JPH1166196A - Document image recognition device and computer-readable recording medium where program allowing computer to function as same device is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate documents in various format complying with usages such as a document whose reproduction is given priority and a document whose contents are made important. SOLUTION: This device is equipped with a document image input part 200 which inputs a document image generated by optically reading a paper document, a preprocess part 201 which performs noise removal and a skew correcting process for the inputted document image, an information extraction part 204 which performs a recognizing and extracting process for a character area including character strings and/or a an image area including images of a graph, a table, a photograph or the like, and a character recognizing process for the character strings in the extracted character area and also analyzes the layout of the document image to extract layout information, a document generation part 210 which generates a PostScript document and an HTML document according to the character recognition result and layout information extraction result, and a data base part 213 which stores the generated PostScript document and HTML document respectively.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention not only performs a character recognition process of extracting a character code from a document image obtained by optically reading a paper document, but also performs a character recognition process.
By extracting and using various information of paper documents, it can be used on various types of computers according to the purpose of use, such as documents that prioritize reproduction of paper documents and documents that emphasize contents of paper documents The present invention relates to a document image recognition device capable of generating a document, and a computer-readable record in which a program for causing a computer to function as the device is recorded.

[0002]

2. Description of the Related Art The rapid spread of personal computers and networks has promoted the digitization of documents in offices. However, because paper has good convenience such as good readability, production of paper documents has not stopped.
At present, there are still a large number of documents stored and distributed in paper form, and this gap is one of the causes of a decrease in production efficiency in office work.

[0003] In other words, document creation, distribution, browsing, management,
In a series of flows such as reuse, documents exist in the form of data recorded on paper and data on a computer. Thus, the document exists in the form of data recorded on paper and data on a computer. The reason is that the mutual conversion cost is high.

As one of the means for solving the above problem, O
CR (Optical Character Read)
er). By using the OCR, a paper document is read by a scanner or the like to generate a document image, and then a character string in the document image can be converted into a character code.

[0005]

However, the above O
The CR can convert a character string in a document image into character code information, but cannot extract the layout of the original paper document or the like. Therefore, the document generated by the OCR is simply a text list. Thus, there is a problem that various information included in the original document image cannot be sufficiently used. In other words, the document is subjected to various layout processes according to the creator's intention, but simply using OCR cannot extract the document layout and reproduce the original document layout. It is not possible to generate a new document newly or to generate a document with a new layout using the layout of the original document.

The present invention has been made in view of the above, and not only performs a simple character recognition process of extracting a character code from a document image obtained by optically reading a paper document, but also performs processing of the paper document. Generates documents that can be used on various types of computers according to the purpose of use, such as documents that prioritize reproduction of paper documents and documents that emphasize contents of paper documents by extracting and using various information possessed The purpose is to be able to.

[0007]

According to a first aspect of the present invention, there is provided a document image recognition apparatus comprising: an input unit for inputting a document image generated by optically reading a paper document; Region extracting means for recognizing and extracting a character region including a character string and / or an image region including an image such as a figure, a table, or a photograph from a document image input through the apparatus;
Character recognition means for performing character recognition processing on a character string of a character area extracted by the area recognition means; and layout information for analyzing a layout of the document image based on an extraction result of the area extraction means and extracting layout information Extracting means; first document generating means for generating a first document using a page description language based on a character recognition result by the character recognizing means and a layout information extracting result by layout information extracting means; Based on the result of character recognition by the means and the result of layout information extraction by the layout information extracting means.
And a storage unit for storing the first and second documents generated by the first and second document generation units, respectively.

According to a second aspect of the present invention, in the document image recognition apparatus of the first aspect, the first document is a document expressed in a PostScript format or a PDF format.

According to a third aspect of the present invention, in the document image recognition apparatus of the first aspect, the second document is a document represented by SGML, HTML, or XML.

A computer-readable recording medium according to a fourth aspect of the present invention stores a program for causing a computer to function as each unit of the document image recognition apparatus according to any one of the first to third aspects. It is.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a document image recognition apparatus of the present invention and a computer-readable recording medium storing a program for causing a computer to function as the apparatus will be described in detail with reference to the accompanying drawings. Will be described.

First Embodiment A document image recognition apparatus according to a first embodiment receives a document image generated by optically reading a paper document, and can use the document image on a computer based on the input document image. Generate documents. In other words, in office work and the like, since there is a series of flows of producing, distributing, browsing, managing, and using documents, the document image recognition apparatus according to the first embodiment performs the above-described distribution and browsing. It is possible to generate a document in consideration of and a document in consideration of use.

Here, the document considering distribution and browsing is intended to faithfully reproduce the layout information of the original paper document so that the intention of the document creator can be conveyed as accurately as possible. A document (a document in which WYSIWYG is stored as a metaphor of a paper document). First Embodiment To enable a paper document to be faithfully reproduced by a document on a computer, the first embodiment
The document image recognition apparatus generates a document using a page description language capable of uniformly describing characters and images. In the document image recognition apparatus according to the first embodiment, PostScript is used as a page description language.
Will be used. In the following, this PostScri
PostScript is a document generated using pt
It is defined as a t document. Also, PostScri
PDF (Portable Docum) instead of pt
ent Format: a file format published by Adobe Systems Incorporated).

A document that is considered to be used is a document that has a form unique to a computer that gives priority to the contents of the document, regardless of the layout of the original paper document. As described above, in order to generate a document in which the content is prioritized over the layout, the document image recognition apparatus according to the first embodiment uses an SGML (Standard General).
size Markup Language) and HTML
(Hypertext Markup Langag
e), XML (Extensible Markup)
A document is generated by expressing it in a structured description language such as Language (Language) or the like, and a paper document in which characters and images are mixed is converted into hypertext. Embodiment 1
In this document image recognition apparatus, HTML is used as a structured description language. In the following, this HT
A document generated using ML is defined as an HTML document.

FIG. 1 is a configuration diagram showing a system configuration of the document image recognition apparatus according to the first embodiment. A document image recognition apparatus shown in FIG. 1 includes a color scanner 100 and a monochrome scanner 10 that optically read a paper document to generate a document image.
1 and a network scanner 102, a fax modem 104 for receiving a document image transmitted from a facsimile apparatus 103, and a color scanner 100, a monochrome scanner 101, a network scanner 102, and a fax modem 104 for inputting a document image and inputting the document image. Based on the PostScript document and H
A document image processing server 105a for generating a TML document;
A digital multifunction peripheral 106 that optically reads a paper document to generate a document image, a document image is input from the digital multifunction peripheral 106, and a Post is input based on the input document image.
A document image processing server 105b that generates a Script document and an HTML document, and a document image processing server 105
a, 105b (hereinafter referred to as “document image processing server 10”).
5)), stores the document in a database, and searches and outputs the corresponding document in response to a search request from the client 107.
8 is comprised. Reference numeral 109 denotes a network such as a LAN.

FIG. 2 is a conceptual block diagram of the document image recognition apparatus shown in FIG. The document image recognition device according to the first embodiment mainly includes a document image input unit 200, a preprocessing unit 201, an information extraction unit 204, a document generation unit 210, and a database unit 213. The document image input unit 20
0, the pre-processing unit 201, the information extraction unit 204, and the document generation unit 210 correspond to the document image processing server 105 shown in FIG. 1, and the database unit 213 corresponds to the search server 108 shown in FIG. Hereinafter, the configuration of each unit will be described.

(1) Document Image Input Unit The document image input unit 200 inputs a document image generated by the color scanner 100, the monochrome scanner 101, the network scanner 102, and the digital MFP 106, and a document image received by the fax modem 104. Things. Further, the document image input unit 200 can also input a document file created by an application program such as a word processor.

(2) Pre-processing unit The pre-processing unit 201 inputs a document image via the document image input unit 200, and removes an isolated point noise from the input document image. Skew correction unit 2 that corrects the inclination when the camera is tilted
03. The pre-processing unit 201 performs noise removal and inclination correction on the document image,
This is to remove factors that have an adverse effect on the area division processing and character recognition processing performed later. If the input document image is full-color, binarization processing for binarizing the full-color document image is performed in order to easily and quickly perform the area division processing and the like. You may do it.

(3) Information Extraction Unit The information extraction unit 204 receives a document image from the preprocessing unit 201 and includes a character area including a character string and / or an image such as a figure, a table, or a photograph from the input document image. In addition to performing processing for recognizing and dividing an image area, processing for identifying what column type the document image is composed of, for example, single column, multi-column, or free column, processing for detecting a centering area, etc. A region division / information extraction unit 205 that performs a layout information extraction process, and extracts a table frame and a ruled line structure when a table is included in the image region divided by the region division / information extraction unit 205. A table processing unit 206 for extracting a character region in the frame, and a character recognition process for a character string of the character region divided by the region dividing / information extracting unit 205 and the table processing unit 206. OCR to do
The processing for identifying whether the font of the character string of the character area divided by the section 207 and the area division / information extraction section 205 is an emphasis type (medium Gothic, etc.) or a non-emphasis type (fine Mincho, etc.) It has a font identification unit 208 and a layout / logic analysis unit 209 that analyzes the layout of the document image and analyzes the logical structure of the document based on the result of processing by each unit.

(4) Document Generation Unit The document generation unit 210 generates a PostScript document based on the result of the character recognition processing by the OCR unit 207 and the result of the analysis processing by the layout / logic analysis unit 209. And H
And an HTML document generation unit 212 that generates a TML document.

(5) Database (DB) The database 213 includes a PostScript document DB 214 for storing the PostScript document generated by the PostScript document generation unit 211, and HTML.
An HTML document DB 215 for storing the HTML document generated by the document generation unit 212.

Next, the operation of the document image recognition apparatus having the above-described configuration will be described in detail. FIG. 3 is a flowchart showing an operation procedure of the document image recognition device.

The document image input unit 200 inputs a document image generated by the color scanner 100, the monochrome scanner 101, the network scanner 102, and the digital MFP 106, and a document image received by the fax modem 104 (S301). Also, the client 107
Can input a document image or a document file created by an application program such as a word processor.

FIG. 4 is an explanatory diagram showing an example of a state in which a document image input via the document image input unit 200 is displayed on a screen. 4, reference numeral 400 denotes a control screen for the document image recognition process, and 401 denotes a display screen for displaying a document image input via the document image input unit 200.

On the control screen 400 shown in FIG. 4, it is possible to specify the execution of various processes described below and to make detailed settings for the processes, and to specify the generation of a PostScript document and / or an HTML document. can do. Note that FIG. 4 shows the control screen 400 on which a document image can be input and execution of various processes can be designated.
It is also possible to automatically execute processing up to generation and storage of a document.

Subsequently, the noise removing unit 202 inputs a document image via the document image input unit 200, and removes isolated point noise from the input document image (S302). Also,
If the input document image is inclined, the skew correction unit 203 corrects the inclination of the document image (S30).
3).

An area dividing / information extracting section 205 receives a document image from a pre-processing section 201 comprising a noise removing section 202 and a skew correcting section 203, and outputs a character area including a character string from the input document image and / or a figure, A process for recognizing and dividing an image area including an image such as a table or a photograph is performed (S
304). For each of the divided regions, a character region, an image region, and in the case of an image region, a type of a region such as a diagram, a table, or a photograph and a position of the region are added as attribute information.

After performing the area division processing, the area division / information extraction processing unit 205 identifies which column type the document image is composed of, for example, single column, multi column, or free column. Then, a process of extracting layout information such as a process of detecting a centering region is performed (S305).

More specifically, the distance (blank portion) between the character regions and the ruled lines are detected, and the detected distance and the number of ruled lines are used together with the attribute information on the position of the character region divided by the region dividing process. , The column type is determined.

Thereafter, the table processing unit 206 determines whether or not there is a region including a table based on the attribute information of each region divided by the region dividing / information extracting unit 205 (S30).
6). This determination may be made automatically on the device side as described above, or may be specified by the user. If no table is included, the process proceeds to step S308.

On the other hand, if a table is included, the table processing unit 2
Step 06 extracts the structure of the table frame and the ruled line, and also extracts the character area in the frame (S307). As described above, by extracting the character area from the table, the next OCR unit 20 is extracted.
In step 7, character recognition processing in the table can be performed.

Then, the OCR section 207 performs a character recognition process on the character string of the character area divided by the area dividing / information extracting section 205 and the table processing section 206 (S308).
That is, the OCR unit 207 performs line cutout and character cutout processing on the character area, and performs character recognition processing on each of the character cutout character patterns. In addition, the OCR unit 207 performs error correction by language processing on the character recognition result.

The font identifying unit 208 determines whether the font of the character string of the character region divided by the region dividing / information extracting unit 205 is emphasized (such as middle Gothic) or non-emphasized (such as Ming Mincho) for each line. Is performed (S309). Specifically, the characteristics of the font are identified based on, for example, the distribution of black pixel density and run length.

Subsequently, the layout / logic analysis unit 209
Analyzes the layout of the document image based on the result of the processing by each unit (S310). The layout analysis performed here includes, for example, detection of a title portion, a subheading portion, a caption, and a header / footer portion.

Here, the title is generally different in size and line pitch from the characters in the text, and the position where the title is present is slightly apart from the text. The title part can be detected using the attribute information on the position of the area and the identification result by the font identification unit 208.

The subheadings are often almost the same in size as the characters in the main text, and because they are located near the main text, they often exist in the same area as the main text.
Therefore, when the character size or font of the first line of each character area is different from those of other characters in the same character area, the first line is determined to be a subtitle line.

A caption is given to an image such as a figure, a table, a photograph, etc., and generally exists near the image area and at a position away from the text.
Since a word indicating a diagram, a table, or the like such as “FIG. 8” is included, a row satisfying these conditions is determined as a caption. Then, a link is generated between the caption and the corresponding figure.

Further, since the header / footer exists above and below the document image, the corresponding line is detected as the header / footer. Specifically, for example, when the region identification / information extraction unit identifies the single column, the line above the centering line can be determined to be the header portion. On this occasion,
Also identifies the level of the header row. If the column is identified as a multi-column, the upper and lower rows that do not belong to the column become header / footer sections.

Thereafter, the layout / logic analysis unit 209
Analyzes the logical structure of the document (S311). Specifically, the same word as the word such as “FIG. 8” in the captions given to the figures, tables, photographs, and the like is found in the text, and a process of generating a link between these characters is performed. In this processing, since the result of character recognition by the OCR unit 207 is used, it is possible to deal with an OCR error by using all the candidate characters of each character pattern.

Then, the document generation unit 210, based on the result of the processing performed by each unit of the information extraction unit 204 described above,
A process for generating a document is performed (S312). That is, P
The ostScript document generation unit 211 includes the information extraction unit 2
Based on the results of character recognition and layout analysis performed in step 04, a document expressing each page in PostScript is generated. In order to cover the OCR error, the original image can be used for a character with low character recognition certainty, instead of the character recognition result.

The HTML document generation unit 212 adds a tag indicating a title, a paragraph, etc. to each character string based on the result of character recognition and layout analysis performed by the information extraction unit 204 to generate an HTML document. I do.

After that, the PostScript document becomes Po
The HTML document is H in the stScript document DB 214.
Each is stored in the TML document DB 215 (S31
3). Note that PostSc generated from the same document image
When registering both a RIP document and an HTML document, they may be associated with each other so that one can call the other.

Note that the PostScript document DB 21
4 and the HTML document stored in the HTML document DB 215 can be displayed on the screen of each client 107 by performing a search process. That is, in response to a search request from the client 107, the search server 108
script document DB 214 and HTML document DB 21
5 and retrieves the corresponding document and outputs it.
In step 07, a search result is input from the search server 108, and a document corresponding to the search request is displayed on the screen.

FIG. 5 is an explanatory diagram showing an example of a state in which a PostScript document is displayed on a screen. As shown in FIG. 5, PostScr based on the document image shown in FIG.
By generating a document represented by ipt, a document having the same layout as the original paper document can be easily generated. That is, by generating a document that reproduces a paper document, a document suitable for distribution and browsing that can convey the intention of the document creator as accurately as possible can be obtained. Although the PostScript document is displayed on the screen almost in the same manner as the original paper document, since the internal information is retained, the PostScript document can be searched or reused.

FIG. 6 is an explanatory diagram showing an example of a state in which an HTML document is displayed on a screen. FIG. 6 is a diagram in which the figure and the figure number are converted into hypertext. For example, when "FIG. 9" in the text is clicked with a mouse or the like, "FIG. 9"
Is displayed on the screen. As described above, by converting into a hypertext without being bound by the layout of the paper document, a document having a form unique to a computer in which the contents of the paper document are prioritized can be generated.

As described above, according to the document image recognition apparatus of the first embodiment, not only the simple character recognition process of extracting the character code from the document image but also various information of the document image is extracted. For this purpose, it is possible to generate a document that can be used on various types of computers according to the purpose of use, such as a document that prioritizes the reproduction of a paper document and a document that emphasizes the contents of a paper document.

In FIG. 1, the network 10
Although the configuration of the document image recognition apparatus according to the first embodiment has been described as a system via a computer 9, the function shown in FIG. 2 is provided in one computer to configure the document image recognition apparatus in a stand-alone form. You can also.

In the document image recognition apparatus according to the first embodiment, the above-described PostScript document and HTML document are generated. However, the present invention is not limited to these, and other types of documents may be generated as needed. You can also.

[Second Embodiment] A document image recognition apparatus according to a second embodiment is capable of efficiently searching for a PostScript document or an HTML document generated as described in the first embodiment. is there. More specifically, a predetermined character string is extracted from the character string recognized by the OCR unit 207, and the extracted character string is assigned to the corresponding PostSc.
associated with a .doc document or an HTML document, and by searching for a corresponding character string, the associated Pos
A tScript document or an HTML document can be output as a search result. Hereinafter, this character string is defined as key text.

FIG. 7 is a conceptual configuration diagram of the document image recognition apparatus according to the second embodiment. 7, the same components as those in FIG. 2 described in the first embodiment are denoted by the same reference numerals, and the detailed description thereof will be omitted.

The document image recognition apparatus according to the second embodiment has the structure shown in FIG.
As shown in FIG. 7, a key text extracting unit 700 for extracting the above-described key text from the character string recognized by the OCR unit 207, and a key text extracted by the key text extracting unit 700 and registered in the key text DB 702. A text registration unit 701 and a search request are input, a key text registered in the key text DB 702 is searched, and PostS associated with the corresponding key text is searched.
a search processing unit 703 that outputs a script document or an HTML document as a search result.

Note that the PostScript document DB 21
4, HTML document DB 215 and key text DB 7
02 is provided in the search processing unit 703. This search processing unit 703 corresponds to the search server 108 in FIG.
The search processing is performed based on a search request from the client 107. When the document image recognition device according to the second embodiment is configured in a stand-alone mode, a search request is directly input to perform a search process.

The key text to be extracted by the key text extraction unit 207 is a character string that simply represents the document, for example, the entire document, chapter and section titles, bibliographic items such as headers and footers, and document summaries. A sentence or the like can be considered. Also,
A character string constituting a caption of a figure, a sentence including a figure number, a paragraph including the sentence, and a character string in page units may be extracted as a key text based on a figure or the like in a document. In order to extract the key text, it is necessary to analyze the layout of the document image. Therefore, the key text extraction unit 700 uses the analysis result of the layout / logic analysis unit 209 to perform a key text extraction process. It may be performed.

The key text extraction unit 701 inputs the key text extracted by the key text extraction unit 700, associates the input key text with the PostScript document or HTML document generated by the document generation unit 210, and generates a key text DB 702. To be stored.

Further, the search processing unit 703 is provided with an input / output unit 704 for inputting a search request and outputting a search result.
And a search engine 705 for inputting a search request from the input / output unit 704 and searching for a corresponding key text from the key text DB 702. Specifically, the input / output unit 7
04 inputs a search request and outputs the search request to the search engine 705. The search engine 705 includes an input / output unit 70
4, a search request is input, a corresponding key text is searched from the key text DB 702, and the corresponding key text is output to the input / output unit 704. The input / output unit 704 includes a PostScript associated with the input key text.
t documents and HTML documents are output as search results.

As described above, according to the document image recognition apparatus of the second embodiment, a document search is performed using a character string in a document image as a key text. , And a high-speed search process can be realized.

In the second embodiment described above,
Although the document search is used using the key text, a full-text search may be performed using all the results of character recognition by the OCR unit 207 instead of the key text. FIG. 8 is a conceptual configuration diagram illustrating a modified example of the document image recognition device of the second embodiment.

As shown in FIG. 8, a text registration unit 800 and a full-text search text DB 801 are provided in place of the above-described key text extraction unit 700, key text registration unit 701, and key text DB 702. The character recognition result, that is, the full text of the text in the document image is input from the unit 207, and the input text is associated with the PostScript document or the HTML document generated by the document generation unit 210, and the text DB for full text search is input.
Register at 801. As a result, full-text search text DB
PostScr using the text registered in 801
Ipt documents and HTML documents can be searched. In this case, since the text is registered in the full-text search text DB 801, it is not necessary to open and close each file every time the search is performed, so that the search process can be sped up.

Third Embodiment Next, a document image recognition apparatus according to a third embodiment will be described. The document image recognition device according to the third embodiment can generate a document that considers distribution and browsing and a document that considers use, as in the case of the first embodiment.

The document in consideration of distribution and browsing in the document image recognition device according to the third embodiment is a document image generated by reading an original paper document, and is defined as image data here. In addition, a document in consideration of use is a result of performing character recognition on a character string in a document image, and is defined here as text data.

FIG. 9 is a conceptual configuration diagram of the document image recognition device of the third embodiment. In FIG. 9, the same components as those described in the first embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

The document image recognition apparatus according to the third embodiment includes a document image input unit 200, a preprocessing unit 201, an information extraction unit 204, a registration unit 900, and a database unit 213. The document image input unit 200 and the preprocessing unit 2
01, the information extraction unit 204 and the registration unit 900 correspond to the document image processing server 105 shown in FIG. 1, and the database unit 903 corresponds to the search server 108 shown in FIG.

Document image input unit 200 and preprocessing unit 20
1 is as described in the first embodiment. In the information extraction unit 204, PostScript
Since no t document or HTML document is generated, the configuration is the same as that described in the first embodiment except that the font identification unit 208 and the layout / logic analysis unit 209 are omitted.

The difference from the document image recognition apparatus of the first embodiment is that a registration unit 900 having an image DB registration unit 901 and a text DB registration unit 902 and an image DB 90
4 and a database unit 9 including a text DB 905
03.

The image DB registration unit 901 includes a pre-processing unit 2
01, image data after noise removal and skew correction has been input, and processing for registering the input image data in the image DB 904 of the database unit 903 is performed.
Then, a process of registering text data as a result of performing the character recognition process in the text DB 905 of the database unit 903 is performed. When both the image data and the text data of the same document are registered, they may be associated with each other so that one can call the other.

The image data and text data stored in the image DB 904 and the text DB 905 can be displayed on the screen of each client 107 by performing a search process. That is, in response to a search request from the client 107, the search server 108 stores the image DB 904 and the text DB
The client 107 searches for and outputs the corresponding document from 905, and the client 107 inputs the search result from the search server 108,
The document corresponding to the search request is displayed on the screen.

As described above, according to the document image recognition apparatus of the third embodiment, since the image data is registered in the database, the paper document itself can be displayed on the screen of the computer, and the text data can be stored in the database. Because of the registration, the information in the paper document can be easily used.

[Fourth Embodiment] A document image recognition apparatus according to a fourth embodiment uses the image data and text data generated as described in the third embodiment by using the key text described in the second embodiment. This makes it possible to search efficiently.

FIG. 10 is a conceptual configuration diagram of the document image recognition device of the fourth embodiment. In FIG. 10, the first embodiment
2 and FIG. 7 described in the second embodiment will be denoted by the same reference numerals, and detailed description thereof will be omitted.

The document image recognition apparatus according to the fourth embodiment has the structure shown in FIG.
0, the layout of the document image is analyzed.
A key text extraction unit 700 for extracting the above-described key text from the character string recognized by the character unit 207, and a key text registration unit 7 for inputting the key text extracted by the key text extraction unit 700 and registering the key text in the key text DB 702.
01 and a search request is input directly, and the key text DB 7
A search processing unit 703 for searching for a key text registered in No. 02 and outputting image data or text data associated with the corresponding key text as a search result;
It has.

The key text to be extracted by the key text extraction unit 207 is a character string that simply represents the document, for example, the entire document, chapter and section titles, bibliographic items such as headers and footers, and document summaries. A sentence or the like can be considered. Also,
A character string constituting a caption of a figure, a sentence including a figure number, a paragraph including the sentence, and a character string in page units may be extracted as a key text based on a figure or the like in a document.

The key text extracting unit 701 inputs the key text extracted by the key text extracting unit 700, associates the key text with the corresponding image data or text data, and stores the key text in the key text DB 702.

Further, the search processing unit 703 receives an input of a search request, and outputs a search result.
And a search engine 705 for inputting a search request from the input / output unit 704 and searching for a corresponding key text from the key text DB 702. Specifically, the input / output unit 7
04 inputs a search request and outputs the search request to the search engine 705. The search engine 705 includes an input / output unit 70
4, a search request is input, a corresponding key text is searched from the key text DB 702, and the corresponding key text is output to the input / output unit 704. The input / output unit 704 outputs image data or text data associated with the input key text as a search result.

The search processing unit 703 corresponds to the search server 108 in FIG. 1, and performs a search process based on a search request from the client 107. When the document image recognition device according to the second embodiment is configured in a stand-alone mode, a search request is directly input to perform a search process.

As described above, according to the document image recognition apparatus of the fourth embodiment, a document search is performed by using a character string in a document image as a key text. , And a high-speed search process can be realized.

In the fourth embodiment described above,
Although the document search is used using the key text, a full-text search may be performed using all the results of character recognition by the OCR unit 207 instead of the key text. FIG. 11 is a conceptual configuration diagram illustrating a modified example of the document image recognition device of the fourth embodiment.

As shown in FIG. 11, the text registration unit 80
0 and a full-text search text DB 801 are provided, and the text registration unit 800 receives the character recognition result from the OCR unit 207, that is, the full text of the text in the document image, associates the text with the corresponding image data or text data, and associates it with the corresponding text data. Register in DB801. As a result, image data and text data can be searched using the text registered in the full-text search text DB 801. In this case, the text is stored in the full-text search text DB8.
01, it is not necessary to open and close each file every time a search is performed, so that the speed of the search process can be increased.

The document image recognition apparatuses according to the first to fourth embodiments can be arbitrarily combined to form one document image recognition apparatus.

Further, a program for causing a computer to function as the document image recognition device according to the above-described first to fourth embodiments is prepared, and these are stored in a computer-readable recording medium such as a hard disk, a floppy disk, a CD-ROM, an MO, and a DVD. And the program can be distributed via a recording medium. Then, the above-described document image recognition device can be realized by reading and executing the program recorded on the recording medium by the computer.

[0080]

As described above, according to the document image recognition apparatus of the present invention, the input means for inputting the document image generated by optically reading the paper document and the input means are provided. Area extracting means for recognizing and extracting a character area including a character string and / or an image area including an image such as a figure, a table, and a photograph from a document image input through the apparatus, and a character string of the character area extracted by the area recognizing means A character recognizing unit that performs a character recognizing process, a layout information extracting unit that analyzes a layout of a document image based on an extraction result of the area extracting unit, and extracts layout information, and a character recognizing result and a layout information by the character recognizing unit. A first document generation unit for generating a first document using a page description language based on a layout information extraction result by the extraction unit, and a character recognition unit A second document generation unit for generating a second document using a structured description language based on the character recognition result and a layout information extraction result by the layout information extraction unit; and a first and second document generation unit. Storage means for storing the first and second documents respectively, so that not only a simple character recognition process of extracting a character code from the document image but also various information of the document image is extracted. By using it, documents that prioritize reproduction of paper documents and documents that emphasize contents of paper documents, etc.
Documents that can be used on various types of computers according to the purpose of use can be generated.

According to the document image recognition device of the present invention (claim 2), in the document image recognition device according to claim 1, the first document is a document expressed in a PostScript format or a PDF format. Therefore, a document suitable for distribution and browsing, which can convey the intention of the document creator as accurately as possible, can be obtained.

According to the document image recognition device of the present invention (claim 3), in the document image recognition device according to claim 1, the second document is SGML, HTML, or XML.
Since the document is represented by a document, it is possible to generate a document having a form unique to a computer in which the contents of the paper document are prioritized, regardless of the layout of the paper document.

Further, according to the computer-readable recording medium of the present invention (claim 4), the computer-readable recording medium for causing a computer to function as each means of the document image recognition apparatus according to any one of claims 1 to 3 is provided. Because the program is recorded, the recorded program is executed on a computer to extract and use various information of the document image, thereby giving priority to the reproduction of paper documents and documents that emphasize the contents of paper documents. For example, documents that can be used on various types of computers according to the purpose of use can be generated.

[Brief description of the drawings]

FIG. 1 is a configuration diagram illustrating a system configuration of a document image recognition device according to a first embodiment;

FIG. 2 is a conceptual configuration diagram of the document image recognition device shown in FIG.

FIG. 3 is a flowchart illustrating an operation procedure of the document image recognition device according to the first embodiment;

FIG. 4 is an explanatory diagram illustrating an example of a state where a document image input via a document image input unit is displayed on a screen in the document image recognition device according to the first embodiment;

FIG. 5 is an explanatory diagram showing an example of a state in which a generated PostScript document is displayed on a screen in the document image recognition device of the first embodiment.

FIG. 6 is an explanatory diagram showing an example of a state where a generated HTML document is displayed on a screen in the document image recognition device of the first embodiment.

FIG. 7 is a conceptual configuration diagram of an image recognition device according to a second embodiment.

FIG. 8 is a conceptual configuration diagram showing a modified example of the document image recognition device of the second embodiment.

FIG. 9 is a conceptual configuration diagram of a document image recognition device according to a third embodiment.

FIG. 10 is a conceptual configuration diagram of a document image recognition device according to a fourth embodiment.

FIG. 11 is a conceptual configuration diagram showing a modified example of the document image recognition device of the fourth embodiment.

[Explanation of symbols]

REFERENCE SIGNS LIST 100 color scanner 101 monochrome scanner 102 network scanner 103 facsimile machine 104 fax modem 105 (105a, 105b) document image processing server 106 digital multifunction peripheral 107 client 108 search server 108 109 network 200 document image input unit 201 preprocessing unit 202 noise removal unit 203 Skew correction unit 204 Information extraction unit 205 Area division / information extraction unit 206 Table processing unit 207 OCR unit 208 Font identification unit 209 Layout / logic analysis unit 210 Document generation unit 211 PostScript document generation unit 212 HTML document generation unit 213 Database unit 214 PostScript document DB 215 HTML document DB 700 Key text extraction unit 701 Key text registration unit 02 key text DB 703 the search processing unit 704 input-output unit 705 search engine 800 text registration unit 801 full-text search for text DB

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code FI H04N 1/40 H04N 1/40 F

Claims (4)

[Claims] 1. An input means for inputting a document image generated by optically reading a paper document, a character area including a character string and / or a figure, table, or photograph from the document image input via the input means. Area extracting means for recognizing and extracting an image area including an image such as an image, character recognizing means for performing character recognition processing on a character string of the character area extracted by the area recognizing means, and based on the extraction result of the area extracting means. A layout information extracting means for analyzing a layout of the document image and extracting layout information; and a page description language based on a character recognition result by the character recognition means and a layout information extraction result by the layout information extracting means. First to generate the first document
Generating a second document using a structured description language based on the character recognition result by the character recognition unit and the layout information extraction result by the layout information extraction unit.
A document image recognizing device comprising: a document generating unit; and a storage unit that stores the first and second documents generated by the first and second document generating units, respectively. 2. The method according to claim 1, wherein the first document is a PostScript.
2. The document image recognition device according to claim 1, wherein the document is a document represented by a t format or a PDF format. 3. The method according to claim 1, wherein the second document is SGML, HTML,
2. The document image recognition device according to claim 1, wherein the document is a document represented by XML. 4. A computer-readable recording medium on which a program for causing a computer to function as each unit of the document image recognition apparatus according to claim 1 is recorded.