JP2005071014A - Document image processor, document image processing method, program and storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document image processor for automatically classifying document images by using character recognition at the time of registration to quickly and highly precisely classify document images. <P>SOLUTION: As for area-divided and attribute-added document images, an area to retrieve keywords is selected. For example, "title", "header", and "caption" are defined as the retrieval targets of keywords. As for the selected area, character recognition is executed, and the keywords are retrieved. The keyword retrieval is executed by extracting keywords matched with the keywords in a keyword management table. When the keywords are detected, link information to an input document image is added to a group associated with the keywords. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a document image processing apparatus, a document image processing method, a program, and a storage medium that store document images in a searchable system, and more particularly to a document image processing apparatus that performs automatic classification using character recognition at the time of registration. The present invention relates to a processing method, a program, and a storage medium.

  A document image processing apparatus such as an electronic document management system is configured to store an electronic document including character information and to search the electronic document using a keyword.

  A document to be stored is input as document image data by an image scanner or the like, and the document image data is registered (filed) in a folder or the like.

  Some of these electronic document management systems are capable of hierarchical grouping using filing cabinets, folders, etc. when filing document images, making it easier to specify search ranges and more operability. An excellent filing function is realized.

  However, in general, the hierarchy of document image registration is left to the operator's judgment, and the operator must grasp the filing system in order to judge the document image to be registered. Moreover, the registration work itself with such a determination is complicated.

Therefore, an electronic document management system has been proposed in which a hierarchy for classifying an input document image is registered in association with a keyword. In this electronic document management system, the input document image is character-recognized, a keyword is searched for the text data of the character recognition result, and the document image is classified into a corresponding hierarchy based on the most detected keyword. (See Patent Document 1)
There has also been proposed an electronic document management system for storing a form layout and its title position in a layout dictionary. In this electronic document management system, matching processing between the input form and the layout dictionary is performed, and characters and the like written in the title position are read in the layout regarded as matching, and the form is specified. (See Patent Document 2)
JP 09-282328 A JP-A-10-240958

  However, the electronic document management system of Patent Document 1 recognizes all the characters included in the document image, so that the registration process takes a long time, and the classification priority is determined by the number of detected keywords. When many words different from the keyword of the hierarchy to be generated appear, the possibility of misclassification increases.

  In the electronic document management system of Patent Document 2, since the character at the title position is read in the specified layout, registration processing cannot be performed unless the form has a layout from which keywords can be extracted, and the application is limited. Furthermore, the complicated work of storing the form in advance in the layout dictionary is indispensable, and classification registration is not considered.

  The present invention has been developed to solve such conventional problems, and an object thereof is to classify document images with high speed and high accuracy.

  A document image processing apparatus according to the present invention includes a keyword management unit that manages a keyword for classifying a document image into a plurality of groups and a group to be classified, and a text from an image of a predetermined attribute area in the document image. Character recognition means for acquiring information, keyword search means for searching whether or not the keyword is included in the acquired text information, and at least one of the document images based on a search result by the keyword search means Classification means for classifying into groups.

  A document image processing method according to the present invention is a document image processing method in which keywords for classifying document images into a plurality of groups are associated with classification target groups, and the document images are classified into groups based on the keywords. A character recognition step of acquiring text information from an image of a predetermined attribute area in the document image, a keyword search step of searching whether or not the keyword is included in the acquired text information, and the keyword search step. A classification step of classifying the document images into at least one group based on the search result.

  According to the present invention, document images can be classified with high speed and high accuracy.

  Next, a first embodiment of a document image processing apparatus and a document image processing method according to the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a first embodiment of a document image processing apparatus according to the present invention, and FIG. 2 creates a new group in the first embodiment of the document image processing method executed by the document image processing apparatus of FIG. 3 is a flowchart showing the document image input process in the first embodiment of the document image processing method, FIG. 4 is a flowchart showing the region dividing and attribute adding process in FIG. 3, and FIG. FIG. 6 is a table showing a keyword management table generated in the first embodiment of the document image processing method. FIG. 6 is a diagram showing an example of a group and hierarchy system generated in the first embodiment of the document image processing method. FIG. 8 shows a screen for generating a new group in the first embodiment of the document image processing method, and FIG. 8 shows a first processing rule for area classification in steps S404 and 405 of FIG. 9 is a diagram showing a second processing rule for area classification in steps S404 and 405 of FIG. 4, and FIG. 10 is a diagram showing a third processing rule for area classification in steps S404 and 405 of FIG. 11 is a diagram showing a fourth processing rule for area classification in steps S404 and 405 of FIG. 4, FIG. 12 is a diagram showing a fifth processing rule for area classification in steps S404 and 405 of FIG. FIG. 14 is a diagram showing a sixth processing rule for area classification in steps S404 and S405 in FIG. 4, and FIG. 14 is a memory of a storage medium for storing various data processing programs that can be read out in the first embodiment of the document image processing apparatus. FIG. 15 is a table for explaining the map, and FIG. 15 is a table showing a table of the number of links managed in the modification of the document image processing method.

  In FIG. 1, the document image processing apparatus of this embodiment includes an external storage device 104 such as a magnetic disk and an image scanner 108, and reads document image data from the external storage device 104 or directly reads a document image document. Read with. The document image processing apparatus includes a CPU 101 for overall control, and a control program for the CPU 101 is stored in the ROM 102. The document image input from the image scanner 108 is held in the RAM 103, and processing for classification and registration is executed by the control program in the ROM 102. A keyword management table or the like for classification and registration is stored in the ROM 102 together with the control program, or is read from the external storage device 104 into the RAM 103 for each necessary keyword management table.

  If necessary, preprocessing such as noise removal of the input document image may be executed prior to the document image processing.

  In FIG. 1, 105 is a display, 106 is a keyboard, and 107 is a pointing device such as a mouse. Reference numeral 109 denotes a network interface, which can communicate with a device located at a remote place (not shown) to transmit / receive a program, a recognition dictionary, data, and the like.

  In addition, the image scanner 108 and the display 105 are not limited to those built in the document image processing apparatus, but are configured as external devices connected to the document image processing apparatus via an external interface so that this embodiment is configured. It doesn't matter.

  The present invention can also be implemented by a general-purpose computer. In this case, a control program provided by a medium or the like may be stored in the external storage device 104 and executed by the CPU 101 according to an operator's instruction or the like.

  Next, a document image processing method executed by the document image processing apparatus of FIG. 1 will be described.

  FIG. 6 shows an example of a system of groups and hierarchies of document images classified by the document image processing method.

  In FIG. 6, a “system department report” group for registering a system department report and a “development department” group for registering a development department report are provided under the highest group “report”. ing. The “Development Department” group has a “Development Section Report” and an “Evaluation Section Report” group for each section.

  The group and hierarchy system of FIG. 6 can be displayed on the display 105 by the operation of the operator's keyboard and pointing device.

  When creating a new group, the hierarchy to which the group belongs is specified in the display of FIG. 6, and the folder name and keyword of the group are specified by the input box shown in FIG. Keywords are used for group classification.

  For example, when a group of “Management Section Report” is newly generated under the “Report” group of FIG. 6, the hierarchy to be registered in FIG. 6 is specified, and the folder name “Management” is entered in the input box of FIG. Enter the section report and the keyword "management section" and register.

  FIG. 2 is a flowchart showing a new group generation process.

  First, in step S201, the operator specifies a group hierarchy, instructs the group creation using the keyboard 106 and the mouse 107, and inputs a group name and a keyword for classification.

  In step S202, the keyword management table shown in FIG. 5 is corrected.

  In the keyword management table, a folder name, a keyword, a related group list, and a document image in which a link is generated are registered.

  For example, the keyword management table in the state where the new group of FIG. 7 is added to the group and hierarchy system of FIG. 6 is added by creating a “Management Section Report” folder as shown in FIG.

  That is, the report group has the folder name “Report”, the keyword “Report”, and the related groups “System Department Report”, “Development Department”, and “Management Section Report”. The group of the system department report has the folder name “system department report”, the keyword “system department”, and the related group “report”. In the group of the development department, the folder name is “development department”, the keyword is “development department”, the related groups are “report”, “development department report”, and “evaluation section report”. In the management section report group, the folder name is “management section report”, the keyword is “management section”, and the related group is “report”. The group of the development section report has the folder name “development section report”, the keyword “development section”, and the related group “development department”. In the evaluation section report group, the folder name is “evaluation section report”, the keyword is “evaluation section”, and the related group is “development department”. Further, for each group, a list of document images in which links are generated is registered.

  For example, links to system department reports 1 to j are generated in the system department report group, links to development department reports 1 to k are generated in the development section report group, and evaluation section report groups. The link for the evaluation section reports 1 to m is generated.

  FIG. 3 is a flowchart showing processing when inputting a document image.

  First, in step S301, the document image is read and stored by the image scanner 108 or the like, and the input image is divided into regions in step S302, and attributes are added to the divided regions. The processing in step S302 will be described in detail with reference to FIG.

  Note that image processing for improving the division accuracy may be performed before the region division.

  In step S303, a region for searching for a keyword is selected for the document image after region division and attribute addition. For example, characters in an area to which attributes such as “title”, “heading”, and “caption” are assigned are searched for keywords. This selection condition is determined in advance, but a method of changing the condition by an operator's operation can also be adopted. As a result, it is possible to specify an area where a keyword is highly likely to exist, and to increase the search processing speed.

  Character recognition is executed for the selected area in step S304, and further, in step S305, it is searched whether a keyword is included in the character recognition result. When character recognition is performed, character recognition may be performed using an original image corresponding to the selected region, or may be performed after conversion to an image having a resolution suitable for character recognition. Further, in the keyword search of the present embodiment, whether or not the character recognition result partially matches with the keyword is searched, but a complete match may be used instead of the partial match.

  Based on the search result, a group that matches the searched keyword is identified in the keyword management table of FIG.

  In step S306, link information to the stored input document image is added to the group associated with the searched keyword.

  If the character recognition result matches a plurality of keywords in the keyword management table, link information may be added for all groups corresponding to each keyword. In this case, the same document image can be referred to from a plurality of groups. As shown in FIG. 15, by registering the number of groups associated with one document image as the number of links, management at the time of document image deletion is facilitated.

  As for the matching between the character recognition result and the keyword, a method of adding a link may be adopted even if the keyword does not completely match in order to prevent omission of classification due to erroneous recognition of character recognition. For example, it is determined that the keyword “occupational disease” matches the keyword even if the recognition result is “intellectual disease”. Thereby, the redundancy of character recognition can be increased.

  Further, a combination of candidate characters extracted in the process of character recognition processing may be created, and a candidate that matches a keyword may be extracted from the combination of candidate characters. This can further increase the redundancy of character recognition.

  As a group classification method, it is also possible to associate a plurality of keywords with one group. In this case, a plurality of keywords are designated by the result of operation with a logical operator such as “and” or “or”. Also good.

  The process of step S302 is executed according to the flowchart of FIG.

  First, in step S401, in order to thin out (reduce) the input document image, the input document image is divided into predetermined M × N pixel unit areas (M and N are preset integers, for example, 16 pixel × 16 pixel unit areas). To do.

  Proceeding to step S402, in each divided area (for example, 64 pixels if 8 × 8 pixels), an area where no black pixel exists is 0 (white), and if there is even one black pixel, 1 (black) is set. Create a thinned image.

  For example, the input document image is divided into regions of 8 × 8 = 64 pixels. If there is even one black pixel among the 64 pixels, the area is set to black, thereby generating a thinned image of 1/64 pixels of the input document image.

  Thereby, the area to be processed is reduced, and the processing speed can be increased.

  Next, in steps S403, S404, and 405, regions having the same attribute are connected based on a predetermined rule to generate a small rectangle.

  For example, when a small rectangle is generated by extracting connected pixels in a thinned image, several adjacent characters are extracted as one small rectangle because the thinned image is used. Further, as a first predetermined rule, in the small rectangle group, when the aspect ratio of each small rectangle is large and adjacent rectangular short sides are close to each other, the small rectangle groups (CL1 to CL3) are combined to form a new one. Small rectangles are generated and classified into character lines (see FIG. 8).

  As a second predetermined rule, in the small rectangle group classified into the character line, when the short sides of the small rectangles are substantially the same and the small rectangles are arranged at almost equal intervals, the small rectangle group is referred to as “text”. The attribute of (SN1) is given (see FIG. 9).

  As a third predetermined rule, a small rectangle group (CAP) adjacent to the top or bottom of the graph gives a “caption” attribute (see FIG. 10).

  As a fourth predetermined rule, among the character lines that are not the body attribute, the small rectangle group (T1) that exists above the document image (DOC) and has a large character size gives the attribute of “title” (see FIG. 11). ).

  As a fifth predetermined rule, among the character lines that are not the body part, the small rectangle group (T2) that is located between the body parts (SN1, SN2) gives the attribute of “heading” (see FIG. 12). .

  As a sixth predetermined rule, an area that is not determined to be a character line is determined to be “noise” (N) if the area size is small (see FIG. 13).

  Rules for classifying “graphs” and “images” are set for areas other than character lines. For example, as a seventh rule, a halftone photo area or an irregular drawing area is given an “image” attribute. As a eighth rule, a line drawing area mainly including a table, a figure, etc. Attribute.

  The region classification method is not limited to the above, and other region classification methods may be used.

  In step S403, each area is classified into a character line area and other areas based on the first rule.

  In step S404, based on the second to fifth rules, small rectangles of “text”, “caption”, “title”, and “heading” are generated, and regions that are not classified into these are classified into others. To do.

  In step S405, “noise” is extracted based on the sixth rule, and “graph” and “image” are classified based on the seventh and eighth rules for the area that is not determined to be a character line. .

  Based on the above classification results, attributes of “text”, “title”, “caption”, “headline”, “other character line”, “graph”, and “image” are added to the small rectangle in step S406. “Noise” is deleted in step S407.

  Since the attribute is given to the divided area in this way, as described above, based on the area selected based on the attribute in step S303, the keyword having the desired attribute can be recognized and the keyword can be obtained. It becomes like this.

  As described above, according to the first embodiment, it is possible to classify into a plurality of groups based on the character recognition result of a region having a desired attribute, and the possibility of being classified into a group different from the operator's expectation is reduced. There is an effect to. Further, even if erroneous recognition occurs in character recognition, matching with a keyword is performed, and there is an effect that classification is performed without omission.

  FIG. 14 is a diagram showing a memory map of a storage medium for storing various data processing programs that can be executed by the document image processing apparatus according to the present invention.

  In the memory map, the directory information of each program and data is stored in the “directory information”, the processing program of the flowchart of FIG. 2 is stored in the “first data processing program”, and the “first data processing program” ”Stores the processing program of the flowcharts of FIGS. 3 and 4.

  Although not specifically illustrated, data depending on various programs, such as a recognition dictionary, is also managed in the directory, and a program for installing various programs in a computer may be stored.

  Next, a second embodiment of the document image processing apparatus and the document processing method according to the present invention will be described with reference to the drawings.

  FIG. 16 is a table showing keyword management tables generated in the second embodiment of the document image processing method, FIG. 17 is a flowchart showing the document image input process in the second embodiment of the document image processing method, and FIG. FIG. 10 is a flowchart showing document image deletion processing in Embodiment 2 of the document image processing method.

  In the second embodiment, each of a plurality of operators can set a keyword and a group for classifying document images independently.

  The document image processing apparatus according to the second embodiment is expressed by the block diagram of FIG.

  In FIG. 16, the keyword management table generated in the second embodiment manages operators and groups (folder names) corresponding to each keyword.

  In the keyword management table of FIG. 16, the operator 1 creates two groups “folder 11” and “folder 12”, associates the keyword “A” with “folder 11”, and associates the keyword “C” with “folder 12”. On the other hand, the operator 2 creates three groups “Folder 21”, “Folder 22”, and “Folder 23”, and adds keywords “B” and “C” to “Folder 21”, “Folder 22”, and “Folder 23”. , “D” are associated with each other.

  That is, since the keyword management table is managed with the operator as one attribute, various search and registration modes can be adopted as compared with the first embodiment.

  In the flowchart of FIG. 17, the document image input process is executed by the following process.

  Steps S401 to S404 in the flowchart of FIG. 17 are the same as steps 301 to S304 (FIG. 3) of the first embodiment, and a description thereof is omitted.

  In step S405, the keyword range can be limited without making the search keyword for the character recognition result the entire keyword management table. That is, the group of classification destinations can be limited.

  That is, the classification range of the input document image can be limited to the operator 2, for example. As a result, only the keywords set by the operator 2 are searched, which is effective for managing the forms that the operator 2 wants to manage individually. Further, when managing document images to be shared by all operators, all keywords are set as a search range. When document images to be shared by several operators are managed, the keyword of each of the operators is used as a search range.

  Next, in step S406, a keyword is searched based on the search range set in step S405. In step S407, link information is added to all groups associated with the matched keyword, and FIG. Similarly, information on the number of links is stored in the document image.

  For example, when the keyword “C” is searched, link information is added to “folder 12” and “folder 22”, and the number of links is increased by “2”.

  In the flowchart of FIG. 18, when the operator deletes a document image, the following processing is executed.

  For example, the document image (FIG. 16) linked to “Folder 12” and “Folder 22” has the number of links “2”. In step S501, the operator links the document image from “Folder 22”. When deleted, in step S502, the document image processing apparatus determines whether or not the number of links of the document image is “1”.

  Here, since the number of links is “2”, the process advances to step S503, and the document image processing apparatus deletes the link information for the document image in “folder 22” and reduces the number of links of the document image by “1” to “1”. "

  On the other hand, when the number of links of the document image is “1”, the number of links becomes “0” due to deletion from the group in step S501, and thus the process proceeds from step S502 to step S504. delete.

  In the second embodiment, keywords can be individually set by a plurality of operators, and document images can be classified into respective groups.

  Further, since links to a plurality of groups are managed by the number of links, it is not necessary to generate a document image copy for each group, and a storage area can be saved.

  In the second embodiment, a keyword and a folder are set for each of a plurality of operators (users). For example, a keyword and a folder are set for each department (affiliation group) instead of the operator. It does not matter.

  The document image processing apparatus and the document image processing method according to the present invention reduce the processing time by performing region segmentation and recognizing only necessary attributes without requiring a complicated operation at the time of registration. . In addition, since only a portion including a character string effective for classification is processed, document images can be classified with high accuracy.

1 is a block diagram showing a first embodiment of a document image processing apparatus according to the present invention. FIG. (Example 1) 6 is a flowchart illustrating a process for creating a new group in the document image processing method according to the first embodiment, which is executed by the document image processing apparatus in FIG. 1. (Example 1) It is a flowchart which shows the process of the document image input in Example 1 of a document image processing method. (Example 1) FIG. 4 is a flowchart showing region division and attribute addition processing in FIG. 3. FIG. (Example 1) It is a table | surface which shows the keyword management table produced | generated in Example 1 of a document image processing method. (Example 1) It is a figure which shows the example of the system of the group and hierarchy produced | generated in Example 1 of the document image processing method. (Example 1) It is a figure which shows the screen for producing | generating a new group in Example 1 of a document image processing method. (Example 1) It is a figure which shows the 1st process rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure which shows the 2nd processing rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure which shows the 3rd process rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure which shows the 4th processing rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure which shows the 5th process rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure which shows the 6th process rule of the area | region classification in step S404, 405 of FIG. (Example 1) It is a figure explaining the memory map of the storage medium which stores the various data processing program which can be read by Example 1 of a document image processing apparatus. (Example 1) It is a table | surface which shows the table | surface of the number of links managed in the modification of a document image processing method. (Example 1) It is a table | surface which shows the keyword management table produced | generated in Example 2 of a document image processing method. (Example 2) It is a flowchart which shows the process of the document image input in Example 2 of a document image processing method. (Example 2) It is a flowchart which shows the document image deletion process in Example 2 of a document image processing method.

Explanation of symbols

101 CPU
102 ROM
103 RAM
104 External storage device 105 Display 106 Keyboard 107 Pointing device 108 Scanner unit 109 Network interface

Claims (20)

A keyword management unit that associates and manages keywords for classifying document images into a plurality of groups and a group to be classified;
Character recognition means for acquiring text information from an image of a predetermined attribute area in the document image;
Keyword search means for searching whether or not the keyword is included in the acquired text information;
Classification means for classifying the document images into at least one group based on a search result by the keyword search means;
A document image processing apparatus comprising: The keyword search means searches for keywords managed by the keyword management means by limiting the group to be classified.
The document image processing apparatus according to claim 1, wherein: The keyword management means manages an operator in association with the keyword and the group,
The document image processing apparatus according to claim 2, wherein the keyword search unit searches for the keyword by specifying a target operator to limit the group to be classified. The document image processing apparatus according to claim 1, wherein the character recognition unit can change the predetermined attribute that is a target of character recognition. The document image processing apparatus according to claim 1, wherein the keyword search unit searches whether or not the keyword partially matches in the text information. 5. The document image processing apparatus according to claim 1, wherein the keyword search unit searches whether or not the keyword completely matches in the text information. 7. The keyword search unit according to claim 1, wherein the keyword search unit searches for whether or not the keyword is included using all or part of a combination of candidate characters obtained by the character recognition unit. Document image processing apparatus. The document image processing apparatus according to claim 1, wherein the classification unit classifies the document image by generating link information that associates the group with the document image. The classification means associates the document image with a link number indicating the number of groups to which the document image belongs,
When the document image is to be deleted from any group, the link information of the document image in the group is deleted, the number of links of the document image is reduced, and the number of links becomes “0” Delete means for deleting the document image,
The document image processing apparatus according to claim 1, further comprising: A document image processing method for associating a keyword for classifying a document image into a plurality of groups and a group to be classified, and classifying the document image into a group based on the keyword,
A character recognition step of acquiring text information from an image of a predetermined attribute area in the document image;
A keyword search step for searching whether or not the keyword is included in the acquired text information;
A step of classifying the document images into at least one group based on a search result obtained by the keyword search step;
A document image processing method comprising: The keyword search step searches the keywords managed in the keyword management step by limiting the group to be classified.
The document image processing method according to claim 10. The document image processing method manages an operator in association with the keyword and the group,
12. The document image processing method according to claim 11, wherein the keyword searching step searches for the keyword by specifying a target operator to limit the group to be classified. The document image processing method according to claim 10, wherein the character recognition step can change the predetermined attribute to be character-recognized. 14. The document image processing method according to claim 10, wherein the keyword search step searches whether or not the keyword partially matches in the text information. The document image processing method according to any one of claims 10 to 13, wherein the keyword search step searches whether or not the keyword completely matches in the text information. 16. The keyword search step according to any one of claims 10 to 15, wherein the keyword search step searches for whether or not the keyword is included using all or part of a combination of candidate characters obtained by the character recognition step. Document image processing method. 17. The document image processing method according to claim 10, wherein the classifying step classifies the document image by generating link information that associates the group with the document image. The classification step associates the document image with a link number indicating the number of groups to which the document image belongs,
When the document image is to be deleted from any group, the link information of the document image in the group is deleted, the number of links of the document image is reduced, and the number of links becomes “0” A deletion step of deleting the document image
The document image processing method according to claim 10, further comprising: A computer-executable program including program code for causing a general-purpose computer to execute each step of the document image processing method according to claim 10. A computer-readable storage medium storing the computer-executable program according to claim 19.

Images