JP2618468B2 - Document processing device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a document processing apparatus capable of recognizing a table in a document.

[Conventional technology]

Conventionally, as this type of table recognition technology, a ruled line is extracted and characters are recognized in an area surrounded by the ruled line.
There is a method in which a format (table structure, character area, etc.) is specified in advance, and recognition is performed in accordance with the format.

In the former method, there are various methods such as a projection method and a tracking method as a ruled line extraction method. In any case, it is necessary that the table elements (characters and the like) are surrounded by the ruled line. In this method, different tables can be recognized for each sheet, and a format can be designated.

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

[Problems to be solved by the invention]

FIG. 10A shows an example in which a name and an address are described.In such a case, only the regions R1 to R4 as shown in FIG.
There is a problem that the correspondence between the address and the address 22 becomes unknown.

On the other hand, the latter method of specifying the format as shown in Fig. 10C can cope with Fig. 10A, but it cannot support the format shown in Fig. 10D, and there is a problem that it is necessary to create a new format. .

[Means for solving the problem]

An image input unit that reads a document, a first storage unit that stores the read image data, a character recognition unit that processes the image data in the first storage unit to recognize characters, and compresses the image data. A compression unit, a second storage unit for storing the compressed image data, and extracting at least a ruled line and a blank line in a predetermined area of the table based on the contents of the second storage unit, And a table recognizing unit for recognizing the table structure from the blank line and the blank line. Characters in an area formed by the ruled line and the blank line are recognized by the character recognizing unit.

[Action]

For example, as shown in FIGS. 10A and 10D, the table area is divided into a fixed area 23 and a variable area 24 as shown in FIG. 10E so that a table having the same ruled line structure but different portions without ruled lines can be recognized.
By extracting blank lines having the same meaning as the ruled lines from the variable area 24 as shown by reference numeral 25 in FIG. 10F,
Make the correspondence between each element such as name and address without error,
To be able to respond flexibly to documents of different formats. At this time, the variable area is set to 24A (9th A
Figure 9A, to divide into 24B (see Figure 9) and 24B (see Figure 9)
The ruled lines are extracted as shown in FIG. 9B. Blank lines obtained from the region 24B and the region 24A do not necessarily match, but by extracting only blank lines that match each other, a blank line having the same meaning as a ruled line as indicated by reference numeral 25 in FIG.10F is obtained. Can be. Thereafter, when performing the continuous processing, at least the structural data of the variable area and the ruled line are stored in a nonvolatile memory or the like so that they can be reused.

For a table having no ruled line, the entire table recognition area can be set as a variable region and the position of the vertical ruled line can be specified separately.

FIG. 1 is a block diagram showing an embodiment of the present invention.
2 is an image input unit, 3 is an image compression unit, 4A and 4B are image memories, 5 is a character recognition unit, 6 is a table recognition unit, 7 is a host CPU,
8 is a CRT and 9 is a keyboard. In addition, 2A, 2B, 3, 4, 5
A, 5B, 6A, 6B, 7A, 7B and FIG. 8 are explanatory views for specifically explaining the present invention.

The document 1 is input from the image input unit 2, one of which is stored in the image memory 4A as it is, the other passes through the image compression unit 3, and the compressed image is stored in the image memory 4B. The compression ratio is selected by the host CPU 7 from, for example, a ratio of 1/2 ⁿ (n = 0, 1, 2, 3). The table recognizing unit 6 uses the compressed image in order to reduce the amount of data stored in the memory 4B and increase the processing time.

After storage, the image data is displayed on CRT8.
A table recognition area indicated by reference numeral 10 in FIG. 2A is designated. In the case of vertical ruled line extraction, the table recognizing unit 6 reads out the area 10 as shown in FIG.
Is divided into horizontally long strips, and projection is performed for each of the strips 11A to 11D.
For example, in the 11B band, since the projection data is as shown in FIG. 3 (b), it is binarized as by connexion Figure 3 the threshold level T _H (ii). Thereafter, to separate the characters and ruled lines in the threshold T _W width, obtaining a third view of (c). FIG. 4 shows the result extracted from the band of FIG. 2B. Thereafter, when the horizontal ruled line is processed in the same manner, the extraction of the ruled line is completed.

Next, after extracting the ruled line 26 to determine a blank line,
Processing similar to the projection method at the time of ruled line extraction is performed on the area.
That is, as shown in FIG. 5A, the region is divided into a plurality of band regions in the vertical direction, and the projection value is obtained for each of the regions. However, in the data to be extracted at this time, a section 12 (blank line candidate) of “0” is extracted in reverse to the case of the ruled line, and the blank line 25 is changed to the 5B Obtain as shown in the figure. Also, for the item data area, the ruled line 26 is designated as shown in FIG. 6A, and blank line candidates 12 are similarly obtained. The W region group 13 is obtained as shown in FIG. 6B. The results are shown in FIG. 7A.
Next, only the blank line 25 where the obtained blank lines 25 and 13 match is left as shown in FIG. 7B. Thus, 25A in FIG. 7A is erased. A rectangular character area 14 separated by ruled lines and blank lines is obtained as shown in FIG. 8, and its coordinates are sent to the character recognition unit 5 to perform character recognition. After that, the recognition results are sent from the recognition units 15 and 16 to the host CPU 7, and are displayed, filled,
Perform transmission, etc.

Next, when a similar table comes, after the ruled line is extracted, for example, in this example, the variable region is the second and third lines from the top of the horizontal ruled line, the region between the first and third lines from the left of the vertical ruled line, By using the previous designation and processing result, recognition can be automatically and continuously performed. To this end, it is needless to say that it is necessary to store at least the processing area (variable area) and the structure data of the ruled lines in a nonvolatile memory or the like.

[Effects of the Invention] According to the present invention, it is possible to recognize a portion having no ruled line, and it is also possible to continuously recognize a table in which a portion having no ruled line is variable. In this way, in the past, it was necessary to specify a format for each sheet, but in the present invention, once it is set, it can be handled automatically, so that the processing time can be shortened and the burden on the operator can be greatly reduced. it can.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2A is an explanatory diagram for explaining a table recognition area, FIG. 2B is an explanatory diagram for explaining a vertical ruled line extraction band area, and FIG. FIG. 4 is an explanatory diagram for explaining projection data and a processing method thereof, and FIG.
Explanatory diagram for explaining vertical ruled lines extracted from FIG. 2A,
FIG. 5A is an explanatory diagram for explaining a blank line extraction band region,
FIG. 5B is an explanatory diagram for explaining all blank lines extracted from FIG. 2A, FIG. 6A is an explanatory diagram for explaining a blank line region in the item data area, and FIG. 6B is a blank in the item data region. FIG. 7A is an explanatory diagram for explaining a row extraction method, and FIG.
FIG. 7B is an explanatory diagram for explaining a blank line area extracted from the item data area of FIG. 7, FIG. 7B is an explanatory diagram for explaining a blank line finally extracted from FIG. 2A, and FIG. FIGS. 9A and 9B are explanatory diagrams for explaining the item data area, and FIGS. 10A, 10B, 10C, 10D, and 10E.
10F and 10F are explanatory diagrams for explaining the problems of the conventional system. Reference numeral 1... Document, 2... Image input unit, 3.
A, 4B: image memory, 5: character recognition unit, 6: table recognition unit, 7: host CPU, 8: CRT, 9: keyboard,
10: table recognition area, 11A to 11D: band, 12: blank line candidate, 13: blank line area, 14: character frame, 21: name data, 22: address data, 23: fixed Area, 24: Variable area, 24A, 24B: Data area by item, 25: Blank line, 26
...... borders, T _H, T _W ...... threshold, D ...... distance, W ...... region width.

Claims (3)

(57) [Claims] An image input unit for reading a document; a first storage unit for storing the read image data; a character recognition unit for processing the image data in the first storage unit to recognize the document; A compression unit for compressing the image data, a second storage unit for storing the compressed image data, and at least a ruled line and a blank line in a predetermined area of the table based on the contents of the second storage unit. A table recognizing unit for extracting and recognizing a table structure from the ruled line and the blank line, wherein a character in an area formed by the ruled line and the blank line can be recognized by the character recognizing unit. A document processing apparatus characterized by the above-mentioned. 2. The document processing apparatus according to claim 1, wherein
The document processing apparatus according to claim 1, wherein the blank line is extracted from a horizontal projection value when the document is horizontal writing, and is extracted from a vertical projection value when the document is vertical writing. 3. The document processing apparatus according to claim 1, wherein
After specifying the processing area and extracting ruled lines and blank lines,
A document processing apparatus wherein continuous processing is enabled based on previously specified and extracted information as long as the structure of a processing area and a ruled line do not change.