JPH1021336A - Method and device for discriminating character string classification - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically discriminate whether a character string included in an optically read original image is the character string in language using alphabets or the character string in Japanese. SOLUTION: Area separation is operated to a bit map picture, and a character area is separated (step S101). Then, the extraction of a line is operated to the separated character area (step S104). Next, the number of spaces having width more than a prescribed value is counted for each extracted line (step S106). Then, a rate R of the number of spaces counted in the step S106 to the width of the extracted line is calculated, and whether the line is a line in Japanese or a line in language using alphabets (English in this case) is discriminated based on the value of the R (steps S107 and S108). Then, the character recognizing processing of the line is operated by using either an OCR(optical character recognition) for Japanese or the OCR for English according to this discriminated result (steps S109 and S110).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character string type determining apparatus and method suitable for performing as preprocessing such as OCR (optical character recognition).

[0002]

2. Description of the Related Art In a modern society full of information, digitization of information for easy information management and retrieval is urgently desired. To digitize information, OCR that converts characters, which are images read by input devices such as scanners, into character codes
(Optical character recognition) is indispensable, and the recognition accuracy is steadily improving.

In this type of OCR, it is difficult to recognize alphabets (especially lowercase letters) in Japanese OCR due to the difference in pattern characteristics between Japanese (Kanji, Hiragana, Katakana) and English or other languages using alphabets. Have difficulty. Therefore, different recognition algorithms are used, or the pattern dictionary is switched even if the algorithms are the same. Therefore, there is a need for a technique for distinguishing between alphabets and Japanese.

Conventionally, there has been no method for distinguishing between the alphabetic language and the Japanese language, and the user has instructed the apparatus using an operation panel or the like according to the document to be processed.

[0005]

However, the method in which the user gives an instruction using an operation panel or the like requires a lot of trouble for the user, and further has the following problems.

When reading data of a document of a plurality of pages, an ADF (Auto Document Feeder) reduces the user's document replacement work. At this time, there are cases where English documents and Japanese documents are mixed in the plurality of pages. In such a case, the advantage of the ADF is ruined if the user instructs whether each page is Japanese or English every time one page is read. In addition, in order to request a user's instruction after reading all pages, a large amount of memory is required, which is not practical. For manuscripts where Japanese and English parts are mixed in one page, OCR
Cannot be switched properly.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and identifies, based on an optically read original image, whether a character string included in the image is a character string in an alphabetic language or a Japanese character string. It is an object of the present invention to provide a character string type determination device and method capable of performing the same.

[0008]

According to the present invention, there is provided a character string type determining apparatus for extracting a line from a character image, wherein the line extracted by the extracting means has a predetermined value or more. Counting means for counting the number of spaces having a width of, and determining whether to determine a Japanese line or a line in an alphabetic language based on the number of spaces counted by the counting unit and the width of the extracted line Means.

Preferably, the apparatus further comprises separating means for separating a character area from image data representing an original image.
The extracting means extracts a line from the character area separated by the separating means. This is because the character string type can be appropriately determined even for image data in which a character area, a graphic area, and the like are mixed in one page.

Preferably, the extracting means extracts a line by projecting the character image data in a predetermined direction.

Preferably, the counting means includes:
For the row extracted by the extraction means, comprises a detection means for detecting a white portion by projecting in a direction perpendicular to the row direction,
Of the white portions detected by the detection means, those whose width is equal to or more than a predetermined value are counted.

Preferably, the apparatus further comprises a removing unit for removing noise from the character image. This is because the number of spaces can be counted more accurately, and the accuracy of determining the type of a character string is improved.

Preferably, the apparatus further comprises correction means for detecting the inclination of the row extracted by the extraction means and correcting the inclination of the row. This is because the determination accuracy of the character string type can be maintained with respect to the inclination of the document.

Preferably, a direction detecting means for extracting a character from the character image and detecting a direction of the character to detect a direction of the character image, based on the direction detected by the direction detecting means. Rotating means for rotating the character image in a predetermined direction. This is because the type of the character string can be determined by the same algorithm regardless of the direction of the character in the character image. Also, for example, even if a plurality of character image regions exist in one page and the direction of the characters in each character image region is different, the type of the character string can be determined by performing image rotation for each character image. The type of character string can be determined by the same algorithm (that is, an algorithm corresponding to a character image in a predetermined direction).

[0015]

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a block diagram illustrating the configuration of the character recognition device according to the present embodiment. In the figure, reference numeral 11 denotes a CPU, which executes various controls in the character recognition device. Reference numeral 12 denotes a RAM, which provides a work area when the CPU 11 executes various processes. A ROM 13 stores various control programs for the character recognition device. The CPU 11 executes various control programs described later by executing a control program stored in the ROM 13. In this embodiment, the control program is R
Although the configuration held in the OM 13 has been shown, it is a matter of course that the control program may be loaded into the RAM 12 from an external storage device such as a hard disk or a floppy disk and executed by the CPU 11.

Reference numeral 14 denotes a disk interface.
The hard disk 141 and the floppy disk 142 are connected to the bus 18 of the character recognition device. Reference numeral 15 denotes a scanner, which optically reads a document image. The image read by the scanner 15 is converted into image data that can be processed by the CPU 11 and output to the bus 18. This image data is stored in the RAM 12, the hard disk 141, or the floppy disk 142 under the control of the CPU 11. An input unit 16 includes a keyboard and a pointing device. Reference numeral 17 denotes a display, which displays image data of a document image, a result of performing character recognition processing, and the like. Reference numeral 18 denotes a bus that connects the above-described components.

The ROM 13 stores a control program as shown in FIG. The area separation process 131 separates a text area composed of characters, a graphic area composed of figures, and the like from the image of the document image read by the scanner 15. The line extraction process 132 is a region separation process 131
Extract lines from the text area separated by. The space counting process counts the number of spaces exceeding a predetermined width among the spaces existing in the line for each line extracted in the line extraction process 132. The type determination process 134 determines the type (Japanese or alphabet language) of each line based on the number of spaces obtained by the space count process and the width of each line. The character recognition process 135 executes a character recognition process by appropriately switching between an algorithm for recognizing Japanese characters and an algorithm for recognizing characters in alphabetical languages based on the determination result obtained in the type determination process 134.

Hereinafter, each of the above processes will be described in detail.

FIG. 2 is a flowchart showing a procedure for determining the type of character string in the present embodiment. In this flowchart, the area separation processing 131 (step S101), the line extraction processing 132 (step S104), the space count processing 133 (step S106), the type determination processing (steps S107 and S108), and the character recognition processing 135 (steps S109 and S110) ).

In step S101, the bitmap image input by the image input device such as the scanner 15 is divided into regions, and blocks are divided for each image type. FIG. 3 is a diagram showing an example of a binary bitmap image of one bit per pixel. In this embodiment, a case where a simple binary image is processed will be described. FIG. 4 is a diagram illustrating a result when the area division processing is performed on the binary bitmap image of FIG.

FIG. 5 is a flowchart showing an example of the procedure of the area separation process. In step S301, m × n windowing processing is performed on the bitmap image. If even one pixel has black, the windowing area is set to black.
The area of the wind ring is preferably set according to the resolution. For example, 8 pixels × 300 pixels at 300 dpi
It is preferable to use about eight pixels. As a result, the resolution of the binary image is greatly reduced, and the character portions are connected. FIG. 6 is a diagram showing a state in which the original shown in FIG. 3 is processed by the resolution conversion in step S301 and the character portions are connected.

Next, when the contour is traced in step S302, it is possible to distinguish an elongated pattern (text portion) unique to a character from other figures. Finally, step S30
In the same group combining process of 3, the adjacent text portions are combined to complete the area separation as shown in FIG. FIG. 7 is a diagram illustrating an example of a data structure for holding a result of the area separation processing. In FIG. 7, (a)
FIG. 4 is a diagram illustrating a data type for holding a result of the region separation processing. In each of the divided areas, attribute information (type) indicating whether the image in the area is a title, text, graphic, or image, the position of each area (the upper left coordinate of the area), and the vertical and horizontal sizes Information (startx, st
arty, width, height). Also, next_address
Is a pointer indicating the storage location of the next area, if any. FIG. 7B shows an example of data storage based on the result of the area separation processing shown in FIG.

Returning to the description of FIG. When the above-described region separation processing in step S101 is completed, according to the extracted block data, until there is no unprocessed block in step S102 (the next_address of the structure is NU
Turn the following processing loop (until LL). That is, if it is determined in step S102 that there is no unprocessed block, the process ends, and if it is determined that there is an unprocessed block, the process proceeds to step S103.

In step S103, it is determined whether the block to be processed at present contains characters. That is, when the type of the structure described in FIG. 7 indicates title or text (when the value of type is 0 or 1), it is determined that the block includes characters, and the process proceeds to step S104. Otherwise, the process returns to step S102.

In step S104, data in the block is projected on the Y axis to extract a row. FIG. 8 is a diagram illustrating an example of execution of projection onto the Y axis. FIG. 7 shows a state in which projection is performed on text B in FIG. 4 as an example of execution of projection on the Y axis. Hereinafter, the projection on the Y axis will be described more specifically with reference to FIG.

FIG. 9 is a flowchart showing a processing procedure for projecting onto the Y axis in this embodiment. Step S90
In step S902, nline indicating the number of lines in the character image is reset, and in step S902, a flag for discriminating whether to search for the beginning of a line or the end of a line is reset. In step S903, the counter j of the loop in the Y direction is reset, and in step S904, the counter i of the loop in the X direction is reset. In step S905,
Flag indicating whether any black pixel exists in the i-th row
Reset kuro. Nline, flag, j, i, k
Each variable of uro is secured in the RAM 12.

Projection on the Y axis is performed using the above variables.
The outline of the flowchart of FIG. 9 will be described.
In the loop from 906 to S908, the range of 0 ≦ i ≦ widh (the width of the block) is scanned for the j-th row in the processing target block. Then, when a black pixel is detected, the processing in steps S909 to S912 causes
Line_sy with the j-th line as the starting position of the line where the character string exists
(line), and the process proceeds to the scanning of the next line (steps S913 to S945). Here, the flag is set to “1”, and the process shifts to a process of searching for the end of the line. In the process of searching for the end of the line, even if a black pixel is detected, the determination as to the step S910 does not hold the start position.

If no black pixel is detected in the loop of steps S906 to S908, step S91 is executed.
In step S3, it is determined whether the current process is a process for searching for the end of a line.
918 holds information indicating the height of the line where the character string exists. The above processing is repeated for the block.

Hereinafter, a more detailed description will be given with reference to the flowchart of FIG.

In step S906, the pixel pixel [startx +
i] [starty + j] is determined to be black. If black, the process proceeds to step S909. If white, the process proceeds to step S907.
Here, since startx and starty represent the upper left coordinates of the processing target block as described in the data structure of FIG. 7, the processing in steps S906 to S908 is performed in the starty + j-th row in the block. Each pixel is sequentially scanned in the x direction to determine whether or not the pixel is a black pixel.

If the result of the determination in step S906 is white, i is incremented to further process the next i (step S908), and the flow returns to step S906. The above processing is executed for the width of the block until a black pixel is detected.

When a black pixel is detected, the corresponding line (j-th line) is detected.
Since it is not necessary to scan the black pixels any more for line (line), the process exits the loop of steps S906 to S908 and proceeds to step S909.

In step S909, kuro is set to 1. Then, in step S910, it is checked whether or not flag is 0 (that is, whether or not the mode is a mode for searching for the beginning of a line of a character string). If the flag is 0, it means that the beginning of the line is being searched, and therefore, in step S911, line_sy [nlin
By assigning j to e], the start position of the line is held.
Then, in step S912, the flag is changed to 1, and the processing state is set to "the mode for searching for the end of the line of the character string". On the other hand, if the flag is 1 in step S910, it means that the end of the character string line is being searched, and the process proceeds to step S913 without performing anything.

The fact that the process has proceeded to step S913 means that
Indicates that the processing of the line (j-th line) has been completed. Therefore, in step S913, it is determined whether kuro is 0 (there is no black pixel on the line) and flag is 1 (the mode for searching for the end of the line). If this condition is satisfied, it indicates the end of the character string line, and the process proceeds to step S916. In step S916, the number of lines indicating the height of the character string line (j-line_sy [nline])
To line_h [nline]. Then, Step S91
7 returns the flag to the mode of searching for the beginning of a line (flag
Is reset to 0), nline is incremented in step S918, and the flow advances to step S914. On the other hand, step S9
If the condition is not satisfied in step 13, the process proceeds to step S914.

In step S914, it is determined whether j is smaller than height. If this condition is not satisfied, it means that the processing has been completed for all the lines in the block, and this processing is terminated. On the other hand, if the condition of step S914 is satisfied, since an unprocessed line exists in the block, j is incremented, and the process returns to step S904.

By the processing of FIG. 9, the number of lines (nline), the starting point (line_sy [nline]), and the height (line_h [nline]) of each line can be extracted from the character image.

Returning to the description of FIG. In step S105, a loop for processing each row extracted in step S104 is formed. When all the rows have been processed, the process returns to step S102 and proceeds to the next block. If there is an unprocessed row, the process proceeds from step S105 to step S106. In step S106, projection is performed on the X axis as shown in FIG. 10, and a space number counting process is performed. FIG. 10 is a diagram for explaining the projection process on the X axis and the space number counting process. It is clear that the procedure of the projection process on the X axis can be obtained by modifying the procedure of the projection process on the Y axis (FIG. 9), and thus the detailed description is omitted here.

FIG. 11 is a flowchart showing the procedure of the space number counting process in this embodiment. Here, it is assumed that the projection on the X axis has been completed, and the projection result is stored in SX [i].

First, variables are initialized in steps S1001 to S1004. That is, in step S1001, the character number counter n_char is set to 0, and the space number counter n_space
Is reset to -1. Where n_space is minus 1
The reason is that the white portion before the actual character portion starts is counted to be 0 (the first white portion is not counted as a space). In step S1002, the white dot counter count used to control the space number counter not to count the space less than the predetermined width is reset to zero. Further, step S10
In 03, a flag state indicating whether it is a character portion or a space portion is set to WHITE. Then, step S1004
Then, i for counting the width of the character area rectangle (block) is reset to 0.

In the following processing, the number of spaces in the line is counted. In step S1005, the width counter i
Is greater than or equal to the character area rectangle width, and if so, terminates this processing. If not exceeded, the flow advances to step S1006 to continue counting the number of spaces for the line.

In step S1006, it is checked whether the projection result SX [i] on the X axis is white or black, and the process branches. Here, when SX [i] is 1, it indicates that the projection result is black, and the process proceeds to step S1007.
On the other hand, when SX [i] is 0, it indicates that the projection result is white, and the process proceeds to step S1011.

In steps S1007 to S1010, if a white area has been detected (state is WHITE), n_space is incremented as the end of the space portion. Otherwise, the count is set to 0, and the process returns to step S1005. Step S1
In 011-S1015, after the black area is detected (state is BLACK
), The count is incremented.
If the white portion of the projection result exceeds the predetermined width, the count exceeds the predetermined value (LIMIT). At this point, it is determined that the black region has ended, and n_char is incremented.
Change state to WHITE to indicate detection of white area. By the above procedure, the number of white areas whose count exceeds LIMIT is n_
It will be stored in space.

Each step will be described. In step S1007, when the projection result is 1 (when it is black),
It is checked whether or not the flag state is WHITE. If YES, the process proceeds to step S1008, and if NO, the process proceeds to step S1008.
Proceed to 1010.

In step S1008, n_space is incremented assuming that the beginning of the character portion has been detected (ie, the end of the space portion has been detected). Then, in step S1009, the flag state is set to BLACK.
To In step S1010, the white dot counter coun
Set t to 0. Thereafter, the process proceeds to step S1016, where i is incremented so as to proceed to the next pixel.

On the other hand, when the projection result is 0 (when it is white), first in step S1011 the flag state is set to BLACK
It is checked whether it is, and if YES, step S101
If the answer is NO, the process proceeds to step S1016. In step S1012, the white dot counter count
Is incremented, and the number of continuous white dots is counted. In step S1013, the white dot counter count
Is compared with a LIMIT predetermined according to the resolution. If the LIMIT is larger than LIMIT, the process proceeds to step S1014 as the start of a space (end of the character portion), and otherwise proceeds to step S1016. Step S1014
Then, the character copy counter n_char is incremented. In step S1015, the flag state is set to WHITE to indicate the detection of a white area.

With the above processing, the number of spaces (n_space) in the extracted row is calculated.

Returning to the description of FIG. In step S107, a ratio R between the character area width width and the number of spaces n_space is obtained. In step S108, it is checked whether the ratio R is equal to or greater than a predetermined threshold T. here,
If YES, the line is determined to be in English, and the flow proceeds to step S110 to execute English OCR. If NO, the line is determined to be in Japanese, and the flow advances to step S109 to execute Japanese OCR. The operation of the space number counting process of the present embodiment will be described with reference to an actual sentence example shown in FIG.

FIG. 10A, which is a Japanese language, is originally divided into 16 parts by projection onto the X axis. However, by setting LIMIT appropriately in consideration of the resolution, most white parts disappear and only one space remains. LIMI
T can be obtained by, for example, character height × 0.357. If the character height is 7 mm as in this example, 7 ×
0.357 = 2.5 mm. Therefore, in this example, LIMI
A case is shown in which T is set to 2.5 mm and a white portion smaller than 2.5 mm is set to be canceled (in FIG. 10, the length of the white portion is shown in millimeters above the character). FIG. 10B, which is English, is originally divided into 30 pieces by the oblique shadow on the X axis. However, by setting LIMIT to cancel white areas less than 2.5 mm, there will be four spaces left.

In FIG. 10, the line (a) of the Japanese character string
And the number of spaces and the width widt of each of the lines (b) of the English string
By determining the ratio to h, it can be determined whether the language is Japanese or English. Here, the ratio R and the threshold T will be briefly described. In the present embodiment, the ratio R is: R = (number of spaces + 1) / (line width / (character height × 3.
2)) For example, in the case of Japanese in FIG. 10, the number of spaces = 1, the line width = 120.0 mm, and the character height = 7.
Since it is 0 mm, R = 0.37. FIG.
In the case of English, the number of spaces = 4, line width = 126.0m
m, character height = 7.0 mm, R = 0.888
Becomes As described above, since the ratio between Japanese and English obtained in step S107 is greatly different, it is possible to identify whether the language is Japanese or English from the ratio R by setting the threshold T to about 0.7, for example.

In the above-described embodiment, the line extracting means uses the oblique shadow of the character image on the Y axis, but the present invention is not limited thereto. For example, line extraction may be performed in the region separation processing in step S101 in FIG. Specifically, the resolution conversion processing in step S301 in FIG.
In the object extracted by the outline tracking process in 02, an elongated object is determined to be a character. After that, the regions are separated into rows without performing the same group combination in step S303. As a result of performing the area separation on the original image of FIG. 3 by this processing, the result becomes as shown in FIG. 12, and the oblique shadow on the Y axis becomes unnecessary.

Further, in the above-described embodiment, if noise is present in the one-bit image of one pixel to be subjected to the present process, the ability of this alphabetic Japanese discrimination cannot be utilized.
Therefore, it is preferable to remove isolated dots and the like that are clearly recognized as noise. As a method of removing isolated dots, there is a method of windling an image and erasing the pattern by pattern matching. For example, pattern matching is performed with a 3 dot × 3 dot window in which the center pixel is “1” and the surrounding eight pixels are “0”, and when matching is performed, the pixel corresponding to the center pixel is set to “0”. As a result, isolated black pixels can be erased.

In the above embodiment, the image to be subjected to the region separation processing in step S101 in FIG. 2 is a binary image of one bit per pixel, but is not limited thereto. For example, a multi-valued image of 8 bits per pixel may be used. In that case, as a region separation process, a high frequency component is extracted by applying a differential filter,
There is a method of dividing into a text part and a photograph part. In addition, in order to distinguish between the above-mentioned alphabet language and Japanese,
It is necessary to binarize the multi-valued image of 8 bits per pixel with a predetermined threshold value.

Further, it is preferable to correct the inclination of the character image before discriminating between the language using the alphabet and Japanese in the present embodiment. By performing such a correction,
Good results can be obtained even when the document image is tilted. The inclination correction can be realized, for example, by obtaining the inclination of the extracted row and performing coordinate transformation.

Further, by detecting the direction of the character image before performing the method of discriminating between the language using the alphabet and the Japanese language according to the embodiment, a result independent of the direction of the document can be obtained. For the detection of the character image direction, for example, a line is extracted from the character image extracted by region division, and character extraction is performed.
OCR is performed by rotating the extracted several images by 0 °, 90 °, 180 °, and 270 °. O for each rotation angle
Since the degree of certainty of the CR can be obtained, the direction of the character image is detected by using them, and the original image is rotated. That is, the direction in which the highest degree of certainty is obtained is set as the direction of the character image in the area.

If this original image rotation is performed on each character data block obtained by the area separation,
Even when the character directions of a plurality of character data blocks present in one page are different, the character string type can be accurately determined.

In the above-described embodiment, the line of the character image is extracted, and after extracting all lines, the character of the first line is extracted. However, the present invention is not limited to this. For example, a line extraction of a character image may be performed, and when the first line is found, character extraction may be performed on that line, and Japanese automatic discrimination may be performed using this alphabet. As a result, there is an advantage that a memory for storing the structure data of a row is smaller than in the above embodiment.

As described above, according to the above-described embodiment, it is possible to satisfactorily determine whether each character string line extracted from an image is Japanese or an alphabetic language, and perform an appropriate character recognition process according to the determination result. It is possible to switch. Even if the original is a mixture of Japanese and English, character recognition can be performed with high accuracy.

Even if the present invention is applied to a system composed of a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), a single device (for example, a copying machine, a facsimile, etc.) Device).

An object of the present invention is to provide a storage medium storing a program code of software for realizing the functions of the above-described embodiments to a system or an apparatus, and to provide a computer (or CPU) of the system or apparatus.
And MPU) read and execute the program code stored in the storage medium.

In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD
-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) May perform some or all of the actual processing, and the processing may realize the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into a memory provided on a function expansion board inserted into the computer or a function expansion unit connected to the computer, based on the instructions of the program code, It goes without saying that the CPU included in the function expansion board or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

When the present invention is applied to the storage medium, the storage medium stores program codes corresponding to the above-described flowcharts. Each module shown will be stored in a storage medium.

That is, at least the program code of each of the "extraction processing module", "count processing module" and "judgment processing module" may be stored in the storage medium. Also, as shown in FIG.
It is desirable to have a “separation processing module”.

Here, the extraction processing module is a program module for implementing an extraction processing for extracting a line from a character image. Also, the counting module
It is a program module that implements a count process for counting the number of spaces having a width equal to or greater than a predetermined value for a row extracted by the extraction process. The determination processing module is a program module that implements a determination process of determining whether a line is a Japanese line or a line in an alphabetic language based on the number of spaces counted by the counting process and the width of the extracted line. is there. Further, the separation processing module is a program module that realizes a separation process of separating a character region from image data indicating a document image and providing the character region to the extraction process as a character image. It is needless to say that the illustrated program module may include a program module for removing noise from the character image, a correction module for correcting the inclination of the document, and the like.

[0068]

As described above, according to the present invention,
From the optically read document image, it is possible to automatically identify whether the character string included in the image is a character string in the alphabet language or a Japanese character string.

For this reason, in the character recognition processing, it is possible to appropriately select the character recognition processing corresponding to each language.

[0070]

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a character recognition device according to an embodiment.

FIG. 2 is a flowchart illustrating a procedure for determining a character string type according to the embodiment.

FIG. 3 is a diagram illustrating an example of a binary bitmap image of one bit per pixel.

FIG. 4 is a diagram illustrating a result when an area division process is performed on the binary bitmap image of FIG. 3;

FIG. 5 is a flowchart illustrating an example of a procedure of a region separation process.

FIG. 6 is a diagram illustrating a state in which the document in FIG. 3 is processed by resolution conversion in step S301 and character portions are connected.

FIG. 7 is a diagram illustrating an example of a data structure for holding a result of the area separation processing.

FIG. 8 is a diagram showing an example of execution of projection onto a Y axis.

FIG. 9 is a flowchart illustrating a processing procedure of projection onto the Y axis in the embodiment.

FIG. 10 is a diagram illustrating a projection process onto the X axis and a space number counting process.

FIG. 11 is a flowchart illustrating a procedure of a space number counting process according to the embodiment.

FIG. 12 is a diagram illustrating a result obtained when an area division process including row extraction is performed on the binary bitmap image of FIG. 3;

FIG. 13 is a diagram showing an example of a memory map of a storage medium storing a control program according to the present invention.

Claims (16)

[Claims] 1. An extracting unit for extracting a line from a character image, a counting unit for counting the number of spaces having a width equal to or more than a predetermined value for the line extracted by the extracting unit, and a space counted by the counting unit A character string type determination device, comprising: determination means for determining whether the line is a Japanese line or a line in an alphabetic language based on the number of characters and the width of the extracted line. 2. The image processing apparatus according to claim 1, further comprising a separating unit that separates a character region from image data indicating a document image, wherein the extracting unit extracts a line from the character region separated by the separating unit. The character string type determination device according to the item. 3. The character string type judging device according to claim 1, wherein said extracting means extracts lines by projecting the character image data in a predetermined direction. 4. The counting means includes a detecting means for projecting a row extracted by the extracting means in a direction perpendicular to a row direction to detect a white part, and the white part detected by the detecting means. 2. The method according to claim 1, wherein: the number of which is greater than a predetermined value is counted.
The character string type determination device according to the item. 5. The character string type determination device according to claim 1, further comprising a removing unit that removes noise from the character image. 6. The character string type determination device according to claim 1, further comprising a correction unit that detects a slope of the line extracted by the extraction unit and corrects the slope of the line. 7. A direction detecting means for extracting a character from the character image and detecting a direction of the character, thereby detecting a direction of the character image; and detecting the character based on the direction detected by the direction detecting means. 2. The character string type determination device according to claim 1, further comprising: a rotation unit configured to rotate the image so that the image is oriented in a predetermined direction. 8. An extraction step of extracting a line from a character image, a counting step of counting the number of spaces having a width equal to or greater than a predetermined value for the line extracted by the extraction step, and a space counted by the counting step A character string type determination method, comprising: a determination step of determining whether the line is a Japanese line or a line in an alphabetic language based on the number of characters and the width of the extracted line. 9. The method according to claim 8, further comprising a separating step of separating a character area from image data representing a document image, wherein the extracting step extracts a line from the character area separated in the separating step. Character string type determination method described in. 10. The method according to claim 8, wherein in the extracting step, lines are extracted by projecting the character image data in a predetermined direction. 11. The counting step includes a detecting step of projecting a row extracted in the extracting step in a direction perpendicular to a row direction to detect a white portion, and detecting the white portion detected in the detecting step. 9. The method according to claim 8, wherein, among those, those whose width is equal to or more than a predetermined value are counted.
Character string type determination method described in. 12. The method according to claim 8, further comprising a step of removing noise from the character image. 13. The character string type determination method according to claim 8, further comprising a correction step of detecting the inclination of the line extracted in the extraction step and correcting the inclination of the line. 14. A direction detecting step of detecting a direction of the character image by extracting a character from the character image and detecting a direction of the character, and detecting the character based on the direction detected in the direction detecting step. The method according to claim 8, further comprising: rotating the image so that the image is oriented in a predetermined direction. 15. A computer-readable memory for storing a control program for determining a type of a character string in image data, wherein the code is for an extraction step of extracting a line from a character image, and is extracted by the extraction step. A code of a counting step for counting the number of spaces having a width equal to or more than a predetermined value for a row, and using a Japanese row or alphabet based on the number of spaces counted in the counting step and the width of the extracted row. A code for a determining step of determining whether the line is a language line. 16. The computer-readable memory according to claim 15, further comprising a code of a separation step of separating a character area from image data representing a document image and providing the character area to the extraction step as a character image.