JPH02288991A - document processing device - Google Patents

Abstract

PURPOSE:To continuously recognize even a table, for which the part of no ruled line is made variable, by extracting the ruled line and a blank line in the area, which is designated in advance, of the table at least based on compressed image data and recognizing the table. CONSTITUTION:An image input part 2 is provided to read a document and a first storage part 4A is provided to store the image data. Then, a character recognition part 5 is provided to process the image data of this first storage part 4A and to recognize characters and a compression part 3 is provided to compress the image data. Further, a second storage part 4B is provided to store these compressed image data and a table recognition part 6 is provided to extract the ruled line and blank line in the area, which is designated in advance, of the table at least based on the contents of this second storage part 4B and to recognize the table. The characters in the area, which is formed by these ruled line and blank line, are recognized by the character recognition part 5. Thus, even the table, for which the part of no ruled line is different although the constitution of the ruled line is same, can be recognized as well.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a document processing device capable of recognizing tables in a document.

[Conventional technology]

Conventionally, this type of table recognition technology involves extracting ruled lines, recognizing characters in the area surrounded by the ruled lines, and specifying the format (table structure, character area, etc.) in advance, and then There are methods of recognition.

In the former method, there are various methods for extracting ruled lines, such as a projection method and a tracing method, but whichever method is used, it is necessary that table elements (characters, etc.) are surrounded by ruled lines. Note that with this method, it is possible to recognize different tables for each sheet, and it is also possible to specify the format.

On the other hand, in the latter method, by specifying the format regardless of the presence or absence of ruled lines, it is possible to successively recognize tables with the same structure, but it is necessary to specify the format for each table with a different structure.

[Problem to be solved by the invention]

Figure 10A shows an example of writing a name and address; in such a case, the former method, which uses the ruled line structure, would not be able to understand table areas R1 to R4L as shown in Figure 10B.
There is a problem that the correspondence between the name 21 and the address 22 becomes unclear.

On the other hand, the latter method of specifying the format as shown in Figure 10C is compatible with Figure 10A, but cannot support the table shown in Figure 10D, and a new format must be created. There is.

[Means to solve the problem]

an image input section that reads a document; a first storage section that stores the read image data; a character recognition section that processes the image data in the first storage section and recognizes characters; and a character recognition section that compresses the image data. a compression unit, a second storage unit that stores the compressed image data, and a table that extracts at least ruled lines and blank lines within a prespecified area of the table based on the contents of the second storage unit. A table recognition section is provided to allow the character recognition section to recognize characters within the area formed by the outside line and the blank line.

[Effect]

For example, tables like Figures 10A and 10D have the same ruled line structure but different parts without ruled lines! b In order to be recognized, the table area is divided into a fixed area 23 and a variable area 24 as shown in Fig. 10E, and a blank line having the same meaning as a ruled line is extracted from the variable area 24 as shown by reference numeral 25 in Fig. 10F. By doing so, correspondence between each element such as name and address can be made without errors, and it is possible to flexibly handle documents with different formats. At this time, ruled lines are extracted as shown in FIGS. 9A and 9B in order to divide the variable area into 24A (see FIG. 9A) and 24B (see FIG. 9B) according to item data. Although the blank lines found from the area 24B and the area 24A do not necessarily match, by extracting only the blank lines that match each other, a blank line such as the one shown at 25 in FIG. 10F, which has the same meaning as a ruled line, can be obtained. I can do it. From then on,
When performing continuous processing, at least the structural data of the variable area and ruled lines are stored in a nonvolatile memory or the like so that they can be reused.

Note that for a table without any ruled lines, this can be achieved by making the entire table recognition area a variable area and separately specifying the position of the vertical ruled lines.

〔Example〕

FIG. 1 is a configuration diagram showing an embodiment of the present invention, where 1 is a document, 2
is an image input section, 3 to the left 11 is an image compression section, 4A, 4B are image memories, 5 is a character recognition section, 6 is a table recognition section, 7 is a host CPU, 8 is a CRT, 9 is: Keyboard - It is de.

Also, 2nd A, 2B, 3, 4° 5A, 5B, 6A, 6
Figures B, 7A, 7B and 8 are explanatory diagrams for specifically explaining the present invention.

Document 1 is input from image input section 2, one is stored as is in image memo 'J4A, and the other is passed through image compression section 3, and the compressed image is stored in image memory 4B.

The compression ratio is determined by the host CPU 7, for example -(n=0.1
Select from the ratio of +2+'). The table recognition unit 6 uses this compressed image in order to reduce the amount of data stored in the memory 4B and increase processing time.

After storage, the image data will be displayed on the CRT 8, so first a table recognition area as shown at 10 in FIG. 2A is designated.

In the case of vertical ruled line extraction, the table recognition unit 6 divides the area 10 into horizontal strips 7 and 1iA-11D as shown in FIG. 2B.
Take a projection for each. For example, in the case of 11B, the projection data is as shown in FIG. 6(a), so it is converted into two straight lines as shown in FIG. 5(0) by the threshold level T□. Thereafter, the characters and ruled lines are separated by a width threshold [TW 17, to obtain Fig. 6(c).

The results extracted from the cross section of FIG. 2B are shown in FIG. 4. Thereafter, the horizontal outside lines are also processed in the same way, and the extraction of the ruled lines is completed.

Next, after extracting ruled lines 26 to determine blank lines,
The region is subjected to processing similar to the projection method used when extracting ruled lines.

Then, as shown in FIG. 5A, the area is vertically divided into a plurality of band areas, and the projection sensitivity is determined for each band area. However, the data to be extracted at this time is the interval 1 of "0", contrary to the case of ruled lines.
2 (blank line candidate) is extracted, and a blank line 25 is determined from the connection relationship of the data group of this blank line candidate 12 as shown in FIG. 5B.

Also, regarding the item data area, as shown in Figure 6A, the ruled line 2
6, find the blank line candidates 12 in the same way, and D from the bottom of the blank line candidates 12 for the data between the character strings.
A region group 13 having a width W on the upper side is obtained as shown in FIG. 6B.

The results are shown in FIG. 7A. Then you get 9 blank lines 25
Only the blank line 25 where and 13 match is left as shown in FIG. 7B. Therefore, 25A in FIG. 7A is deleted. The rectangular character area 14 separated by ruled lines and blank lines is
The coordinates are determined as shown in the figure and sent to the character recognition section 5 for character recognition. Thereafter, each recognition result is sent from each recognition unit 15, 16 to the CPU 7 and displayed.

Perform filing, transmission, etc.

Next, when a similar table comes, after extracting the ruled lines, for example, in this example, the variable area is the second or third horizontal ruled line from the top,
By using the previous designation and processing results for the area between the first and fifth vertical ruled lines from the left, it is possible to automatically and continuously recognize the area. For this purpose,
Needless to say, it is necessary to store at least the structural data of the processing area (variable area) and ruled lines in a non-volatile memory or the like.

〔Effect of the invention〕

According to the present invention, it is possible to recognize areas without ruled lines,
It is also possible to continuously recognize a table in which the part without ruled lines is variable. As a result, in the past, it was necessary to specify the format for each sheet, but with the present invention, once the format is set, it can be handled automatically from then on, reducing processing time and greatly reducing the burden on the operator. can.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2A is an explanatory diagram for explaining the table recognition area, FIG. 2B is an explanatory diagram for explaining the vertical ruled line extraction band area, and FIG. 4 is an explanatory diagram for explaining the projection data and its processing method, FIG. 4 is an explanatory diagram for explaining the vertical ruled lines extracted from FIG. 2A, and FIG. 5A is an explanatory diagram for explaining the blank line extraction band area. , FIG. 5B is an explanatory diagram for explaining all blank lines extracted from FIG. 2A, FIG. 6A is an explanatory diagram for explaining the blank line area of the item data area, and FIG. 6B is an explanatory diagram for explaining the blank line area of the item data area. An explanatory diagram for explaining the method for extracting blank lines in the data area. Figure 7 is an explanatory diagram for explaining the blank line area extracted from the item data area of Figure 6. Figure 7B is an explanatory diagram for explaining the method for extracting blank lines in the data area.
An explanatory diagram for explaining the blank line finally extracted from Figure A, Figure 8 is an explanatory diagram for explaining the character frame, and Figure 9A is an explanatory diagram for explaining the character frame.
Figures 1 and 9B are explanatory diagrams for explaining the item data area.
10E are explanatory diagrams for explaining the problems of the conventional method. Code explanation 1... Document, 2... Image input section, 5.
...Image compression unit, 4A, 4B...Image memory, 5...Character recognition unit, 6...Table recognition unit, 7... host) CPU, 8...--C
RT,? ...Keyboard, 10...Table recognition area, 11A to 11D... Band, 12...
...Blank line candidate, 16...Blank line area, 14
......Character frame, 21...Name data, 2
2...Address data, 23...Fixed area, 24...Variable area, 24A, 24B...
...Data area for each item, 25...Blank line, 26
・・・・・・Ruled line, T□pTw・・・・Threshold straight,
D: Distance, W: Area width.

Claims (1)

[Claims] 1) An image input section that reads a document, a first storage section that stores the read image data, and a character recognition section that processes the image data in the first storage section and recognizes the document. a compression unit that compresses the image data; a second storage unit that stores the compressed image data; and based on the contents of the second storage unit, at least the ruled lines and blank spaces in a prespecified area of the table. a table recognition unit that recognizes tables by extracting rows and rows; and a document, characterized in that the character recognition unit is capable of recognizing characters within an area formed by the ruled lines and blank lines. Processing equipment. 2) In the document processing device according to claim 1, when the document is written horizontally, the blank line is extracted from a horizontal projection value;
A document processing device characterized in that when a document is written vertically, extraction is performed from a vertical projection value. 5) In the document processing device according to claim 1), after specifying a processing area and extracting ruled lines and blank lines, as long as the structure of the processing area and ruled lines remains unchanged, the information previously specified and extracted is A document processing device characterized in that it is capable of continuous processing.