JP2001094711A - Document image processing apparatus and document image processing method - Google Patents

Abstract

(57) [Summary] In a document image processing apparatus, even if it is difficult to accurately extract character information such as drawings and line drawings, a character string included in the drawings and line drawings and a document Linking with other parts with high accuracy. When a first range and a second range to be linked are specified on a document image (ST101)
104), and character recognition is performed on this first range (ST10).
2), the recognized word is registered together with the position information (ST10
3). Further, character recognition is performed for the second range (S
T105), the recognized character string is collated with the registered words in the first range (ST106). Then, when the recognized character string from the second range and the word in the word registered in the first range are collated and matched, the position information of the collated and matched second range character string and the Link information that associates the registered word with the position information is generated (ST107).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a document image processing apparatus and a document image processing method for associating a document input with an image, such as a filing system or a document database, between documents or within a document.

[0002]

2. Description of the Related Art Heretofore, a data position of a document image input by a scanner or the like (coordinate point in processing).
An association (link) for indicating a reference relationship from to another data position (the same coordinate point) is often performed manually. In order to perform this automatically, a method has been attempted in which a character is selected in a data area to be linked and then a keyword is selected and associated.

In this case, if the document has a good drawing quality,
The effect of character recognition is demonstrated, and keywords can be extracted with high accuracy.However, character recognition is used for documents that contain noise, etc., and for which it is difficult to accurately extract character information such as drawings and line drawings. May become inaccurate and linking between related data may not be performed correctly.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and has been developed for a document in which it is difficult to accurately extract character information such as drawings and line drawings. It is an object of the present invention to provide a document image processing apparatus and a document image processing method capable of accurately linking a character string included in a document or a line drawing with another part of the document.

[0005]

That is, a first document image processing apparatus according to a first aspect of the present invention is a document image processing apparatus for linking a plurality of data positions on a document captured as image data. An apparatus, comprising: a range designating unit that designates a first range and a second range to be linked on the document image; and a first range on the document image designated by the range designating unit. Word registration means for registering a word recognized from the first range together with its position information, and a second range on the document image designated by the range designation means. A character string that is recognized from the second range and collated with a word in the first range registered by the word registration unit. Means, if the character string recognized from the second range by the word matching means matches a word in the first range registered by the word registration means, Link information generating means for generating link information in which the position information of the character string in the range of 2 and the position information of the registered word are associated with each other.

In the first document image processing apparatus having such a configuration, when a first range and a second range to be linked are specified on a document image, the document image specified by the range is designated. Character recognition is performed on the first range above, words recognized from the first range are registered together with their position information, and a second range on the document image designated as the range is registered. Is subjected to character recognition, and the character string recognized from the second range is collated with the word in the first range registered as the word. Then, when the character string recognized from the second range by the word matching is matched with the word in the first range registered by the word registration, the second range matched by the matching is matched. Since link information in which the position information of the character string is associated with the position information of the registered word is generated, even when character recognition accuracy is low or when character information such as a table or a line drawing cannot be accurately extracted, linking is performed. Can be performed accurately.

A second document image processing apparatus according to a second aspect of the present invention is a document image processing apparatus for linking a plurality of data positions on a document captured as image data, The first and second areas to be linked on the document image
Range designating means for designating the range, character evaluation means for evaluating the quality of character recognition of two ranges on the document image designated by the range designating means, and the quality evaluating means evaluating the quality as high. Word recognition means for performing character recognition on the first or second range and registering a word recognized from one of the ranges together with its position information; Word matching means for performing character recognition on the other of the first and second ranges and matching a character string recognized from the other range with a word in one of the ranges registered by the word registration means; If the character string recognized from the other range by the word matching unit matches the word in one range registered by the word registration unit, the word is matched. Characterized by comprising a link information generating means for generating link information which associates location information of the character string matching the matched other range as the position information of the registered words.

In the second document image processing apparatus having such a configuration, when the first range and the second range to be linked are specified on the document image, the document image specified by the range is designated. The quality for character recognition in the above two ranges is evaluated, and character recognition is performed on one of the first and second ranges evaluated as having high quality by the quality evaluation, and recognition is performed from one of the ranges. Is registered together with the position information, and character recognition is performed on the other first or second range evaluated as having low quality by the quality evaluation, and the character recognized from the other range is recognized. A column is matched with the words in one of the registered words. Then, when the character string recognized from the other range by the word matching and the word in one range registered by the word registration are matched, the character string in the other matched range is matched. Since link information in which the position information is associated with the position information of the registered word is generated, even when character recognition accuracy is low, or when character information such as tables and line drawings cannot be accurately extracted, linking is more accurate. Can be done.

A third document image processing apparatus according to a fifth aspect of the present invention is a document image processing device for capturing a document including a table portion and a drawing portion as image data, and a document image captured by the document image capturing device. Range extracting means for extracting a range of a table portion and a range of a drawing portion on a document image; and performing character recognition of a character string corresponding to an item name in the range of the table portion of the document image extracted by the range extracting device. Word registration means for registering a word whose character has been recognized as an item name in the range of the table portion, and character recognition of a character string in a range of a drawing portion of the document image extracted by the range extracting means. The character string recognized from the range of the part is stored in the range of the table part registered by the word registration means. Word matching means for matching with a word of an item name to be matched, and a character string recognized from the range of the drawing part by the word matching means and a word of an item name in the range of the table part registered by the word registration means. A link information generating unit that generates link information that associates the position information of the character string in the range of the drawing portion that has been matched and matched with the registered word that is the item name of the range of the table portion. It is characterized by having.

In the third document image processing apparatus having such a configuration, when a document including a table portion and a drawing portion is captured as image data, a range of the table portion and a range of the drawing portion are extracted on the document image. Then, character recognition is performed on a character string corresponding to the item name in the range of the table portion of the extracted document image, and the recognized word is registered as the item name in the range of the table portion. Further, character recognition of a character string in the range of the drawing portion of the document image from which the range is extracted is performed, and the character string recognized from the range of the drawing portion is used for the item name in the range of the table portion in which the word is registered. Matches a word. If the character string recognized from the range of the drawing portion by the word matching matches the word of the item name in the range of the table portion registered by the word registration, the matching drawing portion Link information is generated by associating the position information of the character string in the range with the registered word which is the item name of the table part, so that the item name of the table part and the character string in the drawing are efficiently associated with each other. This means that data can be created by adding position information in the drawing to the items in the table portion.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) In the first embodiment,
When two ranges on the document are specified as the target range of linking, after registering the character strings in one specified range in the word dictionary, the word dictionary is extracted when extracting the character strings in the other specified range. A description will be given of a document image processing function for associating data positions where both character strings exist with reference to collation.

FIG. 1 is a block diagram showing a configuration of an electronic circuit of a document image filing apparatus according to an embodiment of the present invention.

This document image filing apparatus includes a control device (CPU) 21 which is a computer.

A control device (CPU) 21 responds to image data input from an image input device 22 or data input or specified by a data input / instruction device 23 to output a R signal.
The system program stored in the OM 24 is started in advance, or the external recording medium 2 such as a floppy disk
The program data for document image processing stored in the storage medium 5 is read by a recording medium reading unit 26 such as a magnetic disk device and activated to control the operation of each circuit unit.

The control device (CPU) 21 includes the image input device 22, the data input / instruction device 23, the ROM 2
4, a RAM 27 and a display device 28 are connected.

The image input device 22 is sent from another computer terminal device via an image scanner or a communication network which optically reads a document or a document on which a drawing or the like is drawn and converts it into image data. The document image data input by the image input device 22 is stored in an image memory 27a in the RAM 27. The communication interface is configured as a communication interface for receiving and inputting image data.

The data input / instruction device 23 includes a keyboard for inputting characters, symbols, numbers and the like, and a mouse for instructing a data position, specifying a range, and performing a moving operation. When an arbitrary area on the image data stored in the image memory 27a is designated by the instruction device 23, the image data in the designated area is read and stored in the read image memory 27b in the RAM 27. On the other hand, as main control programs for realizing the document image processing function of the document image filing apparatus, the program data stored in the ROM 24 in advance include a character recognition program 24a, a word dictionary creation program 24b, a word dictionary matching program 24
c, a link information generation program 24d and a quality evaluation program 24e (used in the second embodiment) are prepared.

As a main data memory for realizing the document image processing function of the document image filing apparatus, a memory area secured in the RAM 27 includes, in addition to the image memory 27a and the read image memory 27b, A word dictionary memory 27c and a link information memory 27d are prepared.

The various control programs stored in the ROM 24 and the various data memories secured in the RAM 27 will be further described with reference to the functional blocks shown in FIG.

FIG. 2 is a block diagram showing the configuration of the document image processing function of the first embodiment in the document image filing apparatus.

In the functional block of the document image processing function, the corresponding components of the document image filing apparatus in FIG. 1 are indicated by reference numerals in parentheses.

This document image processing function includes an image input unit 1 (22) for reading a paper document as image data, an image storage unit 2 (27a) for filing the read image data, and a filing function. An image reading unit 3 (27b) for reading the read image data as needed, an image display unit 4 (28) for displaying the read image data on a screen such as a display monitor, and the like. An area specifying unit 5 (23) for specifying all or part of the data as an area, a character recognizing unit 6 (24a) for extracting a character string included in the specified area and performing character recognition; A word dictionary creating unit 7 (24b) for extracting words from the result of character recognition and registering them in a word dictionary, and a word dictionary for storing and registering the created word dictionary A storage unit 8 (27c), a word dictionary matching unit 9 (24c) for matching the result of character recognition in another specified area with the registered word dictionary, and using the matching result with the word dictionary A link information generating unit 10 (24d) for generating link information based on coordinates between two data positions (points) on the image data; and storing the generated link information in association with the image data. The link information storage unit 11 (27d).

Next, the document image processing function of the first embodiment in the document image filing apparatus configured as described above will be described.

FIG. 3 is a flowchart showing a link information generation process performed by the document image processing function of the first embodiment of the document image filing apparatus.

First, processing is performed on a document area to be a link destination.

One document image data read by the image input unit 1 (22) and stored in the image storage unit 2 (27a) is displayed on the image display unit 4 (28).
An arbitrary part or all of the image is designated by an area designating unit 5
By (23), it is designated as a link destination target area (step ST101).

As a specific area designating means, the image reading section 3 (27b) searches for and reads out a corresponding document, displays it on the image display section 4 (28), and displays it on the image designating section 5 (23). An area is specified by a drag operation on the screen. In addition, a layout analysis may be performed on the display image data retrieved and read to extract a text area, and the area may be designated as a target area.

Next, the character recognition unit 6 (24)
A character area is extracted by the character recognition processing in a), and character recognition is performed (step ST102). The character recognition result on the image data designation area obtained here is divided into individual words by a word dictionary generation process by the word dictionary generation unit 7 (24b), and this is divided into individual words.
7c) (step ST103).

Next, processing is performed on the document area serving as a link source.

In the same manner as described above, 1 is read by the image input unit 1 (22) and stored in the image storage unit 2 (27a).
The document image data of one sheet is displayed on the image display unit 4 (28), and an arbitrary part or all of the image is designated by the region designation unit 5 (23) as the target region of the link source (step ST104). The specific method is
This is the same as the link destination area specifying operation in step ST101.

Then, a character area is extracted from the specified link source image area by the character recognition processing by the character recognition unit 6 (24a) in the same manner as in step ST102, and character recognition is performed (step ST105). Then, the result of the character recognition is stored in the word dictionary storage unit 8.
The word dictionary stored in (27c) and registered is extracted, and word matching with the registered word is performed by word dictionary matching processing by the word dictionary matching unit 9 (24c) (step ST106).

Here, the result of character recognition of the link source area in step ST105 and the word dictionary storage unit 8 (27
If the word in the link destination area stored and registered in c) can be matched with the same word, the word dictionary is stored from the image data position (coordinates) of the link source from which the result of the character recognition is obtained. The link information to the image data position at the link destination of the collation word stored in the unit 8 (27c) is generated by the link information generating process by the link information generating unit 10 (24d) (step ST107). The link information, for example, associates coordinates indicating an image data position in the link source area that matches and matches the registered word with coordinates indicating the image data position in the link destination area of the registered word that matches and matches. Generated as a data link table,
It is stored in the link information storage unit 11 (27d).

Next, a process of generating link information will be described using a specific example of a document image.

FIG. 4 is a diagram showing an example of a document image including a table portion and a text portion.

First, the specification of the area of the document image data at the link destination (step ST101) will be described.

FIG. 5 is a diagram showing an area designation display state of a link destination for the document image.

A document image Ga as shown in FIG. 4 is read out by the image reading section 3 (27b) and is displayed on the image display section 4 (28) as shown in FIG. As described above, the image area Er of the part to be the link destination is specified by the mouse drag operation by the area specifying unit 5 (23).

As described above, the area designating means may be the whole document of one page, or may be designated over a plurality of pages. In this case, the user explicitly specifies the position of the image area Er. However, the area is automatically specified by, for example, automatically extracting a part in which the item name of the table part is written by a table understanding technique, and specifying the area. You may.

The link destination image area E thus extracted
r, each character string is extracted by the character recognition processing by the character recognition unit 6 (24a), and when the character recognition is performed, the “front part”, “back part”, “tip part”, and “connection part” are obtained. A word is extracted (step ST102), and the word is extracted by the word dictionary creating unit 7 (24b) into the word dictionary storage unit 8.
(27c) (step ST103).

FIG. 6 is a diagram showing a registration state of a plurality of words extracted by character recognition with respect to the link destination area of the document image.

At this time, as shown in FIG. 6, the word dictionary storage unit 8 (27c) identifies a document in association with each word extracted by character recognition from the image area Er of the link destination. Document number and document name,
The start coordinates and end coordinates at which the word is located are recorded.

Next, the specification of the area of the document image data at the link source (step ST104) will be described.

FIG. 7 is a diagram showing an area designation display state of a link source for the document image.

In the same manner as described above, a document image Ga as shown in FIG. 4 is read out by the image reading section 3 (27b), and as shown in FIG. As indicated by the frame, the image area Es of the link source is specified by the mouse drag operation by the area specifying unit 5 (23).

At this time, an arbitrary part of another document image may be designated as the link source region Es, or an appropriate range of the document over several pages may be designated as the region and the link source region Es. Also, besides the user explicitly specifying the position, it is automatically extracted by specifying "table part", "graphic part", "text part", etc.
It may be assigned as a link target area.

Next, the character recognizing unit 6 (24a) performs a character recognition process on the specified link source image area Es (step ST105). Then, knowledge processing of character recognition is performed (step ST106).

This knowledge processing of character recognition is used in post-processing of character recognition.
In this method, a word obtained from the set of recognition candidate characters in s is compared with a registered word in the link destination area Er registered in the word dictionary storage unit 8 (27c). Here, when the recognized character string in the link source area Es and the registered word in the link destination area Er are collated,
The document number and its position information associated with the collation word registered in the dictionary in the word dictionary storage unit 8 (27c) in FIG. 6 are extracted and, for example, as shown in FIG. 8, the cursor is designated as the current link source. From the data position of the character string "front part" r1 of the text part, the word "front part" of the table part which is a link destination according to the position information
Link information up to the data position of r2 is generated (step ST107), and a link is displayed on the image display unit 4 (28).

FIG. 8 is a diagram showing a state of generating link information on a document image according to the document image processing function of the first embodiment.

Therefore, according to the document image processing function of the first embodiment having the above-described configuration, the link destination area Er and the two areas Er and Es specified on the document image Ga are linked. Is registered in the post-processing dictionary, and in the character recognition of the link source area Es, the registered dictionary is used for knowledge processing,
In the case of collation, the link information r1-r2 is generated according to the data position of the collated link destination registered word, so that the character recognition accuracy is low, or character information such as a table or line drawing is used. Even if you can not extract accurately
Linking can be performed accurately.

(Second Embodiment) In the second embodiment,
When two ranges on a document are designated as target ranges for linking, a range to be registered in a word dictionary and a range to be subjected to knowledge processing are determined according to character quality in the two designated ranges, and then one of the ranges is determined. A description will be given of a document image processing function that associates a registered word in a range with an extracted character string in the other range and associates data positions with each other.

FIG. 9 is a block diagram showing the configuration of the document image processing function of the second embodiment in the document image filing apparatus.

In the functional block of the document image processing function, the corresponding components of the document image filing apparatus in FIG. 1 are indicated by reference numerals in parentheses.

The document image processing function includes an image input unit 1 (22) for reading a paper document as image data, an image storage unit 2 (27a) for filing the read image data, and a filing function. An image reading unit 3 (27b) for reading the read image data as needed, an image display unit 4 (28) for displaying the read image data on a screen such as a display monitor, and the like. An area specifying unit 5 (23) for specifying all or part of the data as an area, a character recognizing unit 6 (24a) for extracting a character string included in the specified area and performing character recognition; A word dictionary creating unit 7 (24b) for extracting words from the result of character recognition and registering them in a word dictionary, and a word dictionary for storing and registering the created word dictionary A storage unit 8 (27c), a word dictionary matching unit 9 (24c) for matching the result of character recognition in another specified area with the registered word dictionary, and using the matching result with the word dictionary A link information generating unit 10 (24d) for generating link information based on coordinates between two data positions (points) on the image data; and storing the generated link information in association with the image data. Link information storage unit 11 (27d) and area designation unit 5
The quality evaluation unit 12 (24) for evaluating the quality of the character image in each area specified by (23) and determining the range to be registered in the word dictionary and the range in which knowledge processing is to be performed.
e).

Next, the document image processing function of the second embodiment in the document image filing apparatus having the above configuration will be described.

FIG. 10 is a flowchart showing a link information generation process performed by the document image processing function of the second embodiment of the document image filing apparatus.

First, the document image data read by the image input unit 1 (22) and stored in the image storage unit 2 (27a) is displayed on the image display unit 4 (28).
Two areas to be associated are specified by the area specifying unit 5 (23). That is, designation of area 1 (step ST
201) and designation of the area 2 (step ST202). The area specified here may be a single document or a document spanning a plurality of documents.

Next, the character quality of each of the designated areas is evaluated by quality evaluation processing in the quality evaluation unit 12 (24e). That is, evaluation of the character quality for area 1 (step ST203) and evaluation of the character quality for area 2 (step ST204) are performed. As a means for evaluating the character quality, there is a method of extracting a region feature of each designated region and determining the character quality according to the result. Specifically, a connected component is extracted from the image data in the designated area, a connected component having a size like a character is extracted from the size of the extracted connected component as a character candidate area, and a character candidate in the area is extracted. The regions are classified into several categories based on the distribution of the regions, and character quality is determined according to the categories.

Here, the category in the designated area is a type of a document element such as a text area, a table area, a drawing area, and a photograph area. For example, character quality is set to be higher in the order of a text area, a table area, a drawing area, and a photograph area.

As another means of quality evaluation, character recognition is actually performed on the character candidate area in the designated area, and the certainty at the time of the recognition is measured. Set it. As the certainty factor of the character recognition, the similarity at the time of pattern matching with the character recognition dictionary, the difference between the first and second similarities of the recognition candidates, and the like, or a combination thereof are used.

Based on the result of the quality evaluation of each area performed in this way, these are compared (step ST205). As a result of this comparison, character recognition is performed by the character recognition process in the character recognition unit 6 (24a) on one of the areas determined to be of high quality (step ST206).
In addition, as a result of the quality comparison, when the character quality of each area is evaluated to be substantially the same, for example, selection processing such as treating a smaller area size as a high quality area is performed. And
The recognition result obtained by the character recognition processing performed on the area with the higher character quality is divided into words by the word dictionary creation processing in the word dictionary creation unit 7 (24b), and this is divided into words. It is stored in the storage unit 8 (27c) and registered (step ST207).

Next, a character area is extracted from the other designated area having a low evaluation of the character quality by the character recognition processing in the character recognition unit 6 (24a), and character recognition is performed (step ST208). Then, with respect to the character recognition result in the other area, the word dictionary extracted in the one area stored and registered in the word dictionary storage section 8 (27c) is extracted, and the word dictionary in the word dictionary matching section 9 is extracted. Word matching is performed by dictionary matching processing (step ST20).
9).

Here, the result of character recognition of the other area in step ST208 and the word dictionary storage unit 8
If the registered word in one area stored and registered in (27c) can be collated with the same word, the word dictionary storage unit 8 (2) is used from the data position of the collated character string in the other area.
The link information to the data position of the collation word in one area stored and registered in 7c) is transmitted to the link information generation unit 10 (2).
By the link information generation processing in 4d), the data is generated by associating the coordinates indicating the respective data positions (step ST210). The link information is stored in the link information storage unit 11 (27d), and the link display between one and the other image areas read by the image read unit 3 (27b) is displayed on the image display unit 4 (28). ).

Therefore, according to the document image processing function of the second embodiment having the above-described configuration, one of the areas having higher character quality is compared with the two areas to be linked specified on the document image. First, the word recognition is performed first, and the word dictionary is created with high accuracy. Thereafter, in the character recognition processing of the other area with the lower character quality, the words of the one registered dictionary are used for the knowledge processing, When a column is collated, a link is made immediately to the data position of the registered word in one of the collated regions, so that character recognition accuracy is low or character information such as tables and line drawings can be accurately identified. Even if it cannot be extracted, linking can be performed accurately.

(Third Embodiment) In the third embodiment,
For the document image including the table part and the drawing part, extract the item name from the table part and the character string indicating the position in the drawing from the drawing part and link them. A document image processing function of a tabular document to which a position attribute is given will be described.

FIG. 11 is a flowchart showing a data reading link process performed by the document image processing function of the third embodiment of the document image filing apparatus.

FIG. 12 is a diagram showing an example of a document image Gb composed of a table portion and a drawing portion.

FIG. 13 is a diagram showing a format registration state for the document image Gb consisting of the table part and the drawing part.

FIG. 14 is a diagram showing a state of character recognition and collation on a tabular document image according to the document image processing function of the third embodiment.

FIG. 15 is a diagram showing a state of generating link information on a tabular document image according to the document image processing function of the third embodiment.

First, the document image data read by the image input unit 1 (22) and stored in the image storage unit 2 (27a) is displayed on the image display unit 4 (28).
The document reading unit 3 (27b) reads a document image Gb of a tabular document including a table part and a drawing part as a linking target image as shown in FIG. 12 (step ST301).

In this case, it is necessary to register the format of the target tabular document. If not registered, format registration is performed (step ST3).
02 → ST302 ′). Here, the format to be registered includes ruled line information for forming a table, information on characters to be entered in a table portion, and information on a drawing portion.

For example, when the document image Gb of the tabular document including the drawing portion as shown in FIG. 12 is read by the image reading section 3 (27b), as shown in FIG. Position F1 of the character written in the part
Information on the (grid portion) and the position F2 (hatched portion) of the drawing portion is registered as format information.

Next, ruled line information is extracted by image processing from the read document image data of the tabular document, and a format compatible with the registered format is extracted by using the extracted ruled line information. It is identified (step ST303). Then, the ruled line information registered in the identified format is called, and the table portion is aligned (step ST304). Here, the character entry portion is cut out from the position information F1 of the character to be entered in the table portion, and each entry character in the table portion is recognized by the character recognition process by the character recognition section 6 (24a) (step ST305). . The recognition result of the entered characters is created as a word dictionary by associating the coordinates of the data position of each word by the word dictionary creation processing by the word dictionary creation unit 7 (24b), and the word dictionary storage unit 8 (27)
c) (step ST306).

On the other hand, the drawing part is cut out from the document image Gb of the tabular document based on the position information of the drawing part registered in the format identified in step ST303 (step ST307), and a character string is cut out from the cut out image data. (Step ST30)
8). In the extraction of this character string, an image area is cut out using a set of connected components that match a predetermined character size or a nearby area as a character string candidate.

For example, in the drawing portion of the document image Gb of the tabular document as shown in FIG.
As shown in FIG. 4, portions r2a to r2 actually indicating a character string
In addition to d, some extra parts r2e may be extracted as character string candidates under the influence of blurring or noise components of the image.

Next, the character string image cut out from the drawing portion is subjected to character recognition by the character recognition processing in the character recognition section 5 (24a), and at this time, the word dictionary storage section 8 (step ST306). Post-processing is performed using the word dictionary extracted from the table part registered in 27c), and word matching is performed (step ST30).
9). At this time, the character string portion r2e extraly extracted from the drawing portion cannot be matched with the words registered in the word dictionary and is discarded.

On the other hand, for the portions where the word matching is successful, the character strings r1a to r1d corresponding to the item names in the table portion are associated with the character strings r2a to r2d in the drawing, and each item in the table portion is added to each item in the drawing. Is added (step ST310).

As a result, for example, the association as shown by the arrow in FIG. 15 is performed. The link information generation unit 10 (24d) generates link information obtained by adding the corresponding data position information of the drawing portion to the registered word in the table portion obtained as a result, and stores the link information in the link information storage unit 11 (27d) (step ST311). ).

For example, as shown in FIG. 12, when the item name of the table data is a word indicating a location such as “front” or “back”, the item name itself represents the location. Therefore, it is not necessary to add position information to each item data, but if the item names are “P1” and “P2”
In this case, the position information is indispensable information for linking after data conversion by adding the corresponding position information in the drawing to each item when it is designated by a symbol or number. .

Therefore, according to the document image processing function of the third embodiment having the above-described configuration, in the document image of the tabular document including the drawing portion and the table portion, the character string can be extracted with relatively high accuracy. By registering the recognition result in a word dictionary and using the word dictionary to recognize and match the character strings in the drawing, it is possible to efficiently associate the item names in the table with the character strings in the drawing The position information in the drawing can be added to the data that can be read from the table portion to convert the data.

(Fourth Embodiment) In this fourth embodiment,
A description will be given of a document image processing function that limits the reference direction of linking using the temporal order relation of each document when linking between different documents.

FIG. 16 is a flowchart showing the link information generation processing performed by the document image processing function of the fourth embodiment of the document image filing apparatus.

First, two document images to be linked are specified (step ST401). The specifying means here may specify one document image or a plurality of document images by the area specifying unit 5 (23). Alternatively, an area may be specified such as only a character portion in a document image.

Next, character recognition is performed on the two designated document images by character recognition processing by the character recognition unit 6 (24a) (step ST402).
Here, by the word dictionary creation processing by the word dictionary creation unit 7 (24b), the character recognition result for one document image is divided into words using a method such as morphological analysis, and the like.
The divided words are stored and registered in the word dictionary storage unit 8 (27c) (step ST403).

Next, the character recognition result of the other document image by the character recognizing unit 6 (24a) is converted into the word extracted from the one document image already registered in the word dictionary storage unit 8 (27c) in step ST403. Using the dictionary, word matching is performed by the word dictionary matching unit 9 (24c) (step ST404), and link information that associates the data positions of the matched words with each other is generated by the link information generating unit 10 (24d) ( Step ST4
05).

Then, temporal information such as the creation date is extracted from both linked document images (step ST406). At this time, it is desirable to extract from both document images by the same method. As a method of extracting the time information, a character area is extracted from the upper or lower part of the document image and character recognition is performed, so that the recognized character string is a sequence of numbers, or “month,” “day,” “ It is determined whether the information is temporal information by determining whether the character string includes a keyword such as "". Alternatively, time information may be obtained by specifying several data positions in which the date is predicted to be written in advance, and extracting a character string from the specific data position. If it cannot be obtained by extraction from an image, for example, the time of reading the image may be used as it is, or a user may be input manually.

The temporal information of each document image thus obtained is also added to the link information generated in step ST405, and the display direction of each linked document image is restricted (step ST407). .

FIG. 17 is a diagram showing a state of generating link information between two document images having time information associated with the document image processing function of the fourth embodiment.

For example, as shown in FIG.
Document image Gb1 created on “20th month” and “9
Document image Gb2 created on "August 20, 2009"
On the other hand, regarding the generated link information, the reference direction accompanying the link display is limited to the direction from the document Gb1 to the document Gb2 (r1 → r2).
Alternatively, in accordance with the time information added to the link information between the document images, the temporal information such as “previous document” and “back document” is displayed together with the link display of each document image, and both documents are displayed. Links may be displayed between the two.

If the link position information for another document image already exists from the link position information of the document image generated here, the temporal information of the another document image is used. A so-called sorting process may be performed so that the links are arranged in a temporal order, and a process of rewriting the link information may be added (step ST408).

FIG. 18 is a diagram showing a sort state of link information among a plurality of document images having time information associated with the document image processing function of the fourth embodiment.

For example, as shown in FIG.
When the document image Gb3 dated "August 1, 9" is read and linked to the same point of the document image Gb2, the link information is changed (r1 → r2 → r3) so that the documents can be referred to in chronological order. .

Therefore, according to the document image processing function of the fourth embodiment having the above-described configuration, a plurality of designated document images are linked, and temporal information is extracted. Since the link information to which the additional information such as the reference direction restriction is given is generated, it is possible to browse the documents in chronological order or return to the same document at the previous time.

The method described in each of the above embodiments, namely, the link information generation processing in the first embodiment shown in the flowchart of FIG. 3, the link information generation processing in the second embodiment shown in the flowchart of FIG. Each method such as the data reading link processing in the third embodiment shown in the flowchart of FIG. 11 and the link information generation processing in the fourth embodiment shown in the flowchart of FIG. It can be stored in an external recording medium 25 such as a card (ROM card, RAM card, etc.), a magnetic disk (floppy disk, hard disk, etc.), an optical disk (CD-ROM, DVD, etc.), a semiconductor memory, and distributed. And
The computer reads the program stored in the external recording medium 25 by the recording medium reading unit 26,
By controlling the operation by the read program, the function of generating the link information for the document image described in each embodiment can be realized, and the same processing by the above-described method can be executed.

The program data for implementing each of the above methods can be transmitted over a network in the form of a program code, and the communication control unit of a computer terminal connected to the network transmits the program data. It is also possible to realize various document image processing functions described above.

[0097]

As described above, according to the first document image processing apparatus of the first aspect of the present invention, the first range and the second range to be linked on the document image are designated. Then, character recognition is performed on the first range on the document image designated as the range, the word recognized from the first range is registered together with its position information, and Character recognition is performed on a specified second range on the document image, and a character string recognized from the second range is compared with a word in the first range registered with the word. Then, when the character string recognized from the second range by the word matching is matched with the word in the first range registered by the word registration, the second range matched by the matching is matched. Since link information in which the position information of the character string is associated with the position information of the registered word is generated, even when character recognition accuracy is low or when character information such as a table or a line drawing cannot be accurately extracted, linking is performed. Can be performed accurately.

Further, according to the second document image processing device of the present invention, when the first range and the second range to be linked on the document image are specified, The quality for character recognition of two ranges on the document image whose range is specified is evaluated, and character recognition is performed for one of the first and second ranges evaluated as having high quality by this quality evaluation, Words recognized from one of these ranges are registered along with their location information,
In addition, character recognition is performed on the first or second other range evaluated as having a low quality by the quality evaluation, and a character string recognized from the other range is used in one of the ranges where the word is registered. Matches a word. Then, when the character string recognized from the other range by the word matching and the word in one range registered by the word registration are matched, the character string in the other matched range is matched. Since link information in which the position information is associated with the position information of the registered word is generated, even when character recognition accuracy is low, or when character information such as tables and line drawings cannot be accurately extracted, linking is more accurate. Will be able to do it.

Further, according to the third document image processing apparatus of the present invention, when a document including a table portion and a drawing portion is fetched as image data, the range of the table portion is defined on the document image. The range of the drawing portion is extracted, the character recognition of the character string corresponding to the item name in the range of the table portion of the extracted document image is performed, and the word recognized as the item name of the range of the table portion is recognized. Is registered. Further, character recognition of a character string in the range of the drawing portion of the document image from which the range is extracted is performed. Matches a word. If the character string recognized from the range of the drawing portion by the word matching matches the word of the item name in the range of the table portion registered by the word registration, the matching drawing portion Link information is generated by associating the position information of the character string in the range with the registered word which is the item name of the table part, so that the item name of the table part and the character string in the drawing are efficiently associated with each other. As a result, data can be created by adding position information in the drawing to the items in the table portion.

Therefore, according to the present invention, even for a document in which it is difficult to accurately extract character information such as a drawing or a line drawing, the character string included in the drawing or the line drawing can be compared with another part of the document. Linking can be performed with high accuracy.

Further, according to the fourth document image processing apparatus of the present invention, a plurality of documents are fetched as image data, and each document image is interposed between the plurality of document images. When link information that associates the above location information is generated,
By generating the link information, the time information of each of the plurality of linked document images is extracted, and the line information is generated based on the generated link information in accordance with the temporal order according to the time information of each of the plurality of document images. The reference direction between the plurality of document images is limited, or, at the time of reference reading between the plurality of document images performed based on the generated link information, information of the temporal order is added, In the document image processing apparatus that can perform the linking accurately, it becomes possible to browse the documents in order of time and return to the same document at the previous time.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an electronic circuit of a document image filing apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a document image processing function of the first embodiment in the document image filing apparatus.

FIG. 3 shows a first example of the document image filing apparatus.
9 is a flowchart illustrating link information generation processing performed by the document image processing function of the embodiment.

FIG. 4 is a diagram illustrating an example of a document image including a table portion and a text portion.

FIG. 5 is a view showing an area designation display state of a link destination with respect to the document image.

FIG. 6 is a view showing a registration state of a plurality of words extracted by character recognition with respect to a link destination area of the document image.

FIG. 7 is a diagram showing a state of specifying a link source area with respect to the document image.

FIG. 8 is a diagram showing a state of generating link information on a document image according to the document image processing function of the first embodiment.

FIG. 9 is a block diagram showing a configuration of a document image processing function of a second embodiment in the document image filing apparatus.

FIG. 10 is a flowchart showing link information generation processing performed by the document image processing function of the second embodiment of the document image filing apparatus.

FIG. 11 is a flowchart showing a data reading link process performed by the document image processing function of the third embodiment of the document image filing apparatus.

FIG. 12 is a diagram illustrating an example of a document image Gb including a table portion and a drawing portion.

FIG. 13 is a view showing a format registration state for a document image Gb including the table part and the drawing part.

FIG. 14 is a diagram showing a state of character recognition and collation on a tabular document image according to the document image processing function of the third embodiment.

FIG. 15 is a diagram showing a state of generating link information on a tabular document image according to the document image processing function of the third embodiment.

FIG. 16 is a flowchart showing link information generation processing performed by the document image processing function of the fourth embodiment of the document image filing apparatus.

FIG. 17 is a diagram showing a state of generating link information between two document images having time information associated with the document image processing function of the fourth embodiment.

FIG. 18 is a diagram showing a sort state of link information among a plurality of document images having time information associated with the document image processing function of the fourth embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Image input part 2 ... Image storage part 3 ... Image reading part 4 ... Image display part 5 ... Area designation part 6 ... Character recognition part 7 ... Word dictionary creation part 8 ... Word dictionary storage part 9 ... Word dictionary matching part 10 ... Link information generation unit 11 Link information storage unit 12 Quality evaluation unit 21 Control device (CPU) 22 Image input device 23 Data input / instruction device 24 ROM 24a Character recognition program 24b Word dictionary creation program 24c ... word dictionary matching program 24d ... link information generation program 24e ... quality evaluation program 25 ... external recording medium 26 ... recording medium reading unit 27 ... RAM 27a ... image memory 27b ... readout image memory 27c ... word dictionary memory 27d ... link information memory 28 ... Display device

Continued on front page F term (reference) 5B050 BA10 BA20 CA07 EA04 EA06 FA02 FA09 5B064 BA01 BA05 CA08 DA27 EA19 5C062 AA06 AA13 AB17 AB20 AB24 AB38 AB41 AB43 AB44 AC02 AC05 AC07 AC22 AC58 AE15

Claims (9)

[Claims] 1. A document image processing apparatus for linking a plurality of data positions on a document captured as image data, wherein the first and second links are linked on the document image.
Range specifying means for specifying a range and a second range, and character recognition is performed on a first range on the document image specified by the range specifying means, and a word recognized from the first range And word registration means for registering the position information together with the position information, and character recognition is performed on a second range on the document image specified by the range specification means, and a character string recognized from the second range is Word matching means for matching words in the first range registered by the word registration means; and a character string recognized from the second range by the word matching means and a first character string registered by the word registration means. If the words in the range are collated and matched, a link information that associates the position information of the character string in the collated and matched second range with the position information of the registered word is generated. Document image processing apparatus characterized by being equipped with click information generating means. 2. A document image processing apparatus for linking a plurality of data positions on a document captured as image data, the first image being a link target on the document image.
Range designating means for designating a range and a second range; quality evaluating means for evaluating the quality of character recognition of two ranges on the document image designated by the range designating means; Word recognition means for performing character recognition on one of the first or second range evaluated as being high, and registering a word recognized from this one range together with its position information, Character recognition is performed on the first or second other range evaluated as having low quality, and a character string recognized from the other range is compared with a word in one range registered by the word registration unit. A character string recognized from the other range by the word matching unit and a word in one range registered by the word registration unit. A link information generating unit that generates link information that associates the position information of the character string in the other range matched and matched with the position information of the registered word. Image processing device. 3. The quality evaluation unit measures region features in each of two ranges on the document image designated by the range designation unit, and character recognition of each range is performed according to the measured region features. 3. A quality evaluation means for evaluating the quality of the product.
Document image processing apparatus according to 1. 4. The quality evaluation unit performs character recognition by extracting a character string of all or a part of the two ranges on the document image specified by the range specification unit, 3. The document image processing apparatus according to claim 2, wherein the quality evaluation unit evaluates the quality of each range of the character recognition based on the result of the character recognition. 5. A document image capturing means for capturing a document including a table part and a drawing part as image data, and a range for extracting a range of a table part and a range of a drawing part on a document image captured by the document image capturing means. Extracting means for performing character recognition of a character string corresponding to an item name in a range of a table part of the document image extracted by the range extracting means, and registering a character recognized word as an item name of the range of the table part Word registration means to perform character recognition of a character string in a range of a drawing portion of the document image extracted by the range extraction means,
A word matching unit that matches a character string recognized from the range of the drawing portion with a word of an item name in the range of the table portion registered by the word registration unit; and a word matching unit that is recognized from the range of the drawing portion by the word matching unit. If the matched character string matches the word of the item name in the range of the table part registered by the word registration unit, the position information of the character string in the range of the drawing part matched and matched is stored in the table part. And a link information generating unit that generates link information associated with a registered word that is an item name in the range of. 6. A document image capturing means for capturing a plurality of documents as image data, and link information in which positional information on each document image is associated between the plurality of document images captured by the document image capturing means. Link information generating means, time information extracting means for extracting time information of each of a plurality of document images linked by the link information generating means, and each of the plurality of document images extracted by the time information extracting means. Link direction limiting means for limiting a reference direction between the plurality of document images performed based on the link information generated by the link information generating means according to a temporal order according to time information. Document image processing equipment Place. 7. A document image capturing means for capturing a plurality of documents as image data, and link information in which positional information on each document image is associated between the plurality of document images captured by the document image capturing means. Link information generating means, time information extracting means for extracting time information of each of a plurality of document images linked by the link information generating means, and the link information generated by the link information generating means. A sequence information adding unit that adds information on a temporal order according to time information of each of the plurality of document images extracted by the time information extracting unit when reference reading is performed between the plurality of document images. Document image to feature Processing equipment. 8. The link information generating means performs character recognition on one of the plurality of document images captured by the document image capturing means, and outputs a word recognized from the one document image. Word registration means for registering the document image together with the position information, and character recognition is performed on another document image of the plurality of document images captured by the document image capturing means, and the character is recognized from the other document image. Word matching means for matching a character string with a registered word from one document image registered by the word registration means, and a character string recognized from the other document image by the word matching means and the word registration means Registered words from one document image registered by Link information generating means for generating link information in which, when the collation matches, the positional information of the character string of the other collated document image and the positional information of the registered word are associated with each other. The document image processing device according to claim 6. 9. A document image processing method for linking a plurality of data positions on a document captured as image data, the method comprising:
A range specifying step of specifying a range and a second range, and performing character recognition on a first range on the document image specified by the range specifying step, and a word recognized from the first range And a word registration step of registering the position information together with the position information; and performing a character recognition on a second range on the document image designated by the range designation step, and converting a character string recognized from the second range into A word matching step of matching words in the first range registered by the word registration step; a character string recognized from the second range by the word matching step; and a first character string registered by the word registration step. When the words in the range are collated and matched, the position information of the character string in the collated and matched second range and the position information of the registered word are compared. Document image processing method for a link information generating step of generating the communication with the link information, that it consists characterized.