image that obtains to import;
The intersection point set obtains step, is used to obtain set or the vertically set of the intersection point of line segment that described line segment detects the intersection point of the horizontal line section that step obtained;
Candidate's vanishing Point Detection Method step according to the set that described intersection point is gathered the intersection point that obtains the step acquisition, is determined candidate's end point;
The end point determining step according to candidate's end point that candidate's vanishing Point Detection Method step is determined, is determined final end point,
Wherein, described candidate's vanishing Point Detection Method step adopts one or more kinds in following candidate's vanishing Point Detection Method method:
Based on candidate's vanishing Point Detection Method method of cluster, set obtains the intersection point that step obtains and carries out cluster to described intersection point, obtains a plurality of clusters, with the central point of each cluster as candidate's end point;
Based on candidate's vanishing Point Detection Method method of fladellum projection and density Estimation, the image of described input is carried out the fladellum projection, the dot density of the intersection point in the high density fan-beam district is estimated, with the intersection point of dot density maximum as candidate's end point;
Based on candidate's vanishing Point Detection Method method of fladellum projection and cluster, the image of described input is carried out the fladellum projection, the intersection point in the high density fan-beam district is carried out cluster, obtain a plurality of clusters, with the center of each cluster as candidate's end point.
Intersection point collection fladellum projection step is carried out the fladellum projection to the image of described input, obtains the intersection point set that step obtained according to described intersection point set, obtains the projection value of each fladellum;
Fladellum is selected step, according to the projection value of each fladellum, selects the fladellum of projection value maximum;
The dot density calculation procedure is calculated the density that described fladellum is selected the point in the selected fladellum of step;
Candidate's end point is selected step, and the point of selecting the density maximum that described dot density calculation procedure calculates is as candidate's end point,
And wherein,
Described candidate's vanishing Point Detection Method step comprises when the candidate's vanishing Point Detection Method method that adopts based on fladellum projection and cluster:
Intersection point collection fladellum projection step is carried out the fladellum projection to the image of described input, obtains the intersection point set that step obtained according to described intersection point set, obtains the projection value of each fladellum;
Fladellum is selected step, according to the projection value of each fladellum, selects the fladellum of projection value maximum;
The cluster step selects the intersection point in the selected fladellum of step to carry out cluster to described fladellum, obtains one or more cluster;
Candidate's end point is selected step, and the center of selecting each cluster is as candidate's end point.
The invention 11, according to the invention 9 described vanishing Point Detection Method methods, it is characterized in that, described line segment detects step and comprises that character stroke detects step, and described character stroke detects step and detects the vertical stroke line segment of described file and picture or horizontal stroke line segment as described vertical line segment or horizontal line section.
Invention 12, according to invention 11 described vanishing Point Detection Method methods, it is characterized in that described character stroke detects step and comprises:
The connected domain calculation procedure obtains the connected domain of the image of described input;
Stroke detects step, and the shape and the size of the connected domain that described connected domain calculation procedure is obtained are analyzed, and obtain candidate's horizontal stroke or candidate's vertical stroke;
The stroke segmental determining step detects candidate's horizontal stroke or the candidate's vertical stroke that step obtained according to described stroke, determines horizontal stroke line segment or vertical stroke line segment.
Invention 13, according to invention 12 described vanishing Point Detection Method methods, it is characterized in that described stroke detects step and selects width and character duration is close and its direction is close with horizontal direction connected domain as horizontal stroke.
Invention 14, according to invention 12 described vanishing point detecting devices, it is characterized in that described stroke detects step and selects height is close with character height and its direction is close with vertical direction connected domain as vertical stroke.
In addition, the present invention also provides a kind of computer program, and described program can be carried out to realize above-mentioned vanishing Point Detection Method method of the present invention by computing machine.
In addition, the present invention also provides a kind of computer program, and described program can and make computing machine be used as the present invention above-mentioned various devices or unit by the computing machine execution, comprises described vanishing point detecting device etc.
The invention provides a kind of data storage medium more on the one hand according to of the present invention, described data storage medium stores above-mentioned computer program.Described storage medium can be any storage medium that those skilled in the art can know, as ROM, floppy disk, flash memory, hard disk, CD, DVD, tape, magneto-optic disk or the like.
Though in foregoing description of the present invention, each step is described in order, the order of these steps can be adjusted, also can executed in parallel.
Method of the present invention has overcome the problem of a large amount of text statistical informations of classic method needs, simultaneously also overcome the instability shortcoming of utilizing single method calculated candidate end point, can proofread and correct at the perspective distortion document image that contains a small amount of text effectively.
Description of drawings
The accompanying drawing that is comprised is used to provide to further understanding of the present invention, and it is merged in instructions and constitutes its part, description of drawings embodiments of the invention, and be used from instructions one and explain principle of the present invention.
Fig. 1 is the structural representation of means for correcting of the perspective distortion document image of one embodiment of the present invention.
Fig. 2 is the schematic flow diagram of bearing calibration of the perspective distortion document image of one embodiment of the present invention.
Fig. 3 is based on horizontal candidate's vanishing Point Detection Method iconicity explanation of cluster.
Fig. 4 is based on horizontal candidate's vanishing Point Detection Method explanation of fladellum projection and density Estimation.
Fig. 5 is the synoptic diagram based on horizontal candidate's vanishing point detecting device of fladellum projection and density Estimation of one embodiment of the present invention.
Fig. 6 is the process flow diagram based on horizontal candidate's vanishing Point Detection Method method of fladellum projection and density Estimation of one embodiment of the present invention.
Fig. 7 is the iconicity explanation based on horizontal candidate's vanishing Point Detection Method of fladellum projection and cluster of one embodiment of the present invention.
Fig. 8 is the horizontal candidate's vanishing point detecting device synoptic diagram based on fladellum projection and cluster of one embodiment of the present invention.
Fig. 9 is the process flow diagram based on horizontal candidate's vanishing Point Detection Method of fladellum projection and cluster of one embodiment of the present invention.
Figure 10 is the structural representation of the character horizontal stroke detecting unit of one embodiment of the present invention.
Figure 11 is the process flow diagram that the character horizontal stroke of one embodiment of the present invention detects.
Embodiment
Describe the means for correcting and the method for perspective distortion document image of the present invention below with reference to accompanying drawings in detail.
Fig. 1 is the structural drawing according to the means for correcting of the perspective distortion document image of one embodiment of the present invention.As shown in Figure 1, the means for correcting according to perspective distortion document image of the present invention comprises: perspective distortion document image input block 101, edge calculations unit 102, character horizontal stroke detecting unit 103, horizontal candidate's vanishing Point Detection Method unit 104 based on cluster, horizontal candidate's vanishing Point Detection Method unit 105 based on fladellum projection and density Estimation, horizontal candidate's vanishing Point Detection Method unit 106 based on fladellum projection and cluster, horizontal end point selected cell 107, character vertical stroke detecting unit 108, vertical candidate's vanishing Point Detection Method unit 109 based on cluster, vertical candidate's vanishing Point Detection Method unit 110 based on fladellum projection and density Estimation, vertical candidate's vanishing Point Detection Method unit 111 based on fladellum projection and cluster, vertical end point selected cell 112, perspective distortion is corrected converter unit 113, and image output unit 114 after correcting.
Basic identical based on horizontal candidate's vanishing Point Detection Method unit 104 of cluster with structure based on vertical candidate's vanishing Point Detection Method unit 109 of cluster, corresponding to the candidate's vanishing Point Detection Method unit based on cluster of the present invention.Basic identical based on horizontal candidate's vanishing Point Detection Method unit 105 of fladellum projection and density Estimation with structure based on vertical candidate's vanishing Point Detection Method unit 110 of fladellum projection and density Estimation, corresponding to the candidate's vanishing Point Detection Method unit based on fladellum projection and density Estimation of the present invention.Based on horizontal candidate's vanishing Point Detection Method unit 106 of fladellum projection and cluster and basic identical, corresponding to the candidate's vanishing Point Detection Method unit based on fladellum projection and cluster of the present invention based on the structure of vertical candidate's vanishing Point Detection Method unit 111 of fladellum projection and cluster.Horizontal candidate's vanishing Point Detection Method unit 104 based on cluster, horizontal candidate's vanishing Point Detection Method unit 105 based on fladellum projection and density Estimation, horizontal candidate's vanishing Point Detection Method unit 106 based on fladellum projection and cluster, vertical candidate's vanishing Point Detection Method unit 109 based on cluster, based on vertical candidate's vanishing Point Detection Method unit 110 of fladellum projection and density Estimation and based in vertical candidate's vanishing Point Detection Method unit 111 of fladellum projection and cluster one or more corresponding to candidate's vanishing Point Detection Method of the present invention unit.When need not to distinguish, candidate's vanishing Point Detection Method unit can refer to one or more in them.
The structure of character horizontal stroke detecting unit 103 and character vertical stroke detecting unit 108 is basic identical, corresponding to character stroke detecting unit of the present invention.
Horizontal end point selected cell 107 is basic identical with the structure of vertical end point selected cell 112, corresponding to end point determining unit of the present invention.
Physically, the essentially identical unit of above structure can be realized lumpedly by same parts, also can realize discretely.
In addition, though for the convenience that illustrates, in the means for correcting of the perspective distortion document image shown in Fig. 1, comprised perspective modification file and picture input block 101, edge calculations unit 102, perspective modification rectification converter unit 113 and corrected back image output unit 114, but these unit can be realized in other device.That is, the means for correcting of perspective distortion image of the present invention can not comprise these unit.
Fig. 2 is the structural drawing according to the bearing calibration of the perspective distortion document image of one embodiment of the present invention.As shown in Figure 2, the bearing calibration according to perspective distortion document image of the present invention comprises: perspective distortion document image input step 201, edge calculations step 202, the character horizontal stroke detects step 203, horizontal candidate's vanishing Point Detection Method step 204 based on cluster, horizontal candidate's vanishing Point Detection Method step 205 based on fladellum projection and density Estimation, horizontal candidate's vanishing Point Detection Method step 206 based on fladellum projection and cluster, horizontal end point is selected step 207, the character vertical stroke detects step 208, vertical candidate's vanishing Point Detection Method step 209 based on cluster, vertical candidate's vanishing Point Detection Method step 210 based on fladellum projection and density Estimation, vertical candidate's vanishing Point Detection Method step 211 based on fladellum projection and cluster, vertical end point is selected step 212, the perspective modification is proofreaied and correct shift step 213, and correct the back image and export step 214.
These steps can be realized by the device shown in the correspondence among Fig. 1.And similarly, perspective modification file and picture input step 201, edge calculations step 202, perspective modification are corrected shift step 213 and are corrected back image output step 214 and can realize in other device.That is, the bearing calibration of perspective distortion image of the present invention can not comprise these steps.
Below in conjunction with Fig. 2 and Fig. 1 the bearing calibration of perspective distortion document image of the present invention is described.
After step 201 has been imported perspective distortion document image by perspective distortion document image input block 101, in step 202, the edge image of bianry image is tried to achieve in edge calculations unit 102, and described bianry image detects with extraction by line character that the color and the brightness of original color or gray level image are analyzed, gone forward side by side and obtains.
In step 203, character horizontal stroke detecting unit 103 is tried to achieve the horizontal stroke of character by edge image is carried out the connected domain analysis, thereby obtains indicating the line segment of horizontal direction.
In step 204, based on horizontal candidate's vanishing Point Detection Method unit 104 of cluster by by to all horizontal line sections in twos the point formed of intersection point gather, utilize clustering method (as the K-Means method) to carry out cluster, obtain a plurality of clusters, each cluster centre is as horizontal candidate's end point.As shown in Figure 3, this horizontal candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and the character horizontal stroke is distributed in several main directions.That is, this method is especially effective to this situation.
In step 205, horizontal candidate's vanishing Point Detection Method unit 105 based on fladellum projection and density Estimation carries out following processing: the fladellum projection step, image to input carries out the fladellum projection, according to the some set that the intersection point of the horizontal line section in twos in the fladellum is formed, obtain the projection value of each fladellum; The horizontal end point of candidate is selected step, in the fladellum of maximum (being the projection value maximum) of counting, carries out dot density and estimates that the point of dot density maximum is elected the horizontal end point of candidate as.As shown in Figure 4, this horizontal candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and most of character horizontal stroke is distributed in a main direction.That is, especially effective to this situation.
In one embodiment, as shown in Figure 5, the horizontal candidate's vanishing Point Detection Method unit 105 based on fladellum projection and density Estimation can comprise: horizontal line section input block 501, horizontal line section be intersection point calculation unit 502, intersection point collection fladellum projecting cell 503, fladellum selected cell 504, dot density computing unit 505, the horizontal end point selected cell 506 of candidate and the horizontal end point output unit 507 of candidate in twos.Should be noted that horizontal line section (or horizontal stroke) herein is meant if perspective distortion does not take place image, then should be the line segment (or stroke) of level (or basic horizontal).Under the situation of image generation transmission distortion, these horizontal line sections are level no longer, that is to say, and are no longer parallel to each other between them, but for example when being extended, also have intersection point each other.But, still be referred to as horizontal line section for the convenience that illustrates.
Fig. 6 is based on the workflow diagram of horizontal candidate's vanishing Point Detection Method unit 105 of fladellum projection and density Estimation.At first after step 601 has been imported horizontal line section, in step 602, by the horizontal line section intersection point in twos of intersection point calculation unit 502 calculated level line segments (horizontal line section that draws for character horizontal stroke detecting unit 103 in the present embodiment) in twos.Then, in step 603, intersection point collection fladellum projecting cell 503 is a round dot with the picture centre, and image is divided into a plurality of five equilibriums.In plane, image place, carry out the fladellum projection: promptly in each fan-shaped five equilibrium, calculate the number of its included intersection point of line segments (obtaining intersection point calculation unit 502 in twos), be worth projection value as this fladellum with this by horizontal line section.In step 604, select a maximum fladellum of line segment number of hits (be the intersection point high-density region, or claim high density fan-beam district), just that fladellum of projection value maximum then by fladellum selected cell 504.In step 605, calculate the density of the point in selected that fladellum by dot density computing unit 505.Here, the density of point for example is meant that with this point be the center, the number of the point in certain scope.Here, can be undertaken, for example use the kN neighbour to estimate by the density estimation method of traditional point.In step 606, by the horizontal end point selected cell 506 of candidate, that point of selecting the density maximum is as the horizontal end point of candidate, and output in step 607 at last.
Get back to Fig. 2, in step 206, based on horizontal candidate's vanishing Point Detection Method unit 106 of fladellum projection and cluster carrying out following step: horizontal line section intersection point fladellum projection step, the image of input is carried out the fladellum projection, obtain a plurality of fladellums, and calculate the projection value of each fladellum; The horizontal end point of candidate is selected step, in count (projection value maximum) maximum fladellum, utilize clustering method (as the K-Means method), the intersection point of the line segment in twos in this fladellum is carried out cluster, obtain a plurality of clusters, with each cluster centre point as horizontal candidate's end point.As shown in Figure 7, this horizontal candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and most of character horizontal stroke is distributed in a main direction, and presents the main direction of several height in the zone at main direction place.That is to say that this method is particularly effective to above situation.
In one embodiment, as shown in Figure 8, the horizontal candidate's vanishing Point Detection Method unit 106 based on fladellum projection and cluster can comprise: horizontal line section input block 801, horizontal line section be intersection point calculation unit 802, intersection point collection fladellum projecting cell 803, fladellum selected cell 804, cluster cell 805, the horizontal end point selected cell 806 of candidate and the horizontal end point output unit 807 of candidate in twos.
Fig. 9 is based on the workflow diagram of horizontal candidate's vanishing Point Detection Method unit 106 of fladellum projection and cluster.After step 901 input level line segment, at first in step 902, intersection point calculation unit 802 calculate the intersection point between any two of the horizontal line section of input in twos by horizontal line section, intersection point calculation unit 602 is identical in twos with horizontal line section for method.Then, in step 903, carry out the projection of intersection point collection fladellum by intersection point collection fladellum projecting cell 803, its method is identical with intersection point collection fladellum projecting cell 603.In step 904, carry out fladellum by fladellum selected cell 804 and select then, method is identical with fladellum selected cell 604.In step 905, by cluster cell 805, utilize clustering method (as the K-Means method) to carry out cluster, obtain a plurality of clusters.In step 906, select each cluster centre point as horizontal candidate's end point by the horizontal end point selected cell 806 of candidate, and step 907 is exported again at last.
Get back to Fig. 2, in step 207,107 pairs of horizontal candidate's vanishing Point Detection Method unit 104 of horizontal end point selected cell, select, determine last horizontal end point based on horizontal candidate's vanishing Point Detection Method unit 105 of fladellum projection and density Estimation with based on the horizontal end point of all candidates that horizontal candidate's vanishing Point Detection Method unit 106 of fladellum projection and cluster obtains based on cluster.In one embodiment, adopt the middle disclosed method of No. 200710088355.0 Chinese patent application " means for correcting of perspective distortion document image and bearing calibration " to select.By reference, No. 200710088355.0 Chinese patent application " means for correcting of perspective distortion document image and bearing calibration " is incorporated herein.
Particularly, can be with these horizontal end points as the horizontal end point of candidate, and obtain the weight coefficient of the horizontal end point of each candidate, carry out the perspective projection analysis of horizontal direction then, and obtain another weight coefficient, select final horizontal end point according to this weight coefficient and this another weight coefficient then.
For example, when carrying out cluster, the intersection point number in the cluster can be accounted for the ratio of whole intersection point numbers as first weight coefficient.When carrying out density Estimation, can with the point density itself as this weight coefficient.Here, dot density is a probability, the 65th page to 66 pages " non-parametric estmation of 3.5 population distribution " it " the 3.5.1 basic skills " in " pattern-recognition (second edition) " that people such as that its estimation mode can be published referring to publishing house of Tsing-Hua University 2000, Bian Zhaoqi and Zhang Xuegong write.
In addition,, each character block is carried out the perspective projection analysis of horizontal direction, then the Projection Analysis value of all character blocks is added up, obtain the projection value accumulation variance of the horizontal end point of each candidate at the horizontal end point of each candidate.And with the projection value of the horizontal end point of each candidate accumulation variance all candidates' accumulation variance with middle proportion be second weight coefficient of the horizontal end point of this candidate.
Certainly, also can be set to 0 by one of them coefficient, only adopt direct method or only adopt round-about way to determine final horizontal end point thereby become.
Determine the method for terminal level end point according to first coefficient and second coefficient, also can be referring to above-cited No. 200710088355.0 Chinese patent application.
In step 208, character vertical stroke detecting unit 108 is tried to achieve the vertical stroke of character by edge image is carried out the connected domain analysis, thereby obtains having the line segment of vertical direction indication.
Similarly, should be noted that vertical line segment (or vertical stroke) herein is meant if perspective distortion does not take place image, then should be the line segment (or stroke) of vertical (or vertical substantially).Under the situation of image generation transmission distortion, these vertical line segments are no longer vertical, that is to say, not parallel to each other, but for example when being extended, also have intersection point each other between them.But, still be referred to as vertical line segment for the convenience that illustrates.
In step 209, vertical candidate's vanishing Point Detection Method unit 109 based on cluster passes through by the some set that the intersection point in twos of all vertical stroke line segments is formed, utilize clustering method (as the K-Means method) to carry out cluster, obtain a plurality of clusters, each cluster centre is as vertical candidate's end point.This vertical candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and the character vertical stroke is distributed in several main directions.Basic identical based on the processing of vertical candidate's vanishing Point Detection Method unit 109 of cluster with processing based on horizontal candidate's vanishing Point Detection Method unit 104 of cluster, and for example can be referring to No. 200710088355.0 Chinese patent application " means for correcting of perspective distortion document image and bearing calibration ".
In step 210, vertical candidate's vanishing Point Detection Method unit 110 based on fladellum projection and density Estimation carries out following processing at being gathered by the point that the intersection point in twos of all vertical stroke line segments is formed: vertical intersection point of line segments fladellum projection step, the fladellum projection is carried out in the some set of the intersection point of vertical line segment composition in twos in the character vertical stroke of input, obtained a plurality of fladellums; The vertical end point of candidate is selected step, in maximum fladellum of counting, carries out dot density and estimates, elects the point of dot density maximum as candidate vertical end point.These handle identical with based on horizontal candidate's vanishing Point Detection Method unit 105 of fladellum projection and density Estimation candidate's end point, be at be vertical stroke.Similarly, this vertical candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and most of character vertical stroke is distributed in a main direction.
In step 211, based on vertical candidate's vanishing Point Detection Method unit 111 of fladellum projection and cluster at carrying out following step by the some set formed of intersection point in twos of all vertical line segments: vertical intersection point of line segments fladellum projection step, the fladellum projection is carried out in the some set of the intersection point of vertical line segment composition in twos in the character vertical stroke of input, obtained a plurality of fladellums; The vertical end point of candidate is selected step, in maximum fladellum of counting, utilizes clustering method (as the K-Means method) to carry out cluster, obtains a plurality of clusters, with the central point of each cluster as vertical candidate's end point.These handle identical with horizontal candidate's vanishing Point Detection Method unit 106 based on fladellum projection and cluster, but at be vertical stroke.Similarly, this horizontal candidate's vanishing Point Detection Method method at situation be the file and picture that contains a small amount of text, and most of character horizontal stroke is distributed in a main direction, and presents the main direction of several height in the zone at main direction place.
In step 212, vertical 112 pairs of vertical candidate's vanishing Point Detection Method unit 109 based on cluster of end point selected cell, the vertical end point of all candidates that obtains based on vertical candidate's vanishing Point Detection Method unit 110 of fladellum projection and density Estimation and vertical candidate's vanishing Point Detection Method unit 111 based on fladellum projection and cluster are selected.
Particularly, can be with these vertical end points as the vertical end point of candidate, and obtain the weight coefficient of the vertical end point of each candidate, carry out the perspective projection analysis of vertical direction then, and obtain another weight coefficient, select final vertical end point according to another weight coefficient of this weight coefficient and this then.
Similar with the processing of horizontal end point selected cell 107, for example, when carrying out cluster, the intersection point number in the cluster can be accounted for the ratio of whole intersection point numbers as first weight coefficient.When carrying out density Estimation, can with the point density itself as first weight coefficient.In addition,, each character block is carried out the perspective projection analysis of vertical direction, then the Projection Analysis value of all character blocks is added up, obtain the projection value accumulation variance of the vertical end point of each candidate at the vertical end point of each candidate.And be second weight coefficient of the vertical end point of this candidate in all candidates' accumulation variance with middle proportion with the projection value of the vertical end point of each candidate accumulation variance.
Certainly, also can be set to 0 by one of them coefficient, only adopt direct method or only adopt round-about way to determine final vertical end point thereby become.
Subsequently,, correct converter unit 113 by perspective distortion and carry out perspective distortion rectification conversion in step 213, and the image after step 214 output is corrected.
Figure 10 is the structural representation according to the character horizontal stroke detecting unit 103 of one embodiment of the present invention.As shown in figure 10, character horizontal stroke detecting unit 103 comprises horizontal edge image input block 1001, edge image connected domain computing unit 1002, horizontal stroke detecting unit 1003 and horizontal stroke line segment determining unit 1004.
Figure 11 is the workflow diagram according to the character horizontal stroke detecting unit 103 of one embodiment of the invention.At first, in step 1101, by horizontal edge image input block 1001 input level edge images.In step 1102,1002 pairs of horizontal edge images of edge image connected domain computing unit are asked connected domain then.Afterwards, in step 1103,1003 pairs of connected domain shapes of horizontal stroke detecting unit and size are analyzed, and select width (stand out less than predetermined threshold value) close with character duration and its direction connected domain close with horizontal direction as candidate's horizontal stroke.Then,, analyze the shape of this candidate's horizontal stroke connected domain, determine the horizontal stroke line segment by horizontal stroke line segment determining unit in step 1104.Its concrete steps for example can for, establish C
iBe certain candidate's horizontal stroke connected domain, by points all on this connected domain is used line segment LC of minimum variance (Least-Square) algorithm match
i, the line segment equation is a
i* y+b
i* x+c=0, then point on this connected domain (x, y) distance to this line segment is,
If
Wherein
P
LCi＝N(DIS
i，μ
stroke，σ
stroke)
Wherein, (x, μ are that μ, standard variance are the Gaussian distribution of σ for the average of line segment LC σ) to N.μ
StrokeAnd σ
StrokeBe relevant average and the standard variance of character horizontal stroke that rule of thumb obtains with experiment.P_thres_stroke is a threshold value of approximate 1, can be made as 0.98.The number of stain pixel in approximate this connected domain of n_thres_stroke.If f (LC is arranged
iC is then represented in)=1
iBe the character horizontal stroke.At this moment, the line segment of these character horizontal stroke connected domain institute matches is exactly a horizontal line section.In step 1104, export by output unit 1004.
The processing of the processing of character vertical stroke detecting unit 108 and character horizontal stroke detecting unit 103 is similar, and just it detects the character vertical stroke, and exports vertical line segment (vertical stroke line segment).Character vertical stroke detecting unit 108 connected domain that the selection height is close with character height and its direction is close with vertical direction is as vertical stroke.Those of ordinary skill in the art can realize character vertical stroke detecting unit 108 according to the description to character horizontal stroke detecting unit 103, thereby has omitted detailed description in this article.
Above embodiment only is exemplary, is not limitation of the present invention.Can carry out various modification within the spirit and scope of the present invention.
Modification embodiment 1
In above embodiment, as shown in Figure 1, the horizontal line section that horizontal candidate's vanishing Point Detection Method unit 104,105,106 only utilizes character horizontal stroke detecting unit 103 to obtain, but they can also utilize the horizontal line section that obtains by other method except the horizontal line section that utilizes 103 acquisitions of character horizontal stroke detecting unit.Detect (the indirect horizontal line sections) such as respective horizontal line segments that obtains as the horizontal line section (directly horizontal line section) that edge image carried out the length that the connected domain analysis obtains and by line of text.In addition, also can only use described direct horizontal line section and indirect horizontal line section, and not use the horizontal line section that character horizontal stroke detecting unit 103 obtains (although strictly speaking, it also belongs to a kind of of indirect horizontal line section).
Similarly, for vertical vanishing Point Detection Method unit 109,110,111, they also can add or alternatively use directly vertical line segment and indirect vertical line segment.
Modification embodiment 2
Although in above embodiment, horizontal candidate's vanishing Point Detection Method unit 104,105,106 is integrated in the device carried out concentrated description, in a device, can only comprise in them one or two.During in including only them one, need not the horizontal end point selected cell 107 among Fig. 1, structure that can simplification device.
Similarly, in a device, can only comprise in the vertical vanishing Point Detection Method unit 109,110,111 one or two.
Modification embodiment 3
Although above parts separate, draw independently, this just schematically.Physically they can integrate, can separate, also can be partly integrated.Unit or module with identical function can be realized by identical physical unit, logic module.
Although above-mentioned step is dividually, sequentially describes, their execution sequence can be adjusted, also can executed in parallel.
Modification embodiment 4
In above embodiment, intersection point calculation is carried out in each candidate's vanishing Point Detection Method unit respectively in twos.But each candidate's vanishing Point Detection Method unit also can not comprise the intersection point calculation unit.And by the intersection point in twos of common intersection point calculation unit calculated level line segment or the intersection point in twos of vertical line segment.Intersection point calculation unit 502,802 etc. and the vertical line segment that omits explanation intersection point calculation unit etc. all obtain the unit corresponding to intersection point set of the present invention in twos in twos to be dispersed in horizontal line section in each candidate's vanishing Point Detection Method unit.
The present invention has utilized the character horizontal stroke to detect and has obtained horizontal line section and vertical line segment, and can utilize various line segment information, thereby can effectively utilize limited text.In addition, the present invention can utilize three kinds of candidate's vanishing Point Detection Method methods to detect, and testing result is selected, and can stably detect various possible end points.So the present invention can be effectively to the vanishing Point Detection Method of carrying out of the perspective distortion document image that comprises a small amount of text (even several character), thereby proofread and correct.
Note that in this article " comprising ", " comprising " etc., expression existed, and did not repel the existence of miscellaneous part, was the meaning that includes but not limited to.For example A comprises B, only shows to contain B among the A, also contains C among the possible A.
The embodiment that should be noted that the front is exemplary, is not limitation of the present invention.Those skilled in the art can carry out various variants and modifications within the spirit and scope of the present invention, as long as these variants and modifications have fallen into the scope of claim and equivalent thereof, just in protection domain of the presently claimed invention.
Claims (10)
1. vanishing point detecting device, described vanishing point detecting device detects the end point of file and picture, and described vanishing point detecting device comprises:
The line segment detecting unit is used to the horizontal line section or the vertical line segment of the image that obtains to import;
Intersection point set obtains the unit, is used to obtain the set of the intersection point of the set of intersection point of the horizontal line section that described line segment detecting unit obtained or vertical line segment;
Candidate's end point according to the set that described intersection point is gathered the intersection point that obtains the unit acquisition, is determined in candidate's vanishing Point Detection Method unit;
The end point determining unit according to candidate's end point that candidate's vanishing Point Detection Method unit is determined, is determined final end point,
Wherein, described candidate's vanishing Point Detection Method unit is one or more kinds in following candidate's vanishing Point Detection Method unit:
Based on candidate's vanishing Point Detection Method unit of cluster, set obtains the intersection point that the unit obtains and carries out cluster to described intersection point, obtains a plurality of clusters, with the central point of each cluster as candidate's end point;
Based on candidate's vanishing Point Detection Method unit of fladellum projection and density Estimation, the image of described input is carried out the fladellum projection, the dot density of each intersection point in the high density fan-beam district is estimated, with the intersection point of dot density maximum as candidate's end point;
Based on candidate's vanishing Point Detection Method unit of fladellum projection and cluster, the image of described input is carried out the fladellum projection, the intersection point in the high density fan-beam district is carried out cluster, obtain a plurality of clusters, with the center of each cluster as candidate's end point,
Wherein, the image of described input is carried out the fladellum projection be meant that the center with the image of described input is a round dot, the plane at the image place of described input is divided into a plurality of equal sectors, and the number of intersection point of calculating the number of the intersection point of horizontal line section described in each sector or described vertical line segment is as the projection value of this sector.
2. vanishing point detecting device according to claim 1, it is characterized in that, described line segment detecting unit comprises the character stroke detecting unit, and described character stroke detecting unit detects the vertical stroke line segment of described file and picture or horizontal stroke line segment as described vertical line segment or horizontal line section.
3. vanishing point detecting device according to claim 2 is characterized in that, described character stroke detecting unit comprises:
The connected domain computing unit obtains the connected domain of the image of described input;
The stroke detecting unit, the shape and the size of the connected domain that described connected domain computing unit is obtained are analyzed, and obtain candidate's horizontal stroke or candidate's vertical stroke;
The stroke segmental determining unit according to candidate's horizontal stroke or the candidate's vertical stroke that described stroke detecting unit is obtained, is determined horizontal stroke line segment or vertical stroke line segment.
4. vanishing point detecting device according to claim 3 is characterized in that, described stroke detecting unit selects width and character duration is close and its direction is close with horizontal direction connected domain as horizontal stroke.
5. vanishing point detecting device according to claim 3 is characterized in that, the connected domain that described stroke detecting unit selection height is close with character height and its direction is close with vertical direction is as vertical stroke.
6. vanishing point detecting device according to claim 1 is characterized in that, described candidate's vanishing Point Detection Method unit based on fladellum projection and density Estimation comprises:
Intersection point collection fladellum projecting cell carries out the fladellum projection to the image of described input, obtains the intersection point set that the unit obtained according to described intersection point set, obtains the projection value of each fladellum;
The fladellum selected cell according to the projection value of each fladellum, is selected the fladellum of projection value maximum;
The dot density computing unit calculates the dot density of each intersection point in the selected fladellum of described fladellum selected cell;
Candidate's end point selected cell, the point of selecting the density maximum that described dot density computing unit calculates is as candidate's end point.
7. vanishing point detecting device according to claim 1 is characterized in that, described candidate's vanishing Point Detection Method unit based on fladellum projection and cluster comprises:
Intersection point collection fladellum projecting cell carries out the fladellum projection to the image of described input, obtains the intersection point set that the unit obtained according to described intersection point set, obtains the projection value of each fladellum;
The fladellum selected cell according to the projection value of each fladellum, is selected the fladellum of projection value maximum;
Cluster cell carries out cluster to the intersection point in the selected fladellum of described fladellum selected cell, obtains one or more cluster;
Candidate's end point selected cell, the central point of selecting each cluster is as candidate's end point.
8. vanishing point detecting device according to claim 1 is characterized in that, described end point determining unit comprises:
First weight-coefficient calculating unit obtains first weight coefficient of each candidate's end point;
Second weight-coefficient calculating unit obtains second weight coefficient of each candidate's end point;
The end point analytic unit according to described first weight coefficient and described second weight coefficient of each described candidate's end point, is selected final end point,
Wherein when described candidate's vanishing Point Detection Method unit be during based on candidate's vanishing Point Detection Method unit of cluster or based on candidate's vanishing Point Detection Method unit of fladellum projection and cluster, described first weight-coefficient calculating unit accounts for the ratio of whole intersection point numbers as described first weight coefficient with the number of the intersection point in the cluster, when described candidate's vanishing Point Detection Method unit is candidate's vanishing Point Detection Method unit based on fladellum projection and density Estimation, described first weight-coefficient calculating unit with the dot density of described candidate's end point itself as first weight coefficient;
Wherein said second weight-coefficient calculating unit is at the vertical end point of each candidate or each horizontal end point, each character block in the image of described input is carried out the perspective projection analysis of vertical direction, then the Projection Analysis value of all character blocks is added up, obtain the projection value accumulation variance of the vertical end point of each candidate, and be second weight coefficient of the vertical end point of this candidate in the accumulation variance of the vertical end point of all candidates with middle proportion with the projection value accumulation variance of the vertical end point of each candidate.
9. vanishing Point Detection Method method, described vanishing Point Detection Method method detects the end point of file and picture, and described vanishing Point Detection Method method comprises:
Line segment detects step, is used to the horizontal line section or the vertical line segment of the image that obtains to import;
The intersection point set obtains step, is used to obtain set or the vertically set of the intersection point of line segment that described line segment detects the intersection point of the horizontal line section that step obtained;
Candidate's vanishing Point Detection Method step according to the set that described intersection point is gathered the intersection point that obtains the step acquisition, is determined candidate's end point;
The end point determining step according to candidate's end point that candidate's vanishing Point Detection Method step is determined, is determined final end point,
Wherein, described candidate's vanishing Point Detection Method step adopts one or more kinds in following candidate's vanishing Point Detection Method method:
Based on candidate's vanishing Point Detection Method method of cluster, set obtains the intersection point that step obtains and carries out cluster to described intersection point, obtains a plurality of clusters, with the central point of each cluster as candidate's end point;
Based on candidate's vanishing Point Detection Method method of fladellum projection and density Estimation, the image of described input is carried out the fladellum projection, the dot density of each intersection point in the high density fan-beam district is estimated, with the intersection point of dot density maximum as candidate's end point;
Based on candidate's vanishing Point Detection Method method of fladellum projection and cluster, the image of described input is carried out the fladellum projection, the intersection point in the high density fan-beam district is carried out cluster, obtain a plurality of clusters, with the center of each cluster as candidate's end point,
Wherein, the image of described input is carried out the fladellum projection be meant that the center with the image of described input is a round dot, the plane, image place of described input is divided into a plurality of equal sectors, and the number of intersection point of calculating the number of the intersection point of horizontal line section described in each sector or described vertical line segment is as the projection value of this sector.
10. vanishing Point Detection Method method according to claim 9 is characterized in that, described candidate's vanishing Point Detection Method step comprises when the candidate's vanishing Point Detection Method method that adopts based on fladellum projection and density Estimation:
Intersection point collection fladellum projection step is carried out the fladellum projection to the image of described input, obtains the intersection point set that step obtained according to described intersection point set, obtains the projection value of each fladellum;
Fladellum is selected step, according to the projection value of each fladellum, selects the fladellum of projection value maximum;
The dot density calculation procedure is calculated the density that described fladellum is selected the point in the selected fladellum of step;
Candidate's end point is selected step, and the point of selecting the density maximum that described dot density calculation procedure calculates is as candidate's end point,
And wherein,
Described candidate's vanishing Point Detection Method step comprises when the candidate's vanishing Point Detection Method method that adopts based on fladellum projection and cluster:
Intersection point collection fladellum projection step is carried out the fladellum projection to the image of described input, obtains the intersection point set that step obtained according to described intersection point set, obtains the projection value of each fladellum;
Fladellum is selected step, according to the projection value of each fladellum, selects the fladellum of projection value maximum;
The cluster step selects the intersection point in the selected fladellum of step to carry out cluster to described fladellum, obtains one or more cluster;
Candidate's end point is selected step, and the center of selecting each cluster is as candidate's end point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008100809519A CN101520852B (en) | 2008-02-29 | 2008-02-29 | Vanishing point detecting device and detecting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008100809519A CN101520852B (en) | 2008-02-29 | 2008-02-29 | Vanishing point detecting device and detecting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101520852A CN101520852A (en) | 2009-09-02 |
| CN101520852B true CN101520852B (en) | 2011-09-07 |
Family
ID=41081433
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008100809519A Expired - Fee Related CN101520852B (en) | 2008-02-29 | 2008-02-29 | Vanishing point detecting device and detecting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101520852B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102208022A (en) * | 2010-03-31 | 2011-10-05 | 富士通株式会社 | Shaded character recovery device and method thereof, shaded character recognition device and method thereof |
| CN102609938B (en) * | 2012-01-16 | 2014-04-16 | 浙江大学 | Method for detecting vanishing points of road based on single image |
| US8897600B1 (en) * | 2013-12-20 | 2014-11-25 | I.R.I.S. | Method and system for determining vanishing point candidates for projective correction |
| US8811751B1 (en) * | 2013-12-20 | 2014-08-19 | I.R.I.S. | Method and system for correcting projective distortions with elimination steps on multiple levels |
| US8913836B1 (en) * | 2013-12-20 | 2014-12-16 | I.R.I.S. | Method and system for correcting projective distortions using eigenpoints |
| JP6542230B2 (en) * | 2013-12-20 | 2019-07-10 | イ．エル．イ．エス． | Method and system for correcting projected distortion |
| KR102225620B1 (en) * | 2014-04-03 | 2021-03-12 | 한화테크윈 주식회사 | Camera modeling system |
| CN105654082B (en) * | 2014-11-12 | 2019-04-12 | 佳能株式会社 | Character recognition post-processing approach, equipment and the image pick up equipment including the equipment |
| CN106296745B (en) * | 2015-05-26 | 2019-03-12 | 富士通株式会社 | To the corrected method and apparatus of file and picture |
| CN107305688B (en) * | 2016-04-15 | 2020-10-27 | 株式会社理光 | Method, device and system for detecting road vanishing point |
| CN108492284B (en) * | 2018-03-12 | 2020-03-03 | 百度在线网络技术（北京）有限公司 | Method and apparatus for determining perspective shape of image |
| CN110136182B (en) * | 2019-05-28 | 2021-06-04 | 北京百度网讯科技有限公司 | Registration method, device, equipment and medium for laser point cloud and 2D image |
| CN113409235B (en) * | 2020-03-17 | 2023-08-22 | 杭州海康威视数字技术股份有限公司 | Vanishing point estimation method and apparatus |
| CN112150380B (en) * | 2020-09-22 | 2024-04-16 | 北京百度网讯科技有限公司 | Method, apparatus, electronic device, and readable storage medium for correcting image |
| CN112101321B (en) * | 2020-11-18 | 2021-02-02 | 蘑菇车联信息科技有限公司 | Vanishing point extraction method and device, electronic equipment and storage medium |
| CN113096051B (en) * | 2021-04-30 | 2023-08-15 | 上海零眸智能科技有限公司 | Map correction method based on vanishing point detection |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1471055A (en) * | 2002-07-02 | 2004-01-28 | ��ʿͨ��ʽ���� | Image distortion correction method and apparatus |
| CN1855150A (en) * | 2005-04-28 | 2006-11-01 | 索尼株式会社 | Image processing device, method, program and recording medium |
-
2008
- 2008-02-29 CN CN2008100809519A patent/CN101520852B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1471055A (en) * | 2002-07-02 | 2004-01-28 | ��ʿͨ��ʽ���� | Image distortion correction method and apparatus |
| CN1855150A (en) * | 2005-04-28 | 2006-11-01 | 索尼株式会社 | Image processing device, method, program and recording medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101520852A (en) | 2009-09-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101520852B (en) | Vanishing point detecting device and detecting method | |
| US10282853B2 (en) | Method for tracking object in video in real time in consideration of both color and shape and apparatus therefor | |
| CN101267493B (en) | Correction device and method for perspective distortion document image | |
| JP4545641B2 (en) | Similar image retrieval method, similar image retrieval system, similar image retrieval program, and recording medium | |
| CN102176208B (en) | Robust video fingerprint method based on three-dimensional space-time characteristics | |
| CN106295564B (en) | A kind of action identification method of neighborhood Gaussian structures and video features fusion | |
| CN107103326A (en) | The collaboration conspicuousness detection method clustered based on super-pixel | |
| CN105354578B (en) | A kind of multiple target object image matching method | |
| CN106991686B (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
| CN103778436B (en) | A kind of pedestrian's attitude detecting method based on image procossing | |
| CN104809436A (en) | Curved written text identification method | |
| CN101840514A (en) | Image object classification device and method | |
| CN104240220A (en) | Equipment and method for segmenting blood vessels based on images | |
| CN101770583B (en) | Template matching method based on global features of scene | |
| CN111191649A (en) | Method and equipment for identifying bent multi-line text image | |
| CN104573681A (en) | Face recognition method | |
| CN107622271A (en) | Handwriting text lines extracting method and system | |
| CN109308465A (en) | Table line detecting method, apparatus, equipment and computer-readable medium | |
| KR20190080388A (en) | Photo Horizon Correction Method based on convolutional neural network and residual network structure | |
| CN103745197A (en) | Detection method of license plate and device thereof | |
| CN103559715B (en) | A kind of method for detecting abnormality of high spectrum image and device | |
| CN115861738A (en) | Category semantic information guided remote sensing target detection active sampling method | |
| CN103226824A (en) | Video retargeting system for maintaining visual saliency | |
| CN108876776B (en) | Classification model generation method, fundus image classification method and device | |
| Montagner et al. | Staff removal using image operator learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110907 Termination date: 20190228 |