JPH0660221A - Area extracting method for document image - Google Patents

Abstract

PURPOSE:To set an appropriate integrated threshold to the document image of a newspaper, etc., and to exactly integrate a text character area and other character area. CONSTITUTION:With respect to an area image S1 generated by an area image generation processing 1, a level image S2 is formed by a label image generation processing 2. In a character decision processing 3, a specific character in a document image and other character are discriminated and in a character image generation processing 4, as for a specific character, a first character image S4a is formed and as for other character, a second character image S4b is formed. In an integrated threshold setting processing 5, an integrated threshold S5 is derived from a first character image S4a. In such a manner, an appropriate integrated threshold is obtained. In an area extraction processing 6, an integrated processing is executed separately to a first and a second character image S4a, S4b by using the integrated threshold S5. Thus, the area extraction of the document image of a newspaper, etc., can be executed exactly.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a communication device such as a facsimile, a document image database input device, an optical character reader (OCR), etc., for extracting a document image of a newspaper or the like into its component area. The present invention relates to a region extraction method of.

[0002]

2. Description of the Related Art Heretofore, as a method for extracting a region of a document image of this kind, there has been a method described in, for example, Japanese Patent Laid-Open No. 62-71379. In the document image area extraction method described in this document, the document image data is input, black pixels are counted in the scanning direction (for example, the horizontal direction) to detect lines exceeding a threshold value, and the count value is equal to or less than the threshold value. The first area is cut out by determining the state in which a predetermined number of continuous white lines continue.
Within this first region cutout, black pixels are counted in the sub-scanning direction (for example, in the vertical direction) to detect columns in which the count value exceeds the threshold value, and a predetermined number of white columns in which the count value is less than or equal to the threshold value are consecutive. The second area is cut out by determining the state to be performed.

Further, in the second area cutout, the third area cutout is performed by the same processing as the first area cutout, and in the third area cutout, the same processing as the second area cutout is performed. Thus, the fourth region is cut out.
Therefore, for the area detected by the fourth area cutout, the attributes of the areas of the character portion, the photograph portion, and the graphic portion are discriminated based on the run length information and the black pixel ratio information of the area.

[0004]

However, the document image area extraction method having the above configuration has the following problems. (A) In the conventional method, when the area is divided and extracted,
It was necessary to arbitrarily set various thresholds according to the division target. For example, when performing region segmentation processing on individual journals, it is necessary to reset the thresholds to appropriate values according to each journal, and there is the problem that it takes time to select those values. It was (B) In order to solve the above problem (a), the applicant of the present application has previously described in Japanese Patent Application No. 3-52846 (Proposal 1) and Japanese Patent Application No. 3-195437 (Proposition 2). , And proposed the area extraction method.

In the first proposal, first, an all-black multivalued area image of the same size as the input monochrome binary image data is created. Next, the original image is scanned in the horizontal and vertical directions, and if there are white runs equal to or greater than the thresholds T1 and T2 in each direction, the portion corresponding to the white runs in the area image is made white to form the document image. The element is represented as a black pixel connected area. The black part of the area image is labeled, a number is uniquely assigned to each black connected area of the area image to create a label image, and the area division is performed using this label image. . However, with this method, the character area can be expressed only in a small area such as a line unit or a character unit, and it is difficult to know the document structure, the connection of sentences, and the like when performing character recognition. Therefore, in order to eliminate this defect, in Proposal 2, after performing the region division, an integration threshold is set from the white run width histogram for the character region, and the integration is performed using the integration threshold to extract the region. I'm trying to do. However, when the document image from which the region is to be extracted is, for example, a newspaper, the problems shown in FIGS. 2A and 2B occur.

FIGS. 2A and 2B are explanatory views of the problems of the above-mentioned proposal 2. As shown in FIG. 2A, in a document image containing headline characters A, body characters B, etc., such as a newspaper, the line spacing is narrow and the column spacing is narrow, but it is suitable for such documents. The integrated threshold cannot be obtained. Also, FIG.
As in the area C surrounded by the broken line in (b), there is a problem that different paragraphs are merged, or the area of the body character B and the other character areas such as the heading character A are erroneously integrated. A technically satisfactory area extraction method has not been obtained. The present invention, as a problem that the above-mentioned conventional art has, is that an appropriate integration threshold cannot be obtained for documents such as newspapers, and that the text area and other text areas are erroneously integrated. The present invention provides a method for extracting a region of a document image of a newspaper or the like, which solves the problem that the image is lost.

[0007]

In order to solve the above-mentioned problems, a first aspect of the present invention creates an area image in which a constituent element of a document image is expressed as a black pixel connected area from an area image of the whole black and an original image. An area of a document image for which area image creation processing and label image creation processing for creating a label image in which constituent elements of the area image are uniquely numbered are performed, and the area of the document image is extracted using the label image In the extraction method, a character determination process is performed from the label image to distinguish a specific character from other characters according to the size of the character. Then, in the character image creating process, a first character image in which all areas other than the character area are white from the label image is created for the specific character, and characters other than the specific character are created from the label image. A second character image in which all areas other than the area are white is created.
Then, an integrated threshold value setting process of scanning the first character image to set an integrated threshold value, and a region extraction process of separately extracting an area for the first and second character images using the integrated threshold value. And are executed to extract the area of the document image of a newspaper or the like. In a second aspect of the present invention, in the area extraction process of the first aspect, in the area extraction of the second character image, the integration threshold value is changed to perform the area extraction.

[0008]

According to the first aspect of the present invention, since the area extraction method of the document image is configured as described above, when the document image is created using the label image created by the label image creation process, the character determination process is performed. A specific character is distinguished from other characters depending on the size of the character. In the character image creating process, a first character image composed of a specific character is created, and a second character image composed of characters other than the specific character is created. In the integrated threshold setting process, the integrated threshold is set from the first character image, and the integrated threshold is used to perform the integration process separately for the first character image and the second character image. The area of a document image such as a newspaper is extracted. According to the second aspect, in the area extraction of the second character image, the area extraction is performed with different integration thresholds, so that the area extraction corresponding to the configuration of the document image can be performed. Therefore, the above problem can be solved.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The overall processing (I) of the area extraction method showing the embodiment of the present invention and the contents of each processing (II) will be described below with reference to FIGS. 1 and 3 to 5.

(I) Description of Overall Processing of Region Extraction Method (FIGS. 1 and 3) FIG. 1 is a diagram showing the entire processing contents of the region extraction method, and FIG.
(A), (b) is a figure which shows the example of the character image (for example, newspaper) in FIG. As shown in FIG. 1, first, an area image creating process 1 creates an area image S1 in which constituent elements of a document image are expressed as black pixel connected areas from an entirely black area image and an original image, and a label image creating process 2 is performed. Due to
A label image S2 is created by uniquely giving numbers to the constituent elements of the area image S1.

Next, in the character determination processing 3, for example, the characters that make up the body of the newspaper and the other characters are classified, and the classification result is sent to the character image creation processing 4. In general newspapers,
The size of the characters that make up the text is almost constant, and the characters used for headline characters and the like are obviously larger than the characters that make up the text. Therefore, in the character determination processing 3,
For example, it is possible to easily determine the characters forming the body and the other characters by the size of the characters. In the character image creation process 4,
With respect to the characters forming the body, a first character image S4a in which all the areas other than the character area are white from the label image S2 is created, and regarding the characters other than the characters forming the body, the areas other than the character area are separated from the label image S2. A second character image S4b that is all white is created, and the first character image S4 is generated.
a is sent to the integrated threshold setting process 5.

In the integrated threshold value setting process 5, the first character image S4a is scanned and the integrated threshold value S5 is set. For example, in the case of the character image as shown in FIG. 3A, the same integrated threshold setting operation as that of the proposal 2 is performed. In the previous proposal 2, FIG.
While the integrated threshold value setting process is performed for a character image in which both elements (a) and (b) are mixed, in the present embodiment, only the character image as shown in FIG. 3A is integrated. Since the threshold setting process 5 is performed, an appropriate integrated threshold S5 can be obtained more reliably. After that, in the region extraction processing 6, the first and second character images S4 are used by using the integrated threshold S5.
Region extraction is performed for each of a and S4b. In this area extraction processing 6, for example, the same area extraction processing as that of the above-mentioned proposal 2 is performed separately for the character images as shown in FIGS. 3 (a) and 3 (b). However, in the case of the character image as shown in FIG. 3B, since the block interval is wide, for example, n times the integrated threshold S5 set in the integrated threshold setting process 5 (n; real number)
It is desirable to extract the area with the value of as the integrated threshold.

(II) Details of Processes of (I) (II) (1) Integrated Threshold Setting Process 5 (FIG. 4) FIG. 4 is a process content diagram of the integrated threshold setting process 5 in FIG. In FIG. 4, first, the character image creation process 4 in FIG.
The data 51 of the first character image S4a created by
The white run distribution in the horizontal direction is extracted. The same operation is performed in the vertical direction. Next, a white run width histogram creation process 53 creates horizontal and vertical histograms from the white run distribution. Then, the white run width detection process 54 determines the continuous white run width that gives the maximum value of this histogram, and then the integrated threshold value determination process 55 sets an integrated threshold value S5 composed of horizontal and vertical integrated threshold values.

(II) (2) Region Extraction Process 6 (FIG. 5) FIG. 5 is a diagram showing the contents of the region extraction process 6 in FIG. First, the first and second character images S4a and S4b are each horizontally divided by the horizontal division processing 61, and thereafter, the vertical division processing 62 and the horizontal division processing 64 are alternately performed.
In the determination processes 63 and 65, the first and second character images S4a,
For S4b, it is determined whether the first block number obtained by the horizontal division and the second block number obtained by the vertical division do not match and the number of blocks of the vertical division is equal to the number of blocks of the horizontal division. At that time, the area extraction processing for the first and second character images S4a and S4b is completed.

As described above, in the present embodiment, in the character determination processing 3 of FIG. 1, for example, the characters constituting the text of the newspaper are classified into the characters constituting the text and the other characters, and the first character is created from the characters constituting the text. Since the integrated threshold value S5 is set in the integrated threshold value setting process 5 for the character image S4a, the appropriate integrated threshold value can be obtained. Moreover, in the region extraction processing 6, the first and second character images S4a, S created from two types of characters.
Since the regions are separately extracted for 4b, the above proposal 2
However, while there is a risk that the text area and the other text areas may be erroneously integrated, the method of the present embodiment allows the text areas to be correctly integrated. The present invention is not limited to the above embodiment, and various modifications can be made. For example, in the area extraction processing 6 of FIG. 5, the vertical division processing 62 is performed, then the horizontal division processing 61 is performed, and then the processing is alternately repeated, or the area extraction processing 6 is performed by a method other than FIG. And further, the integrated threshold value setting process 5 by a method other than FIG.
May be performed. The area extraction method of this embodiment can be applied to general document images other than newspapers.

[0016]

As described in detail above, according to the first aspect of the present invention, a specific character in a document image is distinguished from other characters by a character determination process, and a specific character is extracted from the specific character by a document image creation process. Since the integrated threshold is set in the created first character image by the integrated threshold setting process, an appropriate integrated threshold can be obtained. Moreover, since the area extraction processing separately performs area extraction on the first and second character images created from two types of characters, that is, a specific character and other characters, the first character area and the second character image are extracted. It can be integrated correctly without accidentally integrating with the character area of. According to the second invention, in the region extraction of the second character image, by using a value different from the integrated threshold used in the region extraction of the first character image, a document image of a newspaper or the like having a wide block interval. Therefore, it is possible to extract an accurate area.

[Brief description of drawings]

FIG. 1 is a processing content diagram of a document image area extraction method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a problem of the previous proposal 2.

FIG. 3 is a diagram showing an example of the character image of FIG.

FIG. 4 is a diagram showing the processing content of integrated threshold setting processing 5 in FIG.

5 is a diagram showing the processing contents of a region extraction processing 6 in FIG.

[Explanation of symbols]

 1 area image creation processing 2 label image creation processing 3 character determination processing 4 character image creation processing 5 integrated threshold setting processing 6 area extraction processing S1 area image S2 label images S4a, S4b first and second character images S5 integrated threshold

Claims (2)

[Claims] 1. An area image creating process for creating an area image in which a constituent element of a document image is expressed as a black pixel connected area from an entirely black area image and an original image, and a number is uniquely given to the constituent element of the area image. A label image creating process for creating a label image, and a document image area extracting method for extracting an area of the document image using the label image, wherein a specific character according to the size of the character is extracted from the label image. A character determination process for determining other characters, and a first character image in which all areas other than the character area are white from the label image for the specific character, and at the same time, for the character other than the specific character A character image creating process for creating a second character image in which all areas other than the character area are white from the label image; and scanning the first character image to set an integrated threshold value. Performing an integrated threshold value setting process and an area extraction process for separately performing the area extraction on the first and second character images by using the integrated threshold value. A method for extracting a region of a document image as a feature. 2. The document image region extraction method according to claim 1, wherein in the region extraction processing, the region extraction is performed by changing an integration threshold in the region extraction of the second character image.