JPH09297765A - Document image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically extract the section effective for grasping contents out of a document image by extracting and area, of which respectively divided elements directly express the contents of the document image, based on the features on the layout of these elements. SOLUTION: The document image is provided by an image input means 101. Next, an area dividing means 102 divides the inputted document image into a character area and the elements such as drawing, table or ruled lines. A line extracting means 103 extracts lines out of the extracted character area. Then, when the area is divided and line information is extracted, a key area discriminating means 104 discriminates a key area. In this case, the key area is the section effective for grasping the contents of a document. Generally, such a section is emphasized by some kind of method. For example, large characters or the fonts of an emphasis system are used or processing such as making that section independent of the other sections or surrounding it with a frame is performed. While utilizing such a feature on the layout expression of the document, the key area is discriminated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document image processing method for automatically extracting a region useful for grasping the content from a document image.

[0002]

2. Description of the Related Art In recent years, the opportunities for distributing document images have increased dramatically due to improvements in the capabilities of computers and digital devices and the capabilities of data storage devices such as hard disks. However, since the document image has a large amount of data, it has become a bottleneck when processing speed is required or when image data is exchanged through a network.

Generally, when a large amount of document images are browsed, it is not necessary to see all the images in the accumulated high-quality image state, and the contents can be checked for the time being. Often enough. Of course, it is much faster to perform a keyword search or the like than to handle an image. However, it is difficult to enter the appropriate keyword. It is even more difficult when the contents of the stored image have never been seen.

Therefore, in an optical filing apparatus or the like, the reduced image is held as an index image separately from the original image, and the index image is first presented to the user to reduce the amount of data to be processed. There is. However, if the entire image is simply reduced, it is difficult to read the characters in the image even if you can get a feel for the whole image.
In particular, when there are many documents having similar layouts, it is difficult to select a desired document.

As a solution to such a problem, there is a document processing apparatus described in Japanese Patent Laid-Open No. 5-342326. In this apparatus, a document image is divided into areas, logical identifiers are given to the divided elements in accordance with a logical model, only required logical elements are extracted using the identifier as a key, and they are rearranged so that they can be seen easily. Also partially OC
By using R, sorting of the extracted information is also performed.

[0006]

However, in the above-described document processing method, it is necessary to grasp the layout configuration and logical configuration of the document image input in advance and create the corresponding model. Documents are organized according to the model,
If the area dividing unit is perfect, accurate processing can be performed, but in many cases this is not the case, and the applicable range of the model is limited. Further, since the creation of a logical model requires considerable proficiency, it is very difficult to extract a partial image (key area) in order to grasp the contents of a new document group by the above processing method.

The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a document image processing method capable of automatically extracting a portion (key area) useful for grasping the content from a document image. To provide.

[0008]

In order to achieve the above object, according to the invention of claim 1, a document image is divided into a plurality of elements, and the elements are based on layout characteristics of the respective divided elements. Is determined to be a region that directly represents the contents of the document image, and the region is extracted as a partial image.

According to the second aspect of the invention, as the layout feature, an emphasized portion different from the text of the document image is used.

According to the third aspect of the invention, the emphasized portion has a character size different from the basic character size of the entire page.

According to a fourth aspect of the invention, the emphasized portion has a font feature different from the font feature of the entire page.

According to a fifth aspect of the invention, the emphasized portion is a position corresponding to a title.

According to a sixth aspect of the invention, the emphasized portion is a position corresponding to a subheading.

The invention according to claim 7 is characterized in that the emphasized portion is a bibliographic item at the top and bottom of a page.

The invention according to claim 8 is characterized in that the emphasized portion is a character string in a surrounding frame.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 shows the configuration of an embodiment of the present invention. In the figure, 101 is an image input means, 10
Reference numeral 2 is an area dividing means for dividing the document image into elements, 103 is a line extracting means for extracting a line in each character area, 104 is a key area for discriminating the key area from the information extracted by the area dividing means 102 and the line extracting means 103. Discrimination means, 105
Is a data storage unit that stores an input image and various kinds of information being processed, 106 is a control unit that controls the whole, and 107 is a data communication path.

FIG. 2 shows a processing flowchart of the present invention. Hereinafter, the present invention will be described with reference to FIG. First, a document image is obtained by the image input means 101 (step 20).
1). The image input means may be a scanner or a facsimile, or may be a means for obtaining an image from another device via a network.

Next, the area dividing means 102 divides the input document image into character areas and elements such as figures, tables and ruled lines (step 202). As such a region division method, for example, a publicly known technique described in JP-A-6-20092 may be used. The extracted area has, as attributes, the element type (character area, table, etc.) and its position (circumscribing rectangle of the area).

The line extracting means 103 extracts lines from the extracted character area (step 203). Examples of this line extraction method include, for example, a paper of the Institute of Electronics and Communication Engineers of Japan, "Regional Segmentation of Document Image Using Marginal Distribution, Cotton Density, and Circumscribed Rectangle Feature" (Akiyama et al., August 1986, Vol. J69-DNo. 8).
The technique described in 1. may be used.

When the area is divided and the row information is extracted,
The key area determination means 104 determines the key area (step 204). Here, the key area is a portion useful for grasping the content of the document. Generally, such portions are emphasized in some way. For example, processing such as using a large character or an emphasis font, making it independent of other parts, or enclosing it in a frame is performed. In the present invention, the key area is discriminated by utilizing such a feature in the layout expression of the document.

FIG. 3 is a detailed processing flowchart of the key area discrimination processing. First, the character size feature of each line is detected (step 301). FIG. 4 shows step 301
It is a detailed flowchart of (character size feature detection processing). First, the basic character size of the entire page is detected (step 401). This uses a histogram of the heights of the rectangles in the extracted line and sets the most frequent frequency as the basic character size.
At this time, the maximum character size of the character in each line is set as the character size of the line (step 402). FIG. 8 shows an example of line character size.

Then, the line character size is compared with the basic character size and classified into any of "capital line", "medium character line" and "normal character line". A threshold is used for this classification. That is, Th1 (for example, 1.
8) A line with a character size that is at least twice as large is designated as an “uppercase line” (Yes in step 405), and a line with a character string that is No in step 405 is Th2 (for example, 1.3) times larger is designated as a “middle character line” (Yes in step 406)
s). The one that is No in step 406 is defined as a “normal character line”.

Returning to FIG. 3, the font features are then examined (step 302). For this feature, a black pixel density or a run length distribution may be used. Alternatively,
You may use the method described in Unexamined-Japanese-Patent No. 6-208649. The font of each line is estimated from the obtained font characteristics. Accurate identification of fonts here is generally difficult. Therefore, only the basic system / emphasized system may be determined.

Next, the title part is detected (step 303). Basically, the title part is a character part larger than the basic character size. In addition, since the character size and line pitch are different from the text and the positions are far apart, they are usually separated as independent areas. Therefore, as shown in FIG. 5, when it is less than a certain number of lines (Th3) for each region (step 503) and it is a capital line (step 504), it is determined to be a title region.

Then, the subheading portion is detected (step 304). In many cases, the subheadings are not so different in text size from the text, and are often in the same area as the text because they are close to the text. FIG. 6 is a detailed processing flowchart of the subheading portion detection processing.

Therefore, first, the case where the area becomes an independent area is detected, and when the area becomes the same area as the body, the line is divided. In a region where the number of lines is less than Th4 (step 603), when the line in the region is a medium character line (step 604), the line in this region is set as a subheading line. In the area that does not correspond, the subheading in the area is detected (step 605).

The subheading in the area is detected as follows.
That is, if the size or font of the first line of the area is different from the size or font of the line in the middle of the area, the first line is detected as the subheading line. At this time, the subheading determination may be similarly performed for some lines following the first line. FIG. 9 is a diagram illustrating subheading detection.

Next, a bibliographic item detection process (step 30)
Perform 5). Bibliographical material typically exists at the top and legs of the page. Also, unlike the text, the number of lines is small. Therefore,
Detects isolated bibliographical areas above and below the page. Figure 7
For example, 701 is a range in which a character area exists,
702 and 703 are ranges in which bibliographic items exist, 70
4, 705 and 706 are extracted character areas.

Since the character areas 704 and 705 are in the upper and lower bibliographic item existence ranges 702 and 703, respectively, they are extracted as bibliographic item areas. Character area 70
Although 6 is partially included in the lower existence range 703, it is not included therein and therefore is not extracted as a bibliographic item area.

Finally, the processing when the surrounding frame exists is performed (step 306). When a box exists in addition to the text in a page, the text in the box is often a separate article. Therefore, extracting a partial area showing the contents of this portion is useful for understanding the contents. Therefore, the inside of the box is shown in FIG.
Processing similar to that of step 605 is performed. However, the articles in the box may not have subheadings with clear size and font differences. Therefore, it is possible to substitute a subheading by extracting a uniform number of lines from the beginning in the frame.

"Title area", "Subheading area and line", "Bibliographic information", obtained by the above processing
The "first line of the surrounding frame" is the key area.

The present invention is not limited to the above-described embodiment, and the character size information and font characteristics of the entire page obtained in steps 301 and 302 may be obtained at the stage of area division and line extraction. Good. Therefore, those pieces of information may be changed to the configuration obtained by step 203. That is, the area dividing unit 102 and the line extracting unit 103 are configured to extract the above information. Further, the image input unit 101 to the key area discrimination unit 104 and the control unit 106 may be realized by software on one processor.

[0033]

As described above, according to the invention described in claim 1, it is possible to obtain a region useful for grasping the content of a document image without requiring a special model.

According to the second aspect of the invention, the area useful for grasping the content of the document image can be extracted by a simple process.

According to the third aspect of the present invention, it is possible to accurately extract a necessary partial area in various documents without being affected by the variation in character size between documents.

According to the invention described in claim 4, the fonts used vary greatly from document to document. However, since the present invention distinguishes between basic fonts and emphasized fonts within a page, there is font variation. However, it is possible to accurately extract the required partial area.

According to the fifth to eighth aspects of the invention, the extracted area is an area useful for grasping the content in a normal document, and in addition, these areas frequently appear in the normal layout. It is possible to accurately extract a required partial area.

[Brief description of drawings]

FIG. 1 shows a configuration of an embodiment of the present invention.

FIG. 2 shows a processing flowchart of the present invention.

FIG. 3 is a detailed processing flowchart of key area determination processing.

FIG. 4 is a detailed flowchart of step 301 (character size feature detection processing).

FIG. 5 is a detailed flowchart of step 303 (title portion detection processing).

FIG. 6 is a detailed processing flowchart of step 304 (subheading portion detection processing).

FIG. 7 is a diagram illustrating a bibliographic item detection process.

FIG. 8 shows an example of line character size.

FIG. 9 is a diagram illustrating subheading detection.

[Explanation of symbols]

 101 Image Input Means 102 Area Dividing Means 103 Row Extracting Means 104 Key Area Discriminating Means 105 Data Storage Units 106 Control Units 107 Data Communication Channels

Claims (8)

[Claims] 1. A document image is divided into a plurality of elements, and it is determined based on the layout characteristics of each of the divided elements whether or not the element is a region that directly represents the contents of the document image. A document image processing method characterized by extracting the region as a partial image. 2. The document image processing method according to claim 1, wherein a highlighted portion different from the text of the document image is used as the layout feature. 3. The document image processing method according to claim 2, wherein the emphasized portion has a character size different from the basic character size of the entire page. 4. The document image processing method according to claim 2, wherein the emphasized portion has a font feature different from the font feature of the entire page. 5. The document image processing method according to claim 2, wherein the emphasized portion is a position corresponding to a title. 6. The document image processing method according to claim 2, wherein the emphasized portion is a position corresponding to a subheading. 7. The document image processing method according to claim 2, wherein the emphasized portion is a bibliographic item at the top and bottom of the page. 8. The document image processing method according to claim 2, wherein the emphasized portion is a character string in a surrounding frame.