JPH08227440A - Document image processor - Google Patents

Abstract

PURPOSE: To provide the document image processor which gives order to a character area exactly in read order even when a mark indicating the read order is given. CONSTITUTION: This document image processor is equipped with a character area candidate extraction part 5 which extracts a character area candidate from a document image in a processing for extracting a character area from the document area as a preprocessing such as a character recognizing process, a character string direction decision part 6 which decides whether or not the character candidate area is the character area and decides its character direction when so, a read order indication mark detection part 7 which detects the mark indicating the read order of characters in the document, and a read order decision part 8 which gives order to the character area exactly in the read order on the basis of the mark.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention extracts a character area existing in a document image as a preprocessing of character recognition processing for converting a print document into a database, and when a plurality of character areas are extracted, The present invention relates to a document image processing apparatus that sequentially orders each character area in the reading order of a document.

[0002]

2. Description of the Related Art As a basic method for ordering the character areas in a document image in the reading order, first, the character areas in the document image are extracted, and in the case of a horizontally written document, the character areas in the upper left corner are moved to the right. In the case of a vertically-written document, it is conceivable that the character areas are arranged from the upper right character area toward the lower left character area toward the lower character area. FIG. 8A shows a conventional example in which three horizontally written character areas in a document image are correctly ordered by this basic method.

Further, a method improved from this basic method is proposed in Japanese Patent Laid-Open No. 5-282487.
The improvement is that the heading area and its body area in the character image are extracted as a set of areas, and the above-described basic ordering method is applied within each set of areas. Even if the three document areas shown in FIG. 8A are ordered by this improvement method, the result is the same as the above-described basic ordering method.

[0004]

However, in the above-mentioned method of the prior art, as in the example shown in FIG. 8B, it is added to indicate the reading order indicated by the * mark in the figure. Even if there is a mark (hereinafter, referred to as a "reading order instruction mark"), since the character area at the upper left is ranked toward the character area at the lower right, there is a drawback that an erroneous result is output.

The present invention solves the above-mentioned problems and provides a document image processing apparatus capable of correctly ordering a character area in a reading order even when a reading order instruction mark is added. The purpose is to

[0006]

In order to achieve the above object, a document image processing apparatus of the present invention comprises a character area candidate extraction unit for extracting character area candidates from a document image, and a character area candidate is a character area. If there is, and if it is a character area, a character string direction determination unit that determines the character direction, a reading order instruction mark detection unit that detects a mark that indicates the reading order of the characters in the document, and A reading order determination unit that orders the character areas in the reading order based on the mark is configured.

[0007]

Since the document image processing apparatus of the present invention determines the reading order of the character area after detecting the reading order instruction mark, even if the reading order instruction mark is added, the reading is correctly performed on the character area. Can be ordered in order.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is based on one embodiment shown in the drawings.
This will be specifically described. FIG. 1 is a block diagram showing a part of a document image processing apparatus according to the present invention that executes ordering of character areas. The image input device 1 reads a paper document as binary image data, and the image data is stored in the image memory 2. Then, the processing device 3 extracts the character regions from the image data stored in the image memory 2 by using the working memory 4 at appropriate times and executes the ordering. Processor 3
Includes the character area candidate extraction unit 5, the character string direction determination unit 6, the reading order instruction mark detection unit 7, and the reading order determination unit 8 illustrated in FIG.

The procedure for extracting character area candidates in the image data stored in the image memory 2 in the character area candidate extracting section 5 will be described with reference to the explanatory views of FIGS. 3 (a) to 3 (e). explain. In this embodiment, the significant pixels are black pixels and the background pixels are white pixels.

First, a blank band in which the total length of white runs in the vertical direction is continuous by a preset number or more is detected. In the document image of FIG. 3A, the shaded area of FIG. 3B is detected as a blank band. Next, a blank band in which the total length of white runs in the horizontal direction is continuous by a preset number or more is detected. In the document image of FIG. 3A, the shaded area of FIG. 3C is detected as a blank band. Then, the document image is divided by the detected blank band, and character region candidates are extracted as shown by the broken line in FIG. Similarly, for each of the extracted character candidate regions, vertical and horizontal blank bands in each region are detected, and when a blank band is detected, the character region candidate is further divided.

In this way, the above process is repeated until the blank band is detected and disappears. In the document image of FIG. 3A, character region candidates are finally extracted as shown in FIG.

Next, in the character string direction determination unit 6, whether each character region candidate extracted by the character region candidate extraction unit 5 is a character region, and if it is a character region, whether it is horizontal writing. The procedure for determining whether the writing is vertical will be described with reference to the explanatory diagrams of FIGS. 4A and 4B.

First, the vertical and horizontal projections of each extracted character region candidate are detected. Next, FIG. 4 (a)
When the projections in the horizontal direction have periodical peaks and valleys, it is determined that the character area is horizontal writing, and as shown in FIG. 4B, the projections and valleys in the vertical projection have periodicity. If it is, it is determined that it is a vertically written character area, and in other cases, it is determined that it is not a character area. Also, the width of the projection mountain is
Since the height of the character is shown in the case of horizontal writing and the width of the character is shown in the case of vertical writing, the width of the projection mountain is set as the estimated character size.

Next, with respect to the procedure of detecting the reading order instruction mark in each character area extracted by the character string direction determining section 6 in the reading order instruction mark detecting section 7, the flowchart of FIG. 5 and FIG. And FIG. 6A and FIG. 6B.
This will be described with reference to the explanatory diagram of FIG.

First, a vertical full-length white run is detected in a horizontally written character area, and a horizontal full-length white run is detected in a vertical written character area (step 501). It is determined (step 502). If the full length white run is not detected, it is determined that the reading order instruction mark does not exist (step 503). FIG.
In the three horizontally written character areas shown in (e), the full-length white run shown in FIG. 6 (a) is detected. If a full-length white run is detected, a black pixel block in the narrower area divided by the full-length white run is detected (step 504),
It is determined whether or not there is only one black pixel block having the size of the estimated character size (step 505). And 1
If only one exists, it is extracted as a reading order instruction mark candidate (step 506). In other cases, it is determined that the reading order instruction mark does not exist (step 503). Since there may be two or three marks, in step 506, a case where a small number of black pixels exist locally adjacent to each other can be used as a candidate.

In the case of FIG. 6A, the two marks shown in FIG. 6B are extracted as reading order instruction mark candidates. Next, the number of reading order instruction mark candidates extracted from each character area is totaled (step 507), and it is determined whether two or more are extracted (step 508).

When the total number is 1, it is determined that the reading order instruction mark candidate is not the reading order instruction mark (step 503). When the total number is two or more, the pattern matching is performed for every two reading order instruction mark candidates (step 509).
Extract a set of marks that have the same size and shape. As an example of a pattern matching method, the superposition method is described in “Pattern Recognition, Kenichi Mori, Electronic Information Communication Society, pages 47 to 48”.

It is determined whether such a set of marks has been extracted (step 510), it is determined that the extracted set of marks is a reading order instruction mark (511), and other reading order instruction marks are determined. It is determined that the candidate is not the reading order instruction mark (step 503).

Instead of pattern matching, character recognition may be executed and a set of marks having the same recognition result may be determined as a reading order instruction mark. For example, FIG. 6 (b)
Since the two marks shown in 1 are the * marks having the same size and shape, it is determined that they are reading order instruction marks.

Next, the reading order determination unit 8 uses the reading order instruction marks extracted by the reading order instruction mark detection unit 7 to sequence the character areas extracted by the character string direction determination unit 6. Will be described with reference to the explanatory views of FIG. 3 (e) and FIG. 7.

Basically, as shown in the prior art, in the case of a horizontally written document, the character area moves from the upper left character area to the lower right character area, and in the case of a vertically written document, the upper right character area moves to the lower left character area. Order toward the text area. However, if there is a character area with a reading order instruction mark,
Next, the character areas to which the same reading order instruction mark is added are ordered.

In the case of FIG. 3 (e), since it is a horizontally written document, first, the reading order of the character area in the column on the left is 1
Set to. Since the reading order instruction mark is added to this character area, 2 is set next in the reading order of the lower character area of the right column to which the same reading order instruction mark is added. Finally, 3 is set in the reading order of the upper character area of the remaining right column. As a result, as shown in FIG.
It is correctly ordered according to the reading order.

The character area and its reading order are output as described above.

As described above, according to the present invention, even when a mark indicating the reading order of characters is added to a document, the character areas are correctly ordered in the reading order. Is possible. Further, for example, when recognizing and outputting the characters in the extracted character area, the recognition result can be output in the reading order.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a portion that executes ordering of character areas of a document image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing functions of a processing device in the document image processing device of FIG.

FIG. 3 is an explanatory diagram of character area candidate extraction processing according to the present embodiment.

FIG. 4 is an explanatory diagram of the same character string direction determination processing.

FIG. 5 is a flowchart of a reading order instruction mark detection process.

FIG. 6 is an explanatory diagram of a reading order instruction mark detection process.

FIG. 7 is an explanatory diagram of a processing result of ordering the same character areas.

FIG. 8 is an explanatory diagram of a processing result of ordering character areas in a conventional example.

[Explanation of Codes] 1: Image Input Device, 2: Image Memory, 3: Processing Device,
4: Working memory, 5: Character area candidate extraction unit, 6: Character string direction determination unit, 7: Reading order instruction mark detection unit, 8: Reading order determination unit

Claims (1)

[Claims] 1. A character region candidate extraction unit for extracting character region candidates from a document image in a process of extracting a character region in a document image as preprocessing such as character recognition processing, and the character region candidate is a character region. Whether or not, and if it is a character area, a character string direction determination unit that determines the character direction,
A reading order instruction mark detecting section for detecting a mark indicating a reading order of characters in the document, and a reading order determining section for ordering the reading order in the character area based on the mark. Document image processing device.