JP4109738B2 - Image processing method and apparatus and storage medium therefor - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing method and apparatus and a storage medium therefor, and more particularly to processing for character recognition of an image of a document in which different languages are mixed.
[0002]
The present invention relates to an image processing method and apparatus and a storage medium therefor, and more particularly to processing for analyzing an image that can be divided into a plurality of blocks.
[0003]
[Prior art]
Conventional character recognition processing basically targets a single language, and for example, character recognition is performed using a Japanese character recognition device for Japanese and an English character recognition device for English. However, in the case of a specialized Japanese character recognition device, since the target character often includes an alphabet, it is also possible to recognize a language composed of the alphabet, for example, English. However, in such a case, it cannot be denied that the recognition accuracy of the English portion is deteriorated as compared with the English professional character recognition device.
[0004]
In order to eliminate these disadvantages, the user visually determines whether each sentence is English or Japanese, and instructs the Japanese character recognition by specifying the Japanese sentence part. There is also a character recognition device that designates a sentence portion and instructs to perform English-specific character recognition.
[0005]
[Problems to be solved by the invention]
However, in the character recognition device described in the above-mentioned conventional example, when a document in which different languages are mixed is recognized, character recognition is forcibly performed in Japanese for both Japanese and English sentences, or a sentence part is designated and recognized. As the instruction of language type must be repeated twice in Japanese and English, the recognition rate and usability are not compatible, and either one has to be taken. This has also been a major obstacle for the spread of character recognition devices.
[0006]
[Means for Solving the Problems]
In order to solve the above problems, an image processing method according to the present invention includes an extraction step of extracting a text region from an input image , and analyzing each image region by analyzing image characteristics for each text region extracted in the extraction step. The language type determination step for determining whether the language type is Japanese, English, or unknown, and setting the determined language type in each area, and any of the text areas extracted from the input image is Japanese If it is determined in the language type determination step that the language type of the word is determined, the language type of the text region determined to be unknown in the language type determination step is set to Japanese, while being extracted from the input image If it is determined in the language type determination step that all the text regions that have been determined are not Japanese language types, the text region determined to be unknown in the language type determination step The language type set by the language type determination step and the resetting step is set to English, and the language type set in the Japanese language region is suitable for Japanese character recognition. Character recognition is performed using a Japanese character recognition function capable of recognizing alphabets. On the other hand, the language type set by the language type determination step and the resetting step is English for the text region image in English. And a character recognition step for recognizing characters using an English character recognition function suitable for character recognition.
[0007]
In order to solve the above-described problems, an image processing apparatus according to the present invention includes an extracting unit that extracts a text region from an input image, and an image characteristic is analyzed for each text region extracted by the extracting unit. Language type determination means for determining whether the language type is Japanese, English, or unknown, and setting the determined language type in each region, and any of the text regions extracted from the input image is Japanese If the language type determining unit determines that the language type is the language type, the language type of the text region determined to be unknown by the language type determining unit is set to Japanese, while being extracted from the input image If the language type determination unit determines that all the text areas that have been processed are not Japanese language types, the language type of the text area that is determined to be unknown by the language type determination unit is set to English. Japanese characters that are suitable for Japanese character recognition and can also recognize alphabets for images in a Japanese text region whose language type set by the determination unit, the language type determination unit, and the resetting unit Recognize characters using a recognition function, and on the other hand, an English character recognition suitable for English character recognition for an image of an English text region whose language type is set by the language type determination unit and the resetting unit And character recognition means for recognizing characters using the function.
[0008]
In order to solve the above problems, the computer-readable storage medium of the present invention includes an extraction step of extracting a text region from an input image, and analyzing image characteristics for each text region extracted in the extraction step. A language type determination step for determining whether the language type of each text area is Japanese, English, or unknown, and setting the determined language type for each area, and any of the text areas extracted from the input image If the language type determination step determines that is a Japanese language type, the language type of the text area determined to be unknown in the language type determination step is set to Japanese, while the input If it is determined in the language type determination step that all the text regions extracted from the image are not Japanese language types, it is determined that the text type is unknown in the language type determination step. A reset step for setting the language type of the text region set to English, and a language type set by the language type determination step and the reset step for a Japanese text region image. Character recognition is performed using a Japanese character recognition function suitable for character recognition and also capable of recognizing the alphabet, while the language type set by the language type determination step and the resetting step is for an image in an English text region Thus, a control program for causing a computer to execute each step of character recognition using an English character recognition function suitable for English character recognition is stored.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
[0024]
FIG. 5 is a block diagram of an apparatus according to the present invention.
[0025]
A CPU (Central Processing Unit) 51 controls processing according to the present invention in accordance with a control program stored in the MEM 53. The process shown in the flowchart described later is also executed under the control of the CPU 51. 53 is a MEM (comprising a RAM and a ROM), and various data such as a control program for processing executed by the CPU 51, various parameters used for the processing, input images, and a dictionary for character recognition are stored in the MEM 53. . Reference numeral 54 denotes a display such as a CRT or LCD, which displays an input image, text of a processing result, an operation instruction screen, a recognition result read from a file in accordance with document identification information specified by the input means 59, and the like. The character recognition result stored in the MEM 53 is displayed on the display 54, and editing such as selecting a correct character from a plurality of candidate characters using the input means 59 is possible. A printer 55 such as an LBP or BJ printer prints an image, text, or the like. A scanner 56 optically reads an image of a document and inputs it to the apparatus as an electrical signal. A communication I / F 57 controls data transmission / reception via a public line or a LAN. The input of an image according to the present invention and the output of the result of processing the image can be transmitted and received with other terminals via this communication I / F. Reference numeral 58 denotes a storage medium that can be attached to and detached from the apparatus and can be read and written by a computer, such as a CD-ROM, CR-R, and FD. The image according to the present invention may be read from the storage medium 58 and the processing result may be written to the storage medium 58. Further, the control program stored in the MEM 53 may be installed from another terminal via the communication I / F, or may be installed from the storage medium 58. Reference numeral 59 denotes input means such as a keyboard and a pointing device, and instructions from the operator are given via this input means 59. Reference numeral 50 denotes a bus, which exchanges data between each means.
[0026]
FIG. 1 is a process schematic diagram of an apparatus according to the present invention. In the present embodiment, a case where Japanese and English are recognized as examples will be described.
[0027]
In FIG. 1, reference numeral 2 denotes an image input unit for inputting a document image, which is input from another terminal or a storage medium 58 via a scanner 56 or a communication I / F 57 and stored in the MEM 53. 4 is a region identifying unit for the input document image, 6 is a Japanese / English discriminating unit for discriminating whether the region is Japanese or English, 8 is a Japanese / English discriminating control unit for controlling the Japanese / English discriminating unit, and input means In response to whether or not language type automatic discrimination by 59 is to be performed, and in response to an input of a recognition mode instruction in the case where automatic discrimination is not to be performed, a control signal representing the instruction is generated. Reference numeral 10 denotes a character recognition unit for recognizing Japanese and English characters, and each of these processes is executed by the CPU 51 in accordance with a control program stored in the MEM 53 as shown in a flowchart described later.
[0028]
Next, the operation will be described.
[0029]
The document image acquired from the image input unit 2 is divided into small regions (hereinafter referred to as blocks) for each attribute in the region identification unit 4 as shown in FIG. 2, and the block number, block attribute, and block size are divided. Identify the location of the block. This area identification processing performed by the area identification unit 4 analyzes the arrangement of black pixels in the input document image, determines attributes such as text, figure, image, separator, etc. Are identified as one block. Also, even if the text is the same, if the line direction is different or the line spacing is wider than the specified value, it is determined that the text is different in level and is identified as a different block.
[0030]
When the block is identified, a block number is automatically assigned to each block, for example, in order from the top, and width information and height information indicating the size of the block, and the position of the block are defined. Coordinate data is set and stored in the MEM 53 as block data. By referring to this block data, it is possible to extract a desired one block image of the input image, or to reproduce the arrangement of text, diagrams, images, separators, etc. in the document. On the other hand, in the Japanese / English discrimination control unit 8, the user sets whether or not to perform Japanese / English automatic discrimination, and the recognition language mode (when the user does not set automatic Japanese / English discrimination) ( A control signal corresponding to Japanese recognition or English recognition) is generated and input to the Japanese-English discriminating unit 6. The Japanese / English discriminating unit 6 determines whether the character area is Japanese or English based on the control signal generated by the Japanese / English discriminating control unit 8 and the region attribute set by the region identifying unit 4.
[0031]
The image data of the character region to which the Japanese or English attribute is added is sent to the character recognition unit 10, and character recognition corresponding to the language specified by each attribute is performed. The character recognition unit 10 performs a Japanese character recognition routine in S406 and an English character recognition routine in S408, which will be described later. This may include separate character recognition separate units, or cut out character images. A recognition dictionary may be provided for each language, and a common matching algorithm may be used.
[0032]
When the character recognition process in the character recognition unit 10 is completed, the result of character recognition is added with document identification information and stored (filed). This document identification information may be input by a keyboard operation by the user, or may be extracted from a character recognition result. Alternatively, the input date / time, time, user ID, and the like may be automatically given. It may be different for each document so that it can be used as an index for later reading from a file. The data to be filed may include not only the character recognition result but also the layout information of the input image and the data area where the input image is compressed.
[0033]
Here, the Japanese / English discrimination unit 6 and the character recognition unit 10 will be described in detail.
[0034]
FIG. 3 is a flowchart showing the processing flow of the Japanese-English discriminating unit 6.
[0035]
S302 is a routine for reading a control signal generated by the Japanese / English discrimination control unit 8 and stored in the MEM 53 in accordance with a user instruction. When there are a plurality of document images to be processed, the control signal stored in the MEM 53 is selectively read in association with the image currently specified as the processing target. This control signal represents whether or not automatic Japanese / English discrimination is performed and the recognition language mode when it is not performed.
[0036]
In S304, it is determined whether or not to perform Japanese-English automatic discrimination based on the control signal read in S302. If so, control is passed to S306, and if not, control is passed to S322.
[0037]
S306 is a routine for reading the attribute of each block divided as shown in FIG. 2 by the area identification unit 4, and reads the block data stored in the MEM 53 one by one in the order of the block numbers.
[0038]
In S308, based on the block attribute in the block data read in S306, it is determined whether or not the block is a text area and the process should proceed to S310. Control is transferred to discrimination processing. If it is determined in S308 that the block is not a text area and should not proceed to S310, the control is returned to S306 and the next block data is read from the MEM 53. In the example of FIG. 2, since the text attributes are given to the blocks 2, 3, and 4, these blocks are determined to be sentence regions, and are the target blocks for the Japanese-English discrimination processing in S310 and thereafter.
[0039]
S310 is a routine for analyzing the image characteristics of the block to determine whether the language type is Japanese or English, and is specified based on the block position and size data in the block data read in S306. This process is performed on the image of the corresponding area from the input image of the obtained MEM 53. This can be achieved by applying, for example, a technique for automatically discriminating language types such as Japanese and English from the features of input images described in Japanese Patent Application Laid-Open Nos. 8-339424 and 8-305792.
[0040]
In S312, based on the result determined in S310, a new attribute, that is, a language type, is set in the block data and stored in the MEM 53. Three types are set here: Japanese, English, and unknown. “Unknown” is set when the probability of being Japanese and the probability of being English are lower than a threshold in the Japanese-English discrimination routine in S310.
[0041]
In S314, it is determined whether or not the language type determination process in S306 to S312 has been performed for all blocks of the document image stored in the MEM 53. If all the processes have been completed, the process proceeds to S316. Control is transferred to S306.
[0042]
S316, S318, and S320 check the language type of all blocks in the document. If there is any Japanese block in the block, unknown block is set as Japanese block, otherwise it is all English block. If it is, it is a routine for determining the language type of the English block and the unknown block to be set. In this routine, in S316, the attributes related to the language types of all the blocks stored in the MEM 53 are read to determine whether there is at least one block in which the Japanese attribute is set. The language type is set to Japanese for the block whose attribute of the block read in S316 is unknown and stored in the MEM 53 (S318), and it is determined that there is no Japanese block in S316. In this case, the language type is set to English for the block whose attribute of the block read in S316 is unknown and stored in the MEM 53 (S320). Note that the determination criterion of S316 may be determined in advance, and a predetermined number (1, 2,...) Of a specific language type block is a predetermined ratio (5%, 10%,...). There may be.
[0043]
Even if the unknown block is actually an English block even if it is set as a Japanese block in S318, the Japanese character recognition performed by the character recognition unit 10 includes alphabet recognition (ie, Japanese (In addition to kana, kanji, numbers, symbols, etc., the alphabet is also recognized as a recognition character), so it will not become unrecognizable, but in the opposite case, that is, if the Japanese block is set incorrectly as an English block Because it falls into impossibility, it means that the person who makes such a judgment is strictly conducted. That is, it is easier to select a language type having a large number of characters to be recognized. More preferably, when the recognition target character of one language type includes the recognition target character of the other language type as described in the present embodiment, it is easy to select the language type of the other language type. Good.
[0044]
S322 is a routine for setting a recognition language mode when it is determined in S304 that automatic Japanese-English discrimination is not performed. This setting is performed based on a control signal generated by the Japanese / English discrimination control unit 8 according to the language type operated and instructed by the operator via the input means 59.
[0045]
S324 is a routine for setting all the text areas in the document to Japanese and storing them in the MEM 53 when it is determined in Japanese language recognition mode in S322.
[0046]
S326 is a routine for setting all the text areas in the document to English and storing them in the MEM 53 when the English recognition mode is determined in S322.
[0047]
FIG. 4 is a flowchart showing the flow of character recognition processing performed by the character recognition unit 10.
[0048]
S402 is a routine for reading, from the block data stored in the MEM 53, block data of a sentence block whose attribute is set as text and image data in the block.
[0049]
S404 is a routine for determining the language type attribute of the block based on the attribute of the block data read in S402. Attributes determined by this routine are the attributes of the language type set in S310, S318, S320, S324, and S326.
[0050]
S406 is a routine for performing Japanese character recognition in the character recognition unit 10 when it is determined in S404 that the language is Japanese.
[0051]
S408 is a routine for performing English character recognition in the character recognition unit 10 when it is determined that the English is determined in S404.
[0052]
S410 determines whether or not the character recognition processing of S402 to S408 has been performed on all the text blocks of the document stored in the MEM 53. If there are still processing blocks, the control proceeds to S402, and the next Read a sentence block.
[0053]
As described above, according to the present invention, a Japanese-English automatic discrimination technique is used as a core, a user-friendly interface is provided for the control, and an unknown post-processing is also performed for unknown block processing. Since a character recognition device can be constructed, recognition accuracy and usability can be improved at the same time even in a document in which Japanese and English are mixed, which contributes greatly to the spread of the character recognition device.
[0054]
In this embodiment, the language type discrimination between Japanese and English has been described as an example. However, the present invention is not limited to this, and other languages may be discriminated. In this case, basically, the S310 Japanese / English discriminating unit in FIG. 3 can be dealt with only by replacing it with a language type discriminating routine suitable for it.
[0055]
【The invention's effect】
As described above, according to the present invention, it is possible to construct a character recognition device that provides a user-friendly interface for the control based on automatic language type discrimination technology, and also considers post-processing at the time of erroneous determination. Even in a document in which different languages are mixed, recognition accuracy and usability can be improved at the same time, which can greatly contribute to the spread of the character recognition function.
[0056]
As described above, according to the present invention, a sentence divided into a plurality of regions included in a manuscript is distinguished for each region, and character recognition is performed according to the language type of each region. Recognition of mixed documents can be performed at high speed with high accuracy.
[Brief description of the drawings]
FIG. 1 is a process schematic diagram of a character recognition apparatus according to the present invention. FIG. 2 is an exemplary diagram of region identification results. FIG. 3 is a flowchart showing a process flow of a Japanese / English discrimination unit. FIG. 5 is a configuration diagram of an apparatus according to the present invention.

Claims (4)

An extraction step for extracting a text region from the input image;
By analyzing image characteristics for each text area extracted in the extraction step, it is determined whether the language type of each text area is Japanese, English, or unknown, and the determined language type is assigned to each area. and language type determination step of setting,
If any of the text regions extracted from the input image is determined to be a Japanese language type in the language type determination step, the language type of the text region determined to be unknown in the language type determination step On the other hand, if it is determined in the language type determination step that all the text regions extracted from the input image are not Japanese language types, it is determined that the language type is not known in the language type determination step. A reset step for setting the language type of the designated text area to English,
A Japanese character recognition function suitable for Japanese character recognition and capable of recognizing the alphabet is used for an image of a text region in which the language type set in the language type determination step and the resetting step is Japanese. On the other hand, for an image of an English text area whose language type is set in the language type determination step and the resetting step, an English character recognition function suitable for English character recognition is used. image processing method for a character recognizing character recognition step, and Turkey to have a characterized. An automatic determination permission / inhibition step for determining whether or not to automatically determine the language type;
If it is determined in the automatic determination enable / disable step that language type automatic determination is instructed , the language type of each area is set in the language type determination step and the resetting step , and the setting is performed in the character recognition step. Character recognition based on the language type
On the other hand, if it is determined in the automatic determination availability step that the language type automatic determination is not instructed and the language type is instructed by the user, in the character recognition step, the language instructed by the user The image processing method according to claim 1, wherein the image of each region is recognized based on a seed . Extraction means for extracting a text area from the input image;
By analyzing image characteristics for each text area extracted by the extraction means, it is determined whether the language type of each text area is Japanese, English, or unknown, and the determined language type is assigned to each area. and language type determination means for setting,
The language type of the text region determined to be unknown by the language type determination unit when the language type determination unit determines that any of the text regions extracted from the input image is a Japanese language type On the other hand, if it is determined by the language type determination unit that all the text regions extracted from the input image are not Japanese language types, the language type determination unit determines that the text region is unknown. Resetting means for setting the language type of the designated text area to English,
A Japanese character recognition function suitable for Japanese character recognition and capable of recognizing the alphabet is used for an image of a Japanese text region whose language type is set by the language type determination unit and the resetting unit. On the other hand, the language type set by the language type determination unit and the resetting unit uses an English character recognition function suitable for English character recognition for an image of an English text region. the image processing apparatus according to claim and Turkey that Yusuke and character recognition character recognition means. An extraction step for extracting a text region from the input image;
By analyzing image characteristics for each text area extracted in the extraction step, it is determined whether the language type of each text area is Japanese, English, or unknown, and the determined language type is assigned to each area. and language type determination step of setting,
If any of the text regions extracted from the input image is determined to be a Japanese language type in the language type determination step, the language type of the text region determined to be unknown in the language type determination step On the other hand, if it is determined in the language type determination step that all the text regions extracted from the input image are not Japanese language types, it is determined that the language type is not known in the language type determination step. A reset step for setting the language type of the designated text area to English ,
A Japanese character recognition function suitable for Japanese character recognition and capable of recognizing the alphabet is used for an image of a text region in which the language type set in the language type determination step and the resetting step is Japanese. On the other hand, for an image of an English text area whose language type is set in the language type determination step and the resetting step, an English character recognition function suitable for English character recognition is used. A computer-readable storage medium storing a character recognition step and a control program for causing a computer to execute each step .

Images