JPH113433A - Table closing line intersection correcting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To compensate an intersection of ruled lines that restores an intersection when the intersection of table frame lines of a table which is not a square by making the crossed point an intersection of outer frame ruled lines with each other when the sum of length of an extended line of an outer frame ruled line that has intersection absence and an extended line of another outer frame ruled line is shorter than prescribed length. SOLUTION: A vertical segment line and a horizontal segment line in a segmented image are detected from a table area. Thinning of the segment lines are compensated, and vertical ruled lines and horizontal ruled lines are detected from the segment lines after thinning compensation. Outer frame ruled lines are extracted from the detected ruled lines. When an end point of one remarked outer frame ruled line does not cross the other outer frame ruled lines, the remarked outer frame ruled line is extended to an end point that is not crossed, and if the length is below a reference value α, the intersection is made the intersection of outer frame ruled lines. Also, when outer frame ruled lines except the remarked outer frame ruled line are extended, the extended lines of the outer frame ruled lines mutually cross, and also, the sum of the length of two extended lines is bellow reference value β, the intersection is made the intersection of outer frame ruled lines of the table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for correcting intersections of table frame lines, a table recognizing device, and an optical character reading device. More specifically, when an intersection of a table frame line in an image has disappeared, an intersection correction device that restores the intersection of the disappearing table frame line, and a character in an optically read image using the intersection correction device The present invention relates to an optical character reading device that reads characters correctly.

[0002]

2. Description of the Related Art As a conventional technique relating to correction of blur of a ruled line, there is a technique described in Japanese Patent Application Laid-Open No. 8-125855. Hereinafter, this conventional technique will be described.

FIG. 1 is a block diagram showing a table recognizing device including a table frame line intersection correction device according to an embodiment of the prior art.

The table recognizing device 20 includes an image input unit 102 for inputting an image via an image scanner or a communication line.
An intersection correction device 10 that, when the intersection of the table frame lines in the image has disappeared, restores the intersection of the disappeared table frame lines;
Based on the intersection corrected by the intersection corrector 10, the connection relation of the table frame lines in the image is extracted, and the frame line recognition unit 108 recognizes the table frame line based on the connection relation, and recognizes the table frame line. (E.g., data on connection relations of frame lines in the model of the table).
The image processing apparatus includes a frame recognition result output unit 112 that outputs a recognition result by the frame recognition unit 108, and a control unit 100 that controls the image input unit 102 to the frame recognition result output unit 112.

[0005] As a conventional technique for table recognition,
Paper “Structure Recognition of Form Documents Using Frame Rule Line Information (Hiroyuki Naruse et al., IEICE Journal, D-II, Vol. J.
75-D-II, pp. 1372-1385, 1992
Year)].

FIG. 2 is a PAD diagram showing a table extraction process performed by the table frame line intersection correction device 10. Step 200
Then, a table included in an input image is extracted by contour tracing. The contour tracking is a process of tracking a boundary between a black pixel and a white pixel. In step 201, a rectangular area having an upper left point and a lower right point as diagonal points of a table extracted by contour tracking is defined as one table area, and an image of the table area is cut out. The processing of steps 200 and 201 will be described with reference to specific examples shown in FIGS. In step 200, a table is extracted from the image of FIG. 5 by contour tracing (600) as shown in FIG. In step 201, as shown in FIG.
A table area determined by the diagonal points 601 and 602 is cut out. Thereby, the image of FIG. 7 is obtained. Adjacent tables can be separated by the table extraction processing. In addition, noise can be removed to some extent.

FIG. 3 is a PAD diagram of the process of correcting the intersection of the outer frame ruled line of the table by the table frame line intersection correction device 10.

In step 202, a vertical ruled line and a horizontal ruled line in an image cut out from the table area are detected, for example, according to the following detection rule. (1) A ruled line is a vertical or horizontal line segment, and a diagonal line segment is not a ruled line but a noise component. (2) The ruled line is longer than a predetermined length, and a line segment shorter than the predetermined length is not a ruled line but a noise component. (3) At least one of the two end points of the ruled line intersects with another ruled line, and a line segment at which neither of the two end points intersects another ruled line is not a ruled line but a noise component.

In step 203, an outer frame rule is extracted from the detected rule lines according to, for example, the following extraction rule. (A) If there is always another horizontal ruled line when looking upward from an arbitrary position on a certain horizontal ruled line, the horizontal ruled line is not an outer frame ruled line. If there is a position where another horizontal rule does not exist, the horizontal rule is an outer frame rule. (B) If there is always another horizontal ruled line when looking down from an arbitrary position on a certain horizontal ruled line, the horizontal ruled line is not an outer frame ruled line. If there is a position where another horizontal rule does not exist, the horizontal rule is an outer frame rule. (C) If there is always another vertical ruled line when looking left from an arbitrary position on a vertical ruled line, the vertical ruled line is not an outer frame ruled line. If there is a position where no other vertical ruled line exists, the vertical ruled line is an outer frame ruled line. (D) If there is always another vertical ruled line when looking right from an arbitrary position on a vertical ruled line, the vertical ruled line is not an outer frame ruled line. If there is a position where no other vertical ruled line exists, the vertical ruled line is an outer frame ruled line.

In step 204, the execution of step 206 focusing on one of the extracted outer frame rules is repeated by the number of extracted outer frame rules.

In step 206, it is determined whether or not the end point of one outer frame ruled line intersects another outer frame ruled line. If they cross, step 204
Return to If they do not intersect, it is considered that the vicinity of the intersection has disappeared, and the process proceeds to step 208.

In step 208, the outer frame ruled line of interest is extended to an end point not intersecting with another outer frame ruled line. In step 210, it is determined whether or not the extension line in step 208 intersects another outer frame ruled line. If they intersect, go to step 212. If not, the process proceeds to step 214.

In step 212, the intersection is set as the intersection of the outer frame line of the table. Here, when there are a plurality of other outer frame rule lines intersecting the extension line, the intersection with the outer frame rule line closest to the end point is set as the intersection of the outer frame line in the table. In step 213, the same outer frame ruled line extraction processing as in step 203 is performed.

In step 214, outer frame rules other than the outer frame rule of interest are extended. In step 216, it is determined whether the extension line in step 208 and the extension line in step 214 intersect. If so, go to step 218. If they do not cross, the process returns to step 204.

In step 218, the intersection is set as the intersection of the ruled line of the outer frame of the table. Here, when there are a plurality of extension lines of another outer frame rule line that intersect with the extension line of the outer frame rule line of interest, the intersection point with the other outer frame rule line having the shortest extension line is defined as the outer frame line of the table. This is the intersection of the ruled lines. In step 220, the same outer frame ruled line extraction processing as in step 203 is performed.

The processing from step 202 to step 220 will be described with reference to specific examples in FIGS.

In step 202, ruled lines are detected from the image shown in FIG. Thereby, the image of FIG. 8 is obtained.

In step 202, an outer frame ruled line is detected from the image shown in FIG. Thereby, the image of FIG. 9 is obtained.

At step 206, as shown in FIG. 10, it is determined whether or not the end point 402a of the outer frame ruled line 402 intersects another outer frame ruled line. Since the end points 402a do not intersect, the process proceeds to step 208. Step 208
Then, the extension line 402b is extended. In step 210,
It is determined whether or not the extension line 402b intersects another outer frame ruled line. Since the extension line 402b intersects with the outer frame ruled line 418, the process proceeds to step 212. In step 212, the intersection 418a is set as the intersection of the outer frame line of the table. Therefore, the result is as shown in FIG.

In step 213, outer frame ruled lines are detected from the image shown in FIG. Thereby, the image of FIG. 12 is obtained.

Next, in step 206, as shown in FIG. 13, since the end point 418b of the outer frame ruled line 418 does not intersect with another outer frame ruled line, the process proceeds to step 208. In step 208, the extension line 418c is extended. Step 21
At 0, the extension line 418c intersects with the outer frame ruled line 410, so that the process proceeds to step 212. In step 212, the intersection 410a is set as the intersection of the outer frame line of the table. Therefore, the result is as shown in FIG.

Next, in step 206, as shown in FIG. 13, both end points of the outer frame rule 406 are
Since it intersects with 14, 420, the next outer frame ruled line is searched for, and step 206 is repeated.

Next, in step 206, as shown in FIG. 14, since the end point 420a of the outer frame ruled line 420 does not intersect with another outer frame ruled line, the process proceeds to step 208. In step 208, the extension line 420b is extended. Step 21
At 0, the process proceeds to step 214 because the extension line 420b does not intersect with another outer frame ruled line. In step 214, outer frame rules other than the outer frame rule 420 are extended. Step 216
Then, it is determined whether the extension line 420b intersects with another extension line. Since the extension line 420b intersects with another extension line 402d, the process proceeds to step 218. In step 218, the intersection 420c is set as the intersection of the outer frame line of the table. In step 220, an outer frame ruled line is detected. As a result, FIG.
Is obtained.

Next, in step 206, as shown in FIG. 16, since the end point 410a of the outer frame rule 410 does not intersect with another outer frame rule, the process proceeds to step 208. In step 208, the extension line 410b is extended. Step 21
If the value is 0, the extension line 410b does not intersect with another outer frame ruled line, so that the process proceeds to step 214. In step 214, the outer frame ruled lines other than the outer frame ruled line 410 are extended. Step 216
Then, it is determined whether the extension line 410b intersects with another extension line. Since the extension line 410b intersects with another extension line 412b, the process proceeds to step 218. In step 218, the intersection 410a is set as the intersection of the outer frame line of the table. In step 220, an outer frame ruled line is detected. As a result, FIG.
Is obtained.

FIG. 4 is a PAD diagram of the process of correcting the intersection of the internal ruled line of the table by the intersection correction device 10 for the table frame line.

In step 300, an internal ruled line is detected from the image. In step 301, the execution of step 302 by focusing on one of the internal ruled lines in the table is repeated by the number of internal ruled lines in the table. In step 302, it is determined whether the end point of one internal ruled line of interest intersects another ruled line. If so, the process returns to step 300. If they do not intersect, it is considered that the vicinity of the intersection has disappeared, and the process proceeds to step 304.

In step 304, the internal ruled line of interest is extended to an end point that does not intersect with another ruled line. In step 306, it is determined whether or not the extension line in step 304 intersects another ruled line. If they intersect, go to step 308. If not, the process returns to step 300. In step 308, the intersection is set to the intersection of the table frame lines. Here, when there are a plurality of other ruled lines intersecting the extension line, the intersection with the ruled line closest to the end point is set as the intersection of the table frame line.

The processing from step 300 to step 308 will be described with reference to specific examples in FIGS. In step 300, the internal ruled lines 404 to 408 and 414 to 418 are detected from the image whose outer frame has been corrected as shown in FIG. In step 302, as shown in FIG.
Since the end point 418a of the internal ruled line 418 does not intersect with another ruled line, the process proceeds to step 304. In step 304, the extension line 418b is extended. In step 306, since the extension line 418b intersects with the other ruled lines 408 and 410, the process proceeds to step 308. In step 308, since the extension line 418b intersects with a plurality of other ruled lines 408 and 410, another ruled line 408 closest to the end point 418a is selected, and an intersection 418c with the selected ruled line 408 is set as the intersection of the table frame line.

As described above, as shown in FIG. 20, the disappeared frame line is restored, so that the frame line recognition unit 108 can correctly recognize the frame line. That is, according to the table recognition device 20, the table can be correctly recognized. In steps 212, 218, and 308, a rule may be added that “if the distance between the endpoint to be corrected for intersection and the ruled line or the endpoint exceeds a reference value, intersection correction is not performed”.

FIG. 21 is a block diagram showing an optical character reading apparatus including an apparatus for correcting intersections of table frame lines according to one embodiment of the prior art. The optical character reading device 30 includes an intersection correction device 10 described above, an image input unit 102 that inputs an image to the intersection correction device 10 via an image scanner, and a display based on the intersection corrected by the intersection correction device 10. A frame line recognizing unit 108 that extracts the connection relation of the frame lines and recognizes the table frame line based on the connection relation, and data for recognizing the table frame line (for example, data of the connection relation of the frame line in the table model).
, A character line detection unit 902 that detects a character line in a frame, and a character recognition unit 904 that recognizes characters in a character line by comparing them with a character recognition dictionary 906.
And a character recognition result output unit 908 for outputting a character recognition result
From the intersection correction device 10 to the frame line recognition result output unit 11
And a control unit 900 for controlling up to two.

FIG. 22 is a table in which the disappeared frame lines are restored. It can be clearly seen that the characters 422, 424, 426 are in the table frame (cell) and the symbol 428 is not in the table. On the other hand, in FIG. 5, the character 4
It is not clear whether 22, 424, 426 and the symbol 428 are in the table.

Thus, according to the optical character reading device 30, the characters in the table can be correctly recognized.

[0033]

SUMMARY OF THE INVENTION The above-mentioned JP-A-8-125
In the conventional intersection blur correction method described in Japanese Unexamined Patent Publication No. 855, a ruled line is divided into an outer frame ruled line and an inner ruled line, and the intersection blur correction is performed on each of the ruled lines. This method can correct the outer frame ruled line of the square table, but has a problem in correcting the blurring of the intersection of the outer frame ruled line of the non-rectangular table such as the step shape. Hereinafter, blurring of the intersection which is difficult to correct in the related art will be described with reference to specific examples of FIGS.

FIG. 23 is an example of an input image without blurring.

FIG. 24 is an example of an image in which the intersections of the outer frames in FIG. 23 are blurred. This FIG. 24 will be described as an input image.

In step 202 of FIG. 3, ruled lines are detected from the image of FIG. Thereby, the image of FIG. 25 is obtained.

In step 203, outer frame ruled lines are detected from the image shown in FIG. Thereby, the image of FIG. 26 is obtained.

In step 206, as shown in FIG. 27, it is determined whether or not the end point 2416a of the outer frame ruled line 2416 intersects another outer frame ruled line. Since the end points 2416a do not intersect, the process proceeds to step 208. In step 208, the extension line 2416b is extended. Step 21
At 0, it is determined whether or not the extension line 2416b intersects another outer frame ruled line. The extension line 2416b is the outer frame ruled line 240
Since it intersects with No. 6, the process proceeds to step 212. Step 2
In 12, the intersection 2406c is set as the intersection of the outer frame line of the table.
Thereby, the image of FIG. 28 is obtained.

In step 213, outer frame ruled lines are detected from the image shown in FIG. As a result, the image shown in FIG. 29 is obtained.

Next, in step 206, as shown in FIG. 30, since the end point 2404a of the outer frame ruled line 2404 does not intersect with another outer frame ruled line, the process proceeds to step 208. In step 208, the extension line 2404b is extended. In step 210, since the extension line 2404b intersects the outer frame ruled line 2900, the process proceeds to step 212. Step 212
Then, the intersection 2404c is set as the intersection of the outer frame line of the table. Thereby, the image of FIG. 31 is obtained.

In step 213, outer frame ruled lines are detected from the image shown in FIG. As a result, the image shown in FIG. 32 is obtained.

The image obtained by the above processing is shown in FIG.
It is. This is compared with FIG. 23.
It can be seen that the intersection corresponding to 316a is incorrectly corrected. That is, while the outer frame ruled line 2316 in FIG. 23 intersects with the outer frame ruled line 2302 and 2316a, FIG.
The outer frame ruled line 2900 of No. 3 passes through the intersection 2316a and intersects with the ruled line 2306.

Accordingly, an object of the present invention is to provide a table frame line intersection correction device for correcting an intersection point of a ruled line for correctly restoring the intersection point when the intersection of the table frame lines of a non-rectangular table has disappeared. It is in.

[0044]

According to a first aspect of the present invention, there is provided an intersection correcting apparatus for recovering an intersection of a disappearing table frame line when the intersection of the table frame line in the image has disappeared. A line detecting means for detecting a horizontal line and a vertical line from an image; a blur correcting means for correcting a blur of the detected line; and a rule detecting a horizontal rule and a vertical rule of a table from the line. Detecting means, outer frame rule extracting means for extracting the outer frame rule of the table from the detected rule, and intersection missing outer frame detecting an intersection missing outer frame rule having an end point at which an intersection with another outer frame rule is missing Ruled line detecting means, intersection missing outer frame rule extending means for extending the intersection missing outer frame rule to the end point where the intersection is missing, and an extension of the intersection missing outer frame rule intersects another outer frame rule, Is shorter than the predetermined length, the intersection point is regarded as the intersection of the outer frame ruled lines. Outer frame line intersection revival means, if the extension line of the intersection missing outer frame rule line does not intersect with another outer frame rule line, extend another outer frame rule line extending means, and the intersection missing outer frame The extension line of the ruled line intersects with the extension line of the other outer frame rule line, and the sum of the extension line of the intersection missing outer frame rule line and the extension line of the other outer frame rule line is greater than a predetermined length. When the length is short, a second outer frame line intersection reverting means for setting the crossed point to the intersection of the outer frame ruled lines is provided.

According to a second aspect of the present invention, there is provided an intersection correction device for a table frame having the above-described configuration, image input means for inputting an optically read image optically read to the intersection correction device, and an intersection correction device. Line recognition means for recognizing a table frame line based on the intersection corrected by the method, character line recognition means for recognizing a character line in a frame based on the recognition result by the frame line recognition means, and the character line recognition means An optical character reading device comprising: character recognition means for recognizing characters in a character line based on the recognition result by the character recognition means; and character recognition result output means for outputting a recognition result by the character recognition means. provide.

[0046]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below in detail.

FIG. 34 is a PAD showing a process of correcting the intersection of the outer frame ruled line by the table frame intersection correction device according to one embodiment of the present invention.
FIG. The processing in FIG. 34 uses the table extraction processing in FIG. 2 as preprocessing. In FIG. 34, the processing in the 200s is the same as the processing in FIG.

In step 3400, a vertical line segment and a horizontal line segment in the image cut out from the table area are detected in accordance with, for example, the following detection rule. (1) A line segment is a vertical or horizontal line segment, and a diagonal line segment is a noise component. (2) The line segment is longer than a predetermined length, and a line segment shorter than the predetermined length is a noise component.

In step 3401, the process of executing step 3402 by focusing on one of the extracted line segments is repeated by the number of extracted line segments.

In step 3402, the blur of the line segment is corrected according to, for example, the following blur correction rule. “If there is another line segment on the extension of the line segment and the end point interval between these two line segments is equal to or smaller than the reference value γ, two ruled lines are connected.” In step 3404, after the blur correction, A vertical ruled line and a horizontal ruled line are detected from the line segment according to, for example, the following detection rule. (A) Ruled lines are vertical or horizontal line segments, and oblique line segments are noise components. (B) The ruled line satisfies at least one of the following conditions (a) and (i). (A) At least one of the two end points of the ruled line intersects with another ruled line, and a line segment at which neither of the two end points intersects another ruled line is a noise component, not a ruled line. (I) The ruled line is longer than a predetermined length, and a line segment shorter than the predetermined length is a noise component. According to the detection rule (b), a line segment longer than a predetermined length is detected as a ruled line even if both end points do not intersect with another ruled line, and a line segment shorter than the predetermined length has at least one end point of another line. If it crosses a ruled line, it is detected as a ruled line.

In step 203, an outer frame ruled line is extracted from the detected ruled lines.

In step 3406, focusing on one of the extracted outer frame ruled lines, the processing from step 206 onward is repeated by the number of extracted outer frame ruled lines. This process is performed in order from the outer frame ruled line to the inner outer frame ruled line of the table.

In step 206, it is determined whether or not the end point of one outer frame ruled line intersects another outer frame ruled line. If they intersect, go to step 340
Return to 6. If they do not intersect, it is considered that the vicinity of the intersection has disappeared, and the process proceeds to step 208.

In step 208, the outer frame ruled line of interest is extended to an end point not intersecting with another outer frame ruled line.

In step 210, step 208 is performed.
It is determined whether or not the extension line due to intersects with another outer frame ruled line. If they intersect, go to step 3408. If not, the process proceeds to step 3410. In step 3408, it is determined whether or not the length of the line obtained by extending the outer frame ruled line up to the intersection point of interest is equal to or smaller than the reference value α. If it is equal to or smaller than α, the process proceeds to step 212 If it is larger than α,
Proceed to step 3410. In step 212, the intersection is set to the intersection of the outer frame lines of the table.

At step 3410, it is determined whether or not the intersection of the focused outer frame ruled line has been corrected. If the correction has been made, the process proceeds to step 213. If not, the process proceeds to step 3414. In step 213, the same outer frame ruled line extraction processing as in step 203 is performed. Step 34
At 12, the process returns to step 3406.

At step 3414, one of the outer frame ruled lines other than the focused outer frame ruled line is focused on, and step 214 is executed.
The following processing is repeated by the number of outer frame ruled lines that satisfy the following conditions, for example. (A) When the outer frame ruled line of interest is extended upward, only the outer frame ruled line above the table is displayed. (B) When the outer frame ruled line of interest is extended downward, only the lower frame ruled line at the bottom of the table is used. (C) When the outer frame ruled line of interest is extended to the left, only the outer frame ruled line on the left side of the table is used. (D) If the outer frame ruled line of interest is extended to the right, only the outer frame ruled line on the right side of the table. This process is performed in order from the outer frame ruled line to the inner outer frame ruled line of the table.

In step 214, outer frame rules other than the outer frame rule of interest are extended.

In step 3416, it is determined whether or not the extended lines of the outer frame ruled lines intersect with each other and whether the sum of the lengths of the two extended lines is equal to or smaller than the reference value β. If these two conditions are satisfied, the process proceeds to step 218; otherwise, the process returns to step 3414.

In step 218, the intersection is set as the intersection of the ruled line of the outer frame of the table. In step 220, the same outer frame ruled line extraction processing as in step 203 is performed. Step 341
In step 8, the repetition process of step 3414 is interrupted, and the process returns to step 3406.

Steps 3400 to 3 in FIG.
The processing up to 412 will be described with reference to FIG. 24 and specific examples in FIGS.

In step 3400, a vertical line segment and a horizontal line segment are detected from the image shown in FIG. As a result, the image shown in FIG. 35 is obtained. In step 3402, as shown in FIG. 36, blurring correction of line segments based on continuity is performed. In FIG. 36, 2404 exists on the extension of the line segment 2402, and the length of 2404b between the end points 2402a and 2404a is equal to or smaller than the reference value γ, so that the line segments 2402 and 2404 are connected. By this processing, the image of FIG. 37 is obtained. In step 3404, ruled lines are detected from the image of FIG. As a result, the image shown in FIG. 38 is obtained. In step 203, outer frame ruled lines are detected from the image of FIG. As a result, the image shown in FIG. 39 is obtained. In step 206, as shown in FIG.
Is determined whether or not intersects with another outer frame ruled line. Since the end points 2416a do not intersect, the process proceeds to step 208.

In step 208, the extension 2416d is extended as shown in FIG. In step 210,
It is determined whether or not the extension line 2416d intersects another outer frame ruled line. Since the extension line 2416d intersects the outer frame ruled line 3700, the process proceeds to step 3408. Step 3408
Then, since the length of the extension line 2416d is equal to or less than the reference value α,
Proceed to step 212. In step 212, the intersection 24
Let 16e be the intersection of the outer frame lines in the table. Therefore, the image of FIG. 41 is obtained. In step 3410, the outer frame ruled line 241
Since the blurring of No. 6 has been corrected, the process proceeds to step 213.
In step 213, an outer frame ruled line is detected from the image of FIG. Thereby, the image of FIG. 42 is obtained. In step 3412, the process returns to step 3406. However, since all the outer ruled lines intersect with the other ruled lines, the process of correcting the intersection of the outer frame ruled lines ends.

As a result of the above processing, FIG. 43 is obtained as an image in which the blur of the intersection of the ruled lines has been corrected. Since the ruled line in FIG. 43 is the same as the ruled line in FIG. 23, the blur of the outer frame ruled line is correctly corrected.

Next, the processing of FIG. 34 will be described with reference to specific examples of FIGS.

In step 3400, a vertical line segment and a horizontal line segment are detected from the image shown in FIG. As a result, the image shown in FIG. 45 is obtained. In step 3402, there is no ruled line that satisfies the blurring correction condition, so that blurring correction is not performed. In step 3404, ruled lines are detected from the image of FIG.
As a result, the image shown in FIG. 46 is obtained.

In step 203, an outer frame ruled line is detected from the image shown in FIG. As a result, the image shown in FIG. 47 is obtained.

In step 206, as shown in FIG. 48, it is determined whether or not the end point 4400a of the outer frame ruled line 4400 intersects another outer frame ruled line. Since the end point 4400a does not intersect, the process proceeds to step 208. In step 208, the outer frame ruled line 4400 is extended. Step 2
At 10, it is determined whether or not the extension line 4400b intersects with another outer frame ruled line. As shown in FIG.
Since 00b does not intersect with any outer frame rule, step 3
Proceed to 410. In step 3410, the outer frame ruled line 440
Since the blurring of 0 has not been corrected, step 3414 is executed.
Proceed to. In step 214, since the ruled line 4400 is extended to the right, the rightmost (outer) outer frame ruled line 24
Extend 18

In step 3416, as shown in FIG. 49, the extension lines 4400b and 2418b intersect and
It is determined whether the sum of the lengths of b and 2418b is equal to or smaller than a reference value β. In this case, the extension lines 4400b and 2418b intersect, but the process returns to step 214 because the sum of the lengths of 4400b and 2418b exceeds the reference value β. Step 2
At 14, the right (outer) outer frame ruled line 2416 is extended.

In step 3416, as shown in FIG. 50, the extension lines 4400c and 4402b intersect and
It is determined whether the sum of the lengths of c and 4400b is equal to or smaller than a reference value β. In this case, since these two conditions are satisfied, the process proceeds to step 218. At step 218, the intersection 4402
Let c be the intersection of the outer frame lines in the table. Therefore, the image of FIG. 51 is obtained.

In step 220, outer frame ruled lines are detected from the image shown in FIG. Thereby, the image of FIG. 42 is obtained.

In step 3418, step 3406
Return to However, since all the outer ruled lines intersect with the other ruled lines, the process of correcting the intersection of the outer frame ruled lines ends.

As a result of the above processing, FIG. 43 is obtained as an image in which the blur of the intersection of the ruled lines has been corrected. Since the ruled line in FIG. 43 is the same as the ruled line in FIG. 23, the blur of the outer frame ruled line is correctly corrected.

FIG. 53 shows the overall structure of the optical character reader. The optical character reading device includes an image input unit for inputting an image to be recognized via an image scanner or a communication line, a memory for recording data input from the image input unit and a program for executing the present invention, a frame recognition dictionary, It comprises an external storage device for storing a character recognition dictionary and the like, a display for outputting a character recognition result and a recognition result output unit which is a communication line to the outside, and a CPU for controlling the whole. In addition, the memory can be used as normally used in a computer. In the present application, when executing the above-described program, copying of a dictionary and storage of data from an image input unit during recognition are also performed. It is possible.

FIG. 52 is a block diagram of a program for executing the optical character reading including the correction of the intersection of the table frame lines according to the embodiment of the present invention. The optical character reading device (program) 50 includes an intersection correction unit 4 that, when an intersection of a table frame line in an image has disappeared, restores the intersection of the disappeared table frame line.
0, a frame line recognizing unit 108 that extracts a connection relationship of the table frame lines in the image based on the intersection corrected by the intersection correction device 40, and recognizes the table frame line based on the connection relationship. A frame recognition dictionary 11 storing data for recognition (for example, data of connection relation of frame lines in a model of a table).
0 and a character line detection unit 902 that detects a character line in the frame
And a character recognizing unit 904 for recognizing characters in a character line by collating them with a character recognition dictionary 906. The contents of execution of these units are as described above. Further, a control unit 5200 for controlling these units is provided. Control unit 520
0 indicates that the image data input from the image input unit in FIG. 53 is to be processed by the intersection correction unit 40, the frame line recognition unit 108, and the character line detection unit 90.
2. It controls the character recognition unit 904 to recognize the input image and outputs the recognition result to the recognition result output unit in FIG.

An example of the implementation of the frame recognition dictionary 110 is described in the above-mentioned paper "Structural Recognition of Form Documents Using Frame Ruled Line Information".

In the optical character reading device 50, even if the image of FIG. 24 or FIG. 44 is inputted, the disappeared frame line is restored as shown in FIG. 23 by the operation of the intersection correction device 40. Therefore, the frame line recognition unit 108 can correctly recognize the frame line. That is, according to the optical character reading device 50, the table can be correctly recognized.

Further, it can be clearly seen that the character 2320 and the underline 2322 are in the table frame (cell). On the other hand, in FIG. 24 and FIG. 44, since the frame line has disappeared, it is not clearly known whether the character 2320 or the underline 2322 is in the table frame (cell). Thus, according to the optical character reading device 50, the characters in the table can be correctly recognized.

[0079]

According to the apparatus for correcting the intersection of the table frame lines of the present invention, when the intersection of the frame lines of the table having a non-rectangular shape disappears, the intersection can be correctly restored. According to the optical character reading device of the present invention, the accuracy of correctly recognizing characters in a table is improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a table recognition device including a table frame line intersection correction device according to an embodiment of the related art.

FIG. 2 is a PAD diagram of a table extraction process performed by a table frame line intersection correction device according to an embodiment of the related art.

FIG. 3 is a PAD diagram of a process for correcting an intersection of a ruled line of an outer frame of a table by a device for correcting an intersection of a table frame line according to an embodiment of the related art.

FIG. 4 is a PAD diagram of an intersection correction process of an internal ruled line of a table by a table frame line intersection correction device according to an embodiment of the related art.

FIG. 5 is a view showing an example of an input image.

FIG. 6 is an explanatory diagram of contour tracing.

FIG. 7 is an illustration of an image after cutting out a table area.

FIG. 8 is an exemplary diagram of an image after a ruled line is detected.

FIG. 9 is a view showing an example of an image after outer frame ruled lines are extracted.

FIG. 10 is an explanatory diagram of a first example of an intersection correction of an outer frame ruled line.

FIG. 11 is an explanatory diagram of a first example of restoration of an outer frame ruled line.

FIG. 12 is an illustration of an image after the first example of intersection correction of the outer frame ruled line.

FIG. 13 is an illustration of an image after the second example of the intersection correction of the outer frame ruled line.

FIG. 14 is an explanatory diagram of a third example of the intersection correction of the outer frame ruled line.

FIG. 15 is an illustration of an image after the third example of the intersection correction of the outer frame ruled line.

FIG. 16 is an explanatory diagram of a fourth example of the intersection correction of the outer frame ruled line.

FIG. 17 is an illustration of an image after the fourth example of the intersection correction of the outer frame ruled line.

FIG. 18 is a view showing an example of an image after extracting an internal ruled line.

FIG. 19 is an explanatory diagram of an intersection correction of an internal ruled line.

FIG. 20 is a view showing an example of an image after a frame line that has disappeared is restored.

FIG. 21 is a block diagram showing an optical character reading device including a device for correcting intersections of table frame lines according to one embodiment of the prior art.

FIG. 22 is a view showing an example of the image of the table after the disappeared frame line is restored.

FIG. 23 is a view showing an example of an input image without blurring.

FIG. 24 is an illustration of an input image with a blur at intersections.

FIG. 25 is a view showing an example of an image after a ruled line is detected.

FIG. 26 is an illustration of an image after outer frame ruled line extraction.

FIG. 27 is an explanatory diagram of the intersection correction of the outer frame ruled line.

FIG. 28 is an explanatory diagram of a first example of restoration of an incorrect outer frame ruled line.

FIG. 29 is a view showing an example of an image after the first example of the correction of the intersection of erroneous outer ruled lines;

FIG. 30 is an explanatory diagram of a second example of the restoration of an incorrect outer frame ruled line.

FIG. 31 is a view showing an example of an image after a second example of the intersection correction of an incorrect outer frame ruled line;

FIG. 32 is a view showing an example of an image after outer frame ruled lines are extracted.

FIG. 33 is a view showing an example of an image of a table after a frame line that has disappeared is erroneously restored.

FIG. 34 is a PAD diagram of an intersection correction process for an outer frame ruled line by a table frame line intersection correction device according to an embodiment of the present invention.

FIG. 35 is an exemplary diagram of an image after line segment detection.

FIG. 36 is an explanatory diagram of blur correction of line segments.

FIG. 37 is a view showing an example of an image after blur correction of a line segment;

FIG. 38 is a view showing an example of an image after a ruled line is detected.

FIG. 39 is a view showing an example of an image after the outer frame ruled line is extracted.

FIG. 40 is an explanatory diagram of a fifth example of the intersection correction of the outer frame ruled line.

FIG. 41 is an illustration of the image after the fifth example of the intersection correction of the outer frame ruled line.

FIG. 42 is a view showing an example of an image after the outer frame ruled line is extracted;

FIG. 43 is an exemplary view of an image after restoring a disappeared frame line.

FIG. 44 is a view showing an example of an input image in which intersections are blurred.

FIG. 45 is an illustration of an image after line detection.

FIG. 46 is a view showing an example of an image after a ruled line is detected.

FIG. 47 is a view showing an example of an image after extracting an outer frame ruled line;

FIG. 48 is an explanatory diagram of a sixth example of the intersection correction of the outer frame ruled line.

FIG. 49 is an explanatory diagram of a sixth example of the intersection correction of the outer frame ruled line.

FIG. 50 is an explanatory diagram of a sixth example of the intersection point correction of the outer frame ruled line.

FIG. 51 is an illustration of the image after the sixth example of the intersection correction of the outer frame ruled line;

FIG. 52 is a block diagram showing an optical character reading program including an apparatus for correcting intersections of table frame lines according to an embodiment of the present invention.

FIG. 53 is a block diagram showing an optical character reading device including a device for correcting an intersection of table frame lines according to an embodiment of the present invention.

[Explanation of symbols]

10, 40: intersection correction device, 20: table recognition device, 30,
50: Optical character reader, 100, 900, 520
0: control unit, 102: image input unit, 108: frame line recognition unit, 110: frame recognition dictionary, 112: frame line recognition result output unit, 902: character line detection unit, 906: character recognition dictionary, 9
04: character recognition unit, 908: character recognition result output unit.

 ──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Masafumi Koga 1-280 Higashi Koikekubo, Kokubunji-shi, Tokyo Inside the Hitachi, Ltd. Central Research Laboratory (72) Inventor Kazuki Nakajima 1-280 Higashi Koikekubo, Kokubunji-shi, Tokyo Hitachi Central Research Laboratory

Claims (2)

[Claims] An intersection correction device for recovering an intersection of a table frame line that has disappeared when the intersection point of the table frame line in the image has disappeared. Line segment detecting means for detecting, blur correction means for correcting blurring of detected line segments, ruled line detecting means for detecting horizontal ruled lines and vertical ruled lines of the table from the line segments, and outer frame ruled lines of the table from the detected ruled lines Frame line extraction means for extracting a line, an intersection point missing outer frame rule line detecting means for detecting an intersection missing outer frame rule line having an end point at which an intersection point with another outer frame rule line is missing, and an intersection point missing on the end point where the intersection point is missing An intersection missing outer frame rule extending means for extending the outer frame rule; and an extension line of the intersection missing outer frame rule intersects with another outer frame rule. If the extension line is shorter than a predetermined length, the intersection point is excluded. A first outer frame line intersection revival means which is an intersection between frame ruled lines; If the extension of the frame rule does not intersect with the other frame rule, another frame rule extension means for extending the other frame rule, an extension of the intersection-missing frame rule and extension of the other frame rule. If the line intersects, and the sum of the length of the extension of the intersection missing outer frame rule and the length of the extension of the other outer frame rule is shorter than a predetermined length, the crossed point is defined by the outer frame rules. An intersection correction device for a table frame line, comprising: a second outer frame line intersection restoration means serving as an intersection. 2. An intersection correction device for a table frame line according to claim 1, image input means for inputting an optically read image optically read to said intersection correction device, and an intersection corrected by said intersection correction device. Frame line recognition means for recognizing a table frame line based on the frame line;
Character line recognizing means for recognizing a character line in a frame based on the recognition result by the frame line recognizing means; and character recognizing means for recognizing a character in the character line based on the recognition result by the character line recognizing means. And a character recognition result output means for outputting a result of recognition by the character recognition means.