JP3720405B2 - Region identification apparatus and method - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an area identification apparatus and method, and more particularly to an area identification apparatus and method for identifying an area in an image.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, there is an apparatus that optically reads a business card, specifies an item of each character image on the business card, recognizes each item, stores each item, and searches efficiently.
[0003]
This conventional item specifying processing method will be described with reference to a flowchart shown in FIG. 9 and an apparatus configuration diagram shown in FIG.
[0004]
Here, reference numeral 801 denotes an image input unit for inputting a character pattern, and reference numeral 802 denotes a data separation unit for separating the input character pattern data by a predetermined method and cutting characters in each region. The data separation unit 802 also stores the number of regions separated from the character pattern data, the number of characters, and how they are arranged. There are various known methods for region separation and character cutting, and further, a method for determining the arrangement of characters, and the details thereof will not be described here.
[0005]
However, in the embodiment, when individual character patterns are distributed within an appropriate distance from each other, the character patterns are determined to be the same region. However, as described above, a known technique is used. May be.
[0006]
A control unit 803 controls the entire apparatus. A character recognition unit 804 receives character pattern image data from the control unit 803, performs character recognition for each character, acquires a character code, and stores the character code in the character code storage unit 809. 805 is a character used for character recognition A data ROM for recognition 806 is an item specifying unit that receives a character code (group) from the control unit 803 and specifies an item of the character code (group), 807 is an item specifying data ROM used for specifying an item, and 808 is specified. And an item data storage unit for storing the items and their data.
[0007]
Next, the item specifying process in a certain area will be described. Now, the region separation of the character pattern data input to the image input unit 801 and the character cut processing in each region are completed, and the character pattern data is stored in the data separation unit 802 and the number of separated regions and It is assumed that the number of characters in each area and the order of characters are stored. At this time, the control unit 803 transfers the character pattern data for one character in a certain area to the character recognition unit 804 according to the order of characters.
[0008]
If the character recognition unit 804 determines in step S901 that the character pattern data has been transferred from the control unit 803, the character code storage unit 809 is cleared in step S902, and the number of characters N = 1 is set. In step S903, character recognition processing is performed. That is, the received character pattern data is converted into a recognition format and matched with the data in the character recognition ROM 805 to perform character recognition.
[0009]
As a recognition format or matching method, for example, there is a method in which character pattern data is converted into a normal vector and vector matching is performed. The recognized character code is stored in the character code storage unit 809 in step S904. In step S905, it is determined whether recognition processing has been performed for all character pattern data in the area. This is performed by issuing an in-region end inquiry signal from the character recognition unit 804 to the control unit 803 and the control unit 803 responding to the inquiry signal.
[0010]
If it is determined that recognition processing has not been performed for all character pattern data in the area, N = N + 1 is set in step S906, and the flow advances to step S907 to request transfer of character pattern data for the next character. In response to the transfer request, the control unit 803 transfers character pattern data for the next character in step S908. The transferred character pattern data undergoes character recognition processing in step S903, and the recognition result is stored in addition to the character code data stored in the character code storage unit 809 in step S904.
[0011]
In this way, the process proceeds to step S905, where it is determined whether the recognition process has been performed for all the character pattern data in the area, and the above process is repeated until it is determined that the recognition process in that area has been completed.
[0012]
If it is determined in step S905 that the recognition process has been performed for all character pattern data in the area, the process proceeds to step S909, and the item of the area is determined based on the data stored in the character code storage unit 809. Identify. As for the item specifying processing method here, for example, a method for matching all characters, a feature character is extracted from characters in a region regardless of the order of characters, and matching with the feature character is performed. A method for specifying an item is known.
[0013]
The item specifying result is transferred to the control unit 803 in step S910, and the area data and the item information are associated with each other and stored in the item data storage unit 808 in step S910. As for this storage method, for example, it is conceivable to store and hold in the form as shown in FIG.
[0014]
In FIG. 10, the item data includes an address, a company name, a name, and a telephone number.
[0015]
[Problems to be solved by the invention]
However, in the conventional method, since all character recognition processing in each area is performed and then the item identification processing is performed, the character recognition processing takes time, and as a result, much time is required until the item identification is completed. There was a problem.
[0016]
[Means for Solving the Problems]
And [Action]
The present invention has been made in view of such problems, and an object of the present invention is to provide an area identification apparatus and method that can specify the type of an area composed of character pattern groups at high speed.
[0017]
In order to solve this problem, for example, the region identification device of the present invention has the following configuration. That is,
An area identification device for determining an item attribute of a document image including an area composed of a plurality of character images,
A first character string data for specifying the attribute of the region, a storage unit storing a second character string data to identify that it is not a predetermined attribute,
Character recognition means for sequentially recognizing character images in the region and outputting a recognition result every time one character is recognized;
Each time a recognition result is output from the character recognition means, a second character string data stored in the storage means is a character string composed of the recognition result and the recognition result output so far. Exclusion target item storage means for sequentially storing item attributes corresponding to the second character string data determined to match as item attributes to be excluded when it is determined that they match. When,
Each time a recognition result is output from the character recognition means, the first character string corresponding to the item attribute stored in the exclusion target storage means from the first character string data stored in the storage means After the data is excluded, it is determined whether the character string composed of the recognition result and the recognition result output so far matches with any of the first character string data after the exclusion , And determining means for determining an item attribute corresponding to the specified first character string data as an item attribute of the region when the matching first character string data can be specified .
[0018]
Here, according to a preferred embodiment of the present invention, the order of the character pattern images recognized by the character recognition means is from the top to the bottom if the arrangement of the character pattern images in the region is vertical, and to the left if horizontal. It is desirable to go from right to right. As a result, a correct character string can be extracted, the configuration of the determination unit can be simplified, and the storage content of the storage unit can be simplified.
[0019]
The character string unique to each area stored in the storage means is a character string that is permitted to be used in each item, and if it is determined by the determination means to be matched, the area corresponding to the matched character string It is desirable to use the classification type as the identification result. As a result, the area identification is completed at the time of collation, and the number of characters to be recognized can be minimized.
[0020]
The character string unique to each area stored in the storage means is a character string that is not permitted to be used in each item, and corresponds to the matched character string when it is determined by the determination means. The region types may be sequentially excluded, and when only one region type remains, the region type may be used as the recognition result. As a result, candidates are narrowed down each time collation is performed, so that the type of area can be identified efficiently. Furthermore, a means for storing data for each type of the identified area may be provided. As a result, a database can be easily constructed.
[0021]
【Example】
Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.
[0022]
FIG. 2 is a block configuration diagram of the electronic device in the present embodiment. In the figure, 201 is an image input unit (for example, an image scanner device) for reading a character pattern image such as a business card, 202 is an area separation of the input character pattern data by a preset method, and character cutting in each area is performed. It is a data separation unit that performs. The data separation unit also stores the number of areas separated from the character pattern data, the number of characters, and how they are arranged.
[0023]
203 is a control unit that controls the entire apparatus, 204 is a character recognition unit that receives character pattern image data for each character sent from the control unit 203 and performs character recognition to acquire a character code, and 205 is used for character recognition. A character recognition data ROM 206, an item specifying unit for receiving a character code from the character recognition unit 204 and specifying an item of the character code (group), 207, an item specifying data ROM used for specifying the item, 208, is specified. And an item data storage unit (an external storage device such as an HDD or the like if a database is used).
[0024]
Next, the operation will be described in detail. Now, when a character pattern image is input via the image input means 201, the data is sent to the image data separation unit 202 and separated into areas for each item by a preset method. Further, the separated data is separated into character pattern data for each character by a preset method, and the number of areas and the number of characters separated from the character pattern data are also stored. Next, the control unit 203 sends the character pattern data to the character recognition unit 204 character by character based on the data group.
[0025]
Here, the order of sending the character pattern data depends on how the characters are arranged in the area. Here, considering the arrangement of sentences in Japanese, when the character pattern is vertical writing, it is from the upper character to the lower character, and when writing horizontally, from left to right. Note that these directions may be specified by the user or automatically determined.
[0026]
The character recognition unit 204 converts the received character pattern data into a recognition format, matches it with the data in the character recognition ROM 205, and performs character recognition. As a recognition format or matching method, for example, there is a method in which character pattern data is converted into a normal vector and vector matching is performed.
[0027]
The recognition result is sent to the item specifying unit 206 as a character code. The item specifying unit 206 uses the character code and the item specifying ROM 207 to specify items in an area including the character code. Here, the item is specified every time a character code is sent to the item specifying unit 206, but if the item cannot be specified by the character code (group) sent so far, the next character is sent to the control unit 203. When a data request signal is sent and an item can be specified, the item data is sent. If the received signal is the next character data request signal, the control unit 203 extracts the next character pattern data from the data of the data separation unit 202 and sends it to the character recognition unit 204. If the received signal is item data, the item data and character code data are stored in the item data storage unit 208, and the character pattern data of the next item is extracted from the data of the data separation unit 202, and the character recognition unit 204 Send to. Here, when the item identification is completed for all items, the process is terminated.
[0028]
Here, the details of the item specifying process in one area will be described with reference to the flowchart of FIG. In the present embodiment, processing when the item is “address” will be described. FIG. 3 shows a detailed hardware configuration diagram of the item specifying unit 206. 301 is a character code storage unit for storing the character code sent from the character recognition unit 204, 302 is a character code count unit for counting the number of character codes sent from the character recognition unit 204, and 303 is a character code storage unit. This is a matching unit that performs matching between 301 and the item specifying data ROM 207. FIG. 4 shows an example of the contents of the item specifying data ROM 207.
[0029]
First, it is assumed that character pattern data for one character is sent from the control unit 203 to the character recognition unit 204 in step S101. Then, in step S102, the character recognition unit 204 sends a clear signal to the item specifying unit 206, clears the character code storage unit 301, and sets the count number N = 1 of the stored character code count unit 302. In step S103, the character pattern data is recognized. In step S <b> 104, the character code that is the recognition result is transferred to the item specifying unit 206. The item specifying unit 206 stores the transmitted character code in the character code storage unit 301 in step S105. Next, in step S106, matching processing is performed by the matching unit 303 on the data having 1 character between the stored data and the item specifying data ROM 207.
[0030]
If there is no match in the matching process with the item specifying data ROM, the process proceeds to step S107, and 1 is added to the count N of the stored character code count unit. In step S108, the control unit 203 is requested to transfer the Nth character pattern data. In response to the request, the control unit 203 transfers the character pattern data of the Nth character to the character recognition unit 204 in step S109. In step S103, the character recognition unit 204 recognizes character pattern data for the Nth character. When the recognition result is obtained, the character code data is transferred to the item specifying unit 206 in step S104. In step S105, the item specifying unit 206 adds the character code to the character code data stored in the character code storage unit 301. The process then proceeds to step S106, and between the stored data and the item specifying data ROM 207 again. A matching process is performed with data having N characters. These processes are repeated until they match in the matching process.
[0031]
For example, in the case of a character code “Shimomaruko, Ota-ku, Tokyo”, it matches the data of “Tokyo” in the item specifying data ROM 207 when the third character is sent. Accordingly, at this point, it is found that the area belongs to the item “address”.
[0032]
If they match in the matching process with the item specifying data ROM, the process advances to step S110 to send the specified item data to the control unit 203. As a result, the control unit 203 knows that the item specifying process has ended, and transfers all character pattern data and item data in the area to the item / data storage unit 208 in step S111. If there remains an area that does not specify an item, the character code and item information of the next area are transferred to the item specifying unit.
[0033]
By performing the above processing, there is an effect that the item specifying process of the character pattern data in the separated area can be performed in a shorter time than the conventional method. In addition, there is an effect that the amount of item specifying data can be made smaller than before.
[0034]
<Second embodiment>
In the above embodiment (first embodiment), the item specifying unit 206 specifies a character pattern in a certain area as a certain item, but this may be used to determine that this is not a certain item. This example will be described as a second embodiment.
[0035]
FIG. 5 shows an example of the contents of the item specifying data ROM 207 in the second embodiment. Here, reverse specific items are provided as compared with FIG. Here, reverse specification is used to mean that it is not a certain item.
[0036]
FIG. 6 shows a detailed hardware configuration block diagram of the item specifying unit 206 in this embodiment. For example, when the character pattern data is “Iueo Co., Ltd.”, the first character is recognized by the character recognition unit, the character code is sent to the item specifying unit 207, and when the data 501 is matched, It is determined that the item is not “address”. Therefore, in the recognition after the second character, the data whose item is “address” is not matched. The contents of the item specifying data ROM here are based on the fact that in Japan, a character string with katakana at the beginning is not an address. In any case, since there is no possible address as an item, the matching with the second and subsequent characters will be performed by excluding the “address” as an item from the beginning. The items to be referred to will be narrowed down gradually.
[0037]
FIG. 7 shows an operation flowchart of the item specifying process in a certain area in this embodiment. Only operations that differ from the processing in the first embodiment will be described below.
[0038]
In step S701, the matching unit 601 performs matching processing on the data having the reverse specific item whose number of characters is N between the storage data stored in the character code storage unit 301 in step S105 and the item specifying data ROM 207. To do. If there is matching data, the reverse specific item indicated by the data is stored in the reverse specific item storage unit 602 in step S702. Next, the process proceeds to step S703, and the matching process is performed only for data having N characters except for the items stored in the reverse specific item storage unit 602 in the item specifying data ROM. If they match, the process proceeds to step S110, and if they do not match, the process proceeds to step S107.
[0039]
If the processing as described above is performed, it is detected that the character pattern of a certain area does not belong to a certain item, and the matching of the item data is not performed. There is an effect that the item specifying process can be performed in a shorter time.
[0040]
In the embodiment, only the part for determining the item attribute of the character in the input image has been described. However, application of this technique may be applicable to various fields.
[0041]
For example, in a system for creating a business card database, the business card is read by an image scanner, the item type is automatically identified, and the image of the item or the character code of the recognition result is stored in an external storage device such as an HDD. It is also possible to search freely. In particular, when creating a database as a character code, all characters will be recognized, but the time required to specify at least the type of item in each area is reduced.
[0042]
The image to be input is not limited to a business card. This is because any document image can be processed in the same way as long as the image conforms to some rules. However, in general, a business card is clearly arranged as to what the characters written there mean, as is well known, and is therefore a good target for identifying items and creating a database.
[0043]
In the embodiment, an example in which an image is optically read from a scanner or the like has been described. However, for example, an image stored in a storage medium may be input and processed, or may be input via a communication line or the like. Since it is adaptable, it is not limited by the said Example.
[0044]
In the embodiment, the character recognition data (recognition dictionary) and the item specifying data are both ROM, but may be a rewritable storage device (for example, an external storage device such as RAM or HDD). The processing program need not be stored in the ROM. In other words, the technology disclosed in the embodiments can also be realized by loading a general-purpose information processing apparatus as a program from the outside. Further, it can be realized by a single device or a system constituted by a plurality of devices.
[0045]
As described above, according to this embodiment, there is an effect that the item specifying process of the character pattern data in the separated area can be performed in a shorter time than the conventional method. Also, there is an effect that the amount of item specifying data can be reduced as compared with the conventional method.
[0046]
【The invention's effect】
As described above, according to the present invention, it is possible to specify the type of the area constituted by the character pattern group at high speed.
[0047]
[Brief description of the drawings]
FIG. 1 is an operation flowchart of item specifying processing in a first embodiment;
FIG. 2 is a block configuration diagram of an apparatus according to the first embodiment.
FIG. 3 is a block configuration diagram of an item specifying unit in the first embodiment.
FIG. 4 is a diagram showing an example of contents of an item specifying data ROM in the first embodiment.
FIG. 5 is a diagram showing an example of contents of an item specifying data ROM in the second embodiment.
FIG. 6 is a block diagram of an item specifying unit in the second embodiment.
FIG. 7 is an operation flowchart of an item specifying process in the second embodiment.
FIG. 8 is a block diagram of a conventional apparatus.
FIG. 9 is an operation flowchart of an item specifying process in a conventional apparatus.
FIG. 10 is a diagram showing an example of a storage form of items / data in the prior art.
[Explanation of symbols]
201 Image input unit 202 Data separation unit 203 Control unit 204 Character recognition unit 205 Character recognition data ROM
206 Item specifying unit 207 Item specifying data ROM
208 items / data storage

Claims (6)

An area identification device for determining an item attribute of a document image including an area composed of a plurality of character images,
A first character string data for specifying the attribute of the region, a storage unit storing a second character string data to identify that it is not a predetermined attribute,
Character recognition means for sequentially recognizing character images in the region and outputting a recognition result every time one character is recognized;
Each time a recognition result is output from the character recognition means, a second character string data stored in the storage means is a character string composed of the recognition result and the recognition result output so far. Exclusion target item storage means for sequentially storing item attributes corresponding to the second character string data determined to match as item attributes to be excluded when it is determined that they match. When,
Each time a recognition result is output from the character recognition means, the first character string corresponding to the item attribute stored in the exclusion target storage means from the first character string data stored in the storage means After the data is excluded, it is determined whether the character string composed of the recognition result and the recognition result output so far matches with any of the first character string data after the exclusion , And determining means for determining an item attribute corresponding to the specified first character string data as an item attribute of the area when the matching first character string data can be specified. An area identification device.   The character recognition of the character image by the character recognition means is performed in order from top to bottom if the character image in the region is vertical writing, and from left to right if it is horizontal writing. 2. The area identification device according to 1. 2. The area identification device according to claim 1 , further comprising storage means for storing all the character images included in the area for which the item attribute is determined by the determining means together with the item attribute. An area identification method for determining an item attribute of a document image including an area composed of a plurality of character images,
A character recognition step of sequentially recognizing character images in the region and outputting a recognition result each time a character is recognized;
Each time a recognition result is output from the character recognition step, a character string composed of the recognition result and the recognition result output so far is not a predetermined item attribute stored in advance in the storage means. It is determined whether it matches any of the second character string data for specifying this, and if it is determined to match, the item attribute corresponding to the second character string data determined to match is the item to be excluded Exclusion target item storage step of sequentially storing in the exclusion target item storage means as an attribute,
Each time a recognition result is output from the character recognition step, it is stored in the exclusion target storage means from the first character string data for specifying the item attributes of the area stored in advance in the storage means. After excluding the first character string data corresponding to the item attribute, the first character string data after the character string constituted by the recognition result and the recognition result output so far is excluded. When the first character string data that matches the first character string data can be identified , the item attribute corresponding to the identified first character string data is determined as the item attribute of the area. A region identifying method comprising: a determining step for determining as The character recognition of the character image in the character recognition step is performed in the order from top to bottom if the character image in the region is vertical writing, and from left to right if it is horizontal writing. 5. The area identification method according to 4 . The area identifying method according to claim 4 , further comprising an accumulating step of accumulating all character images included in the area for which the item attribute is determined in the determining step together with the item attribute.

Images