JP3162414B2 - Ruled line recognition method and table processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the recognition of ruled lines and the processing of tables in binary images of documents and forms.

[0002]

2. Description of the Related Art In general, when a document or a form is processed by a character recognition device, the image is divided into a character area, an image area (photograph, figure, etc.) and a table area, and each area is subjected to another processing. Many.

Conventionally, in the processing of a table area, a black pixel connection rectangle is obtained from a binary image of the table area, and a rectangle having a certain length or more is recognized as a ruled line. Then, the area surrounded by the ruled lines in the main scanning direction and the sub-scanning direction is recognized as one frame of the table, an image in the frame is cut out, and characters are cut out from this image and recognized. Examples of such processing methods are described in Japanese Patent Application Nos. 1-314519, 1-312868, 2-134876 and 2-14878 by the applicant of the present invention. In the specification and the drawings.

[0004]

According to the conventional method,
If it is not a rectangle longer than a certain length, it will not be recognized as a ruled line.If even a single ruled line such as a dotted line is included in the table, the table frame cannot be recognized properly. However, there is a problem that an error in character recognition in a table occurs. However, in recent years, since tables including non-solid ruled lines are frequently used, demands for normal processing are increasing.

Accordingly, an object of the present invention is to provide a ruled line recognition method and a table processing method for enabling accurate processing of a table including a non-solid ruled line such as a dotted line.

[0006]

According to the Summary of the invention of claim 1, with the binary image for every predetermined number of lines in the main scanning direction and the sub-scanning direction, first extracts a black pixel connecting rectangular, width, and height of each extracted black pixel connected rectangles within a predetermined range as a solid line ruled <br/> line candidate, the distance of the target scanning direction integrates adjacent border candidates in the predetermined range, recognized as solid borders.
Further, a histogram of the width, height, and distance between adjacent rectangles in the target scanning direction is generated for black pixel connected rectangles not recognized as solid ruled lines at this stage, and the width, height, and The respective ranges of the distance between adjacent rectangles are obtained, and among the black pixel connection rectangles that are not recognized as the solid ruled lines, the black pixel connection rectangles in which all of the width, the height, and the distance between the adjacent rectangles are included in the range are determined. Integration
Then, it is recognized as a non- ruled line.

According to the second aspect of the present invention, in the binary image of the table area of the document or the form, the ruled lines in the main scanning direction and the sub-scanning direction are recognized by the ruled line recognition method according to the first aspect. The table frame area is recognized from the information, and characters in the frame area are cut out and recognized.

[0008]

[Action] In the portion of the solid borders are extracted black pixel connected rectangle with a relatively large width (length) and height within a certain range (thickness) and, even if temporarily interrupted, the same The distance between adjacent rectangles belonging to the ruled line is within a certain range.
Therefore, according to the first aspect of the present invention, by extracting ruled line candidates focusing on the width and height of the connected black pixel rectangle, a black pixel connected rectangle having a high possibility of forming a solid ruled line is extracted, and the distance between adjacent rectangles is extracted. , The solid ruled line can be recognized.

On the other hand, in the case of a dotted line, a broken line, or a chain line, the width of the extracted black pixel connection rectangle is smaller and the distance between adjacent rectangles is larger than in the case of the solid line. Depends on line type. However, the ruled line of a dotted line, a broken line, or a chain line is a pattern in which lines having a specific length according to the line type are connected at a specific interval according to a specific rule, so that the width, height, and distance between adjacent black pixel connection rectangles are determined. Takes a peak at a width, height and distance unique to the line type. If the ruled line has an effective length greater than a certain length, the peak value of the histogram exceeds the certain value. Therefore, according to the first aspect of the present invention, a non-solid ruled line such as a dotted line, a broken line, or a chain line can be recognized by such ruled line recognition focusing on the histogram.

According to the second aspect of the present invention, the ruled lines in the main scanning direction and the sub-scanning direction are recognized irrespective of solid lines or non-solid lines, and a table consisting only of solid lines and a table consisting only of non-solid lines are provided. Regardless of a table or a table in which solid lines and non-solid lines are mixed, characters in a frame area surrounded by the ruled lines can be normally cut out and recognized.

[0011]

FIG. 1 is a block diagram of a processing apparatus according to an embodiment of the present invention, FIG. 2 is a flowchart of the processing, and FIGS.
Is a diagram for explaining ruled line recognition.

In the binary image input unit 1, binary data of a document image is input by a scanner or the like and stored in the binary image memory 2 (processing step 100). Table area recognition unit 3
The table area is automatically recognized for the image data in the binary image memory 2, or the table area is recognized by manually specifying the area using a mouse or the like, and the image data of the table area is recognized. It is cut out and stored in the table area image memory 4 (processing step 102).

Next, processing for extracting ruled lines from the table area image data (processing steps 104 to 114) is performed by a predetermined fixed line number (number) in each of the main scanning direction and the sub-scanning direction. (Line) Execute the unit. The contents of this processing are as follows.

In the black pixel connection rectangle generation unit 5, the table area
Attention in the table area image data of the image memory 4
For the fixed number of lines, black pixels connected by tracking black pixels are extracted as rectangles (black pixel connection rectangles) including the black pixels, and information on the black pixel connection rectangles (coordinates and heights of diagonal vertices) , Width, etc.) are stored in the black pixel connection rectangular memory 6 (processing step 104).

The ruled-line candidate extraction unit 7 refers to the data in the black pixel connection rectangle memory 6 and refers to the width (size in the scanning direction of interest) and height (vertical to the attention scanning direction) of the extracted black pixel connection rectangle. the direction size), respectively beforehand
By comparing a predetermined parameter that defines because extracts black pixel connected rectangles within the width and height can be regarded as a candidate of the ruled line (solid line border) as ruled line candidate, and stores the information in the ruled line candidate memory 8 (Processing step 106).

Next, the first ruled line extraction unit 9 refers to the information in the ruled line candidate memory 8 and integrates black pixel connection rectangles whose distance in the scanning direction of interest is within a predetermined integration allowable parameter , and forms a ruled line ( It is recognized as a solid ruled line , and the information is stored in the ruled line memory 10 (processing step 10).
8). This is where the ruled lines intersect in the actual table
In, ruled lines may be divided even with solid lines,
This is performed to correct it. Other than this
Black image extracted as a ruled line candidate in the processing step 106
Regarding the element connection rectangle, the information is stored in the ruled line memory 10.
To pay. At this stage of the process, the solid ruled line is recognized.
And

Next, in the black pixel connected rectangle counting section 11,
The information of the black pixel connected rectangle except for those extracted as the ruled lines (solid line ruled lines) in the processing step 108 is read from the black pixel connected rectangle memory 6, the distance between adjacent rectangles is calculated, and their width and height are calculated. And a histogram of the distance are calculated, and the histogram data is stored in the histogram memory 12 (processing step 110).

The second ruled line extracting unit 13 refers to each histogram data in the histogram memory 12 to determine a range of the width, height and distance where the histogram value exceeds a certain value. For example, when a histogram as shown in FIG. 4 is obtained for the width, the range of the hatched width is obtained. Then, among the black pixel connection rectangles targeted in the processing step 110, the black pixel connection rectangles whose width, height and distance are all included in the respective ranges are integrated, and a ruled line (non-solid line)
(Ruled line) , and the information is stored in the ruled line memory 10 (processing step 112). In this way, the black pixel connection rectangle A corresponding to a non-solid ruled line such as a broken line or a dotted line shown in FIG. 3 can be integrated and recognized as a ruled line.

The above-described ruled line recognition processing loop (processing steps 104 to 112) is repeated for each of the main scanning direction and the sub-scanning direction in units of a fixed number of lines, and when the entire table area is completed (processing step 114), the table recognition is performed. (Processing steps 116 to 126).

First, in the frame recognition section 14, the ruled line memory 1
With reference to the ruled line information in 0, the frames surrounded by the recognized ruled lines in the main scanning direction and the sub-scanning direction are recognized, and the coordinates of each frame (such as the coordinates of two diagonal vertices) are stored in the frame coordinate memory 15. (Step 116). Next, the frame area extracting unit 16 cuts out the image data of each frame area from the table area image data in the table area image memory 4 by referring to the information in the frame coordinate memory 15 and stores it in the frame image memory 17. (Processing step 118). The line cutout unit 18 cuts out line (character string) image data from the data in the frame area image memory 17 by, for example, a method using projection in the main scanning direction and stores it in the line image memory 19 (processing step). 12
0). Next, the character cutout unit 20 cuts out character image data from the line image data by, for example, a method using projection in the sub-scanning direction, and transfers it to the character recognition unit 21 (processing step 122). The character recognizing unit 21 performs character recognition on the character image data by, for example, feature extraction and matching with a recognition dictionary, and outputs the result (processing steps 124 and 12).
6).

[0021]

As described above, according to the present invention, recognition processing focusing on the width and height of a black pixel connection rectangle and the distance between adjacent rectangles, and the width, height, and distance between adjacent rectangles of the black pixel connection rectangle are performed. , It is possible to recognize both a solid ruled line and a non-solid ruled line, and the solid ruled line is cut off in the middle.
It is also possible to perform character recognition by extracting the frame area of a table using a non-solid ruled line using such a method. The range of simple tables is greatly expanded.

[Brief description of the drawings]

FIG. 1 is a block diagram of a processing apparatus according to an embodiment of the present invention.

FIG. 2 is a schematic flowchart of a process.

FIG. 3 is an explanatory diagram of an example of a black pixel connection rectangle corresponding to a non-solid ruled line;

FIG. 4 is a graph showing an example of a histogram of the width of a black pixel connection rectangle.

[Explanation of symbols]

Reference Signs List 1 binary image input unit 2 binary image memory 3 table area recognition unit 4 table area image memory 5 black pixel connection rectangle generation unit 6 black pixel connection rectangle memory 7 ruled line candidate extraction unit 8 ruled line candidate memory 9 first ruled line extraction unit 10 Ruled line memory 11 Black pixel connected rectangle counting unit 12 Histogram memory 13 Second ruled line extraction unit 14 Frame recognition unit 15 Frame coordinate memory 16 Frame region extraction unit 17 Frame image memory 18 Line extraction unit 19 Line image memory 20 Character extraction unit 21 Character recognition Department

Claims (2)

(57) [Claims] About 1. A binary image for every predetermined number of lines in the main scanning direction and the sub-scanning direction, and extracted black pixel connected rectangular,
Width and height are each real black pixel connected rectangles within a predetermined range
Line rule candidates are extracted as line rule candidates, and adjacent rule line candidates whose distance in the scanning direction of interest is within a predetermined range are integrated and recognized as solid rule lines . For black pixel connection rectangles not recognized as solid line rule lines , Generate a histogram of the width, height and distance between adjacent rectangles in the scanning direction of interest, and determine the respective ranges of the width, height, and distance between adjacent rectangles where the histogram value exceeds a certain value,
Among the black pixel connection rectangles not recognized as solid rule lines , the width, height, and distance between adjacent rectangles are all included in the range, and the black pixel connection rectangles are integrated and recognized as rule lines other than the solid line . A ruled line recognition method characterized in that: 2. In a binary image of a table area of a document or a form, ruled lines in the main scanning direction and sub-scanning direction are recognized by the ruled line recognition method according to claim 1, and a table frame area is recognized based on ruled line information. And extracting a character in the frame region and recognizing the character.