JP2002366899A - Method and device for character information recognition - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device for character information recognition which can efficiently detect an error in character recognition and correct character data even if the number of recognized deviates. SOLUTION: For the recognizing method for recognizing character information from image information, a plurality of optical character recognition devices are used and if characters recognized by the optical character recognition devices are different, a character recognized by making a majority decision among the optical character recognition devices is determined. One arbitrary optical character recognition device among the optical character recognition devices is determined as a standard optical character device and the majority decision is made while the recognition results of other optical character recognition devices having performed recognition different in the character recognition number of a row-by-row or column-by-column character recognition from the recognition result of the standard device are excluded from reference to the majority decision making.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character information recognizing method for recognizing character pattern data (hereinafter referred to as character data) from image information including characters, and more particularly, to optically reading image information from an image and reconstructing the character data. The present invention relates to an optical character recognition device (hereinafter referred to as OCR) for recognition, and a character information recognition method and device for detecting an error in character recognition and correcting character data.

[0002]

2. Description of the Related Art In recent years, the use of DTP in prepress in printing has rapidly progressed, and a large amount of character data has been read from image information. Accordingly, the accuracy of character recognition by 0CR is also required to be accurate, but there is still no effective means for completely eliminating a reading error. For this reason, it is necessary to manually detect an error in character recognition and correct character data, which increases time and labor.

On the other hand, there is known a method of determining character information recognition by majority decision using a plurality of OCRs to mutually complement reading errors. However, even with this method, one character may not be recognized or two characters may be originally recognized due to a character position shift such as insertion of a character font and its size, line spacing, and Western font. In such a case, in character recognition by majority vote, the inconvenience that the characters recognized correctly after the occurrence of such a difference in the number of recognized characters is regarded as a different character by majority rule. Will occur.

[0004]

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and an object thereof is to solve the problem of character recognition even if the number of character recognition is shifted. It is an object of the present invention to provide a character information recognizing method and a character information recognizing device capable of efficiently detecting an error and correcting character data.

[0005]

According to the present invention, there is provided a method for recognizing character information from image information using a plurality of optical character recognition devices. In the character information recognition method for determining a character recognized by a majority decision between a plurality of optical character recognition devices when there is a difference in characters recognized between the plurality of optical character recognition devices, the plurality of optical character recognition Any one of the devices may be a standard optical character recognition device, and the other optical character recognition devices for the recognition result may recognize the different number of character recognition of the character recognition result in units of rows or columns. This is a character information recognition method characterized by performing a majority decision by excluding a recognition result of an optical character recognition device from a majority decision reference.

According to the present invention, by this means, a majority decision is made only for those having the same character recognition number for each row or column.
It is possible to make a majority decision in a state where the characters to be compared always match, and it is possible to perform more accurate character recognition.

According to a second aspect of the present invention, there is provided a recognition method for recognizing character information from image information, wherein a plurality of optical character recognition devices are used. In some cases, in a character information recognition method for determining a character recognized by majority vote between a plurality of optical character recognition devices, any one of the plurality of optical character recognition devices as a standard optical character recognition device, The other optical character recognition device for the recognition result measures the matching rate of the character recognition result in units of rows or columns, and when the matching rate is equal to or less than a predetermined value, the other optical character recognition for the row or column. This is a character information recognition method characterized in that a majority decision is made by excluding a recognition result of the apparatus from a majority decision reference.

According to the present invention, by this means, the recognition result when the character recognition number is shifted near the end of each row or column can be adopted as a majority decision object, and more accurate character recognition can be performed. Becomes

According to a third aspect of the present invention, the image information and the character information are displayed in a hierarchical display by color-coding the result of the majority decision for each character. 2. The character information recognition method according to any one of 2.

According to the present invention, by this means, the result of the majority decision is color-coded and superimposed and displayed on the character of the original image information, so that the comparison of the character in the image information detects the unrecognized error at the time of the majority decision. The result can be visually recognized, and the error can be more easily detected.

According to a fourth aspect of the present invention, in the character information recognition apparatus as set forth in the third aspect, the color classification of each character as a result of the majority decision is displayed by changing a color gradation for each of the coincidence rates. Is the way.

According to the present invention, it is possible to visually recognize the result of the coincidence rate by changing the color gradation for each coincidence rate and displaying the result, thereby making it easier to detect an error. Become.

According to a fifth aspect of the present invention, there is provided the image processing apparatus according to the first aspect of the invention, further comprising:
Regarding the majority decision between the recognition result of the character information of one means and the recognition result of the character information of the other plurality of means, the number of characters recognized in one line or one column of the recognition result is obtained, and When the recognition result of the character information of the means does not match the recognition result of the character information of the one means,
Means for performing a majority vote by excluding the recognition result of the other character recognition means for the row or column from the majority vote reference,
The image information, the character information, means for color-coding the result of the majority decision for each character, displaying the information in a hierarchical display in a superimposed manner, and means for correcting the character pattern. It is a character information recognition device.

According to the present invention, by this means, a majority decision is made only for those having the same character recognition number in row or column units.
It is possible to make a majority decision in a state where the characters to be compared always match, and it is possible to perform more accurate character recognition.

According to a sixth aspect of the present invention, there is provided one means for recognizing character information from image information, a plurality of other means for recognizing character information from the image information, and the character information recognition of the one means. Regarding the majority decision of the result and the character information recognition result of the other plurality of means, one line or one column unit of the character information recognition result of the other plurality of means and the character information recognition result of the one means Means for performing a majority decision by excluding the recognition result of the other character recognition means for the row or column from reference to a majority decision, when the match rate is equal to or less than a predetermined value, the image information, and the character The information is color-coded by character for the result of the majority decision,
A character information recognizing device comprising: means for displaying a hierarchical display in a superimposed manner; and means for enabling the character pattern to be corrected.

According to the present invention, by this means, the recognition result when the character recognition number is shifted near the end of each row or column can be adopted as a majority decision object, and more accurate character recognition can be performed. Becomes

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 shows steps of a character information recognition method according to an embodiment of the present invention. First, original data 1 serving as image information is converted into a standard OCR 1 and other OCRs 2.
Character recognition is performed in 3 and 4. If the image is shown on a paper medium, the image information is read by one scanner and converted into data, and then a plurality of OCRs are read.
It may be recognized by software, or may be read by a plurality of scanners and each of them may be recognized by OCR software.

Next, in each OCR software, the number of recognized characters in the first one line or one line of a character string in each line or column is counted. Next, the standard OCR
The character recognition number of the software is compared with the character recognition number of another OCR. If there is no match, the characters of the entire line are color-coded as non-comparison characters, the characters recognized by the standard OCR are displayed, and the process proceeds to the next line.

If there is an OCR whose character recognition number matches, a majority decision is made for each character in the OCR. If all match, the result of majority decision is standard OC
Characters recognized by the standard OCR are displayed by being color-coded as standard character matching characters when the character matches R, and as non-standard character when the result of majority rule does not match the standard character. After that, the process moves to the next character. If there is no next character, the process proceeds to the comparison of the character recognition numbers of the next line.

FIG. 2 shows another embodiment of the present invention.
The steps of the character information recognition method will be described. Until image information is recognized by each OCR software, it is the same as FIG. Then the first line or 1 of OCR software other than the standard OCR software
The character recognition results of the columns are compared to measure the matching rate. If there is no one having a matching rate of 70% or more, characters in the entire line are color-coded as non-comparison characters, characters recognized by the standard OCR are displayed, and the process proceeds to the next line.

If there is a character recognition result having a matching rate of 70% or more, a majority decision is made for each character. If all match, all characters are matched, if majority result matches standard OCR, standard character match character, and if majority result does not match standard character, non-standard character is color-coded. And standard OC
The character recognized by R is displayed. Thereafter, the process proceeds to the next character, and if there is no next character, the process proceeds to measurement of the matching rate of the character recognition result of the next line.

FIG. 3 schematically shows the structure of the character information recognition apparatus of the present invention. The CPU is a central processing unit and has a read-only memory (ROM) having font data and character code data therein, and a writable memory (RAM) for temporarily storing various programs, data from an input unit, font data, and the like. And each of the following units is connected by a bus line. In addition, the bus line can be connected to an external device (not shown) (LAN, CD-ROM drive, network, etc.).

The display unit is a display such as a color LCD,
The input unit indicates a keyboard, a mouse, and the like. VRAM indicates a display write-only memory, and HD indicates a hard disk serving as an external storage device, in which various font data are recorded and stored. In some cases, new font data is input, registered, and stored from outside. Alternatively, the corrected character data is recorded and saved. The PRT is an output device to another medium such as a printer, and outputs the corrected character data. However, the PRT is not particularly necessary and may be provided as appropriate.

The image information reading section reads image data of an image as a bitmap image, and stores the data in an image information memory. In the present invention, the image information may be obtained by reading the image with the image information reading unit, or the image information originally obtained as a bitmap image may be directly stored in the image information memory. Information may be stored directly in the image information memory.

The character data recognizing section cuts out characters from the image information and recognizes them as character data, and stores the data in a character data memory. Regarding the extraction of characters, a majority decision is made on the recognition result using a plurality of character data recognition units due to differences in the character data recognition units. Determines whether or not to make a majority decision on the line by the coincidence rate measuring unit. If it is determined from the determination result that a majority decision is to be made, a majority decision is made in the majority decision processing unit, and processing for displaying characters is determined from these results.

As the optical character recognition device used in the present invention, in the image information, for characters such as ruby and the like having different fonts and sizes, one line or one line in the font and size of the character recognized as the standard without reading. Those that recognize the characters in the column are preferred. Further, if there is a setting to leave one line between the lines, it is preferable that the setting can be eliminated or one line can be set between all the lines. In addition, it is preferable that the recognition setting that the half-width space and full-width space are skipped and compared, and that the recognition setting can be changed also for the end of line setting (recognition and insertion of a line feed mark) are preferable, but there is no particular limitation. Any commercially available OCR software can be used as appropriate

In the standard optical character recognition apparatus of the present invention, it is preferable to prepare a sample of image information in advance and visually compare the recognition result with the character of the image information to use the one having the highest consistency. However, it may also mean that the OCR software having the function of enabling the setting change regarding the recognition as described above is the most preferable.

In the present invention, the predetermined value of the coincidence rate of the character recognition results in units of rows or columns of the plurality of optical character recognition devices is preferably 50% or more.
Taking into account that the matching rate of CR is about 90%, and considering the possibility of a shift in the number of characters, it is preferably 60 to 80%, most preferably 70% or more.

In FIG. 4, the result of the majority decision in the present invention is classified by character and displayed in a hierarchical display in a superimposed manner, and the character by character is displayed by changing the color gradation for each matching rate. An example of the method will be described. As a result of the majority decision, all matching characters 1 are displayed in the same color tone as the normal display, and the standard O
Standard characters 2 that match CR are displayed in a slightly darker color, and non-standard characters 3 that do not match the standard OCR by majority are displayed in a darker color, allowing comparison with visual image information. Can be performed more easily and efficiently. Although not shown in the figure, the difference is italicized for each character, or the color tone is a color difference other than light and shade, for example, all matching characters are blue like a traffic light, standard characters are yellow,
Non-standard characters may be represented as red. Also,
When displaying the result of majority decision in different colors, the display of characters such as non-comparison lines is displayed and expressed in a completely different way, such as displaying the entire line in italics or underlining. It is preferable.

[0030]

As described above, according to the first aspect of the present invention, the majority decision is made only on the character recognition number in the row or column unit, so that the characters to be compared always match. It is possible to make a majority decision with
There is an operational effect that more accurate character recognition can be performed.

According to the second aspect of the present invention, the recognition result when the character recognition number is shifted near the end of each row or column can be adopted as a majority decision object, and more accurate character recognition can be performed. This has the effect of being able to perform the operation.

Further, according to the third aspect of the present invention, the result of the majority decision is color-coded and superimposed on the character of the original image information and displayed, so that the error of the unrecognized error in the comparison of the character in the image information is obtained. At the time of detection, it is possible to visually recognize the result of the majority decision, and it is possible to detect an error more easily.

According to the fourth aspect of the present invention, since the color gradation is changed and displayed for each coincidence rate, it is possible to visually recognize the result of the coincidence rate, and the error can be further detected. There is an effect of being easily possible.

According to the fifth aspect of the present invention, a majority decision is made only on a character having the same number of recognized characters in units of rows or columns, so that a majority decision can be made in a state where the characters to be compared always match. Thus, there is an operational effect that more accurate character recognition can be performed.

According to the present invention, the recognition result when the number of recognized characters is shifted near the end of each row or column can be adopted as a majority decision object, and more accurate character recognition can be performed. This has the effect of being able to perform the operation.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing steps of a character information recognition method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing steps of a character information recognition method according to another embodiment of the present invention.

FIG. 3 is an explanatory view schematically showing the structure of the character information recognition device of the present invention.

FIG. 4 is a diagram showing a method of displaying a color by color for each character as a result of a majority decision and superimposing them in a hierarchical display according to the present invention, and displaying a color for each character by changing a color gradation for each coincidence rate. It is explanatory drawing which shows an example.

[Explanation of symbols]

 1: All matched characters 2: Standard characters 3: Non-standard characters

Claims (6)

[Claims] 1. A method for recognizing character information from image information, wherein a plurality of optical character recognition devices are used, and when there is a difference in recognized characters among the plurality of optical character recognition devices, a plurality of optical character recognition devices are used. In a character information recognition method for determining a character recognized by a majority decision between character recognition devices, any one of the plurality of optical character recognition devices may be used as a standard optical character recognition device, and other optics for the recognition result may be used. The character recognition result of the other optical character recognition device that recognizes the character recognition result of the character recognition result in a unit of row or column different from each other, and performs majority voting by excluding from the reference of majority voting. Character information recognition method. 2. A method for recognizing character information from image information, wherein a plurality of optical character recognition devices are used, and when there is a difference in recognized characters among the plurality of optical character recognition devices, a plurality of optical character recognition devices are used. In a character information recognition method for determining a character recognized by a majority decision between character recognition devices, any one of the plurality of optical character recognition devices may be used as a standard optical character recognition device, and other optics for the recognition result may be used. The matching rate of the character recognition result of the target character recognition device in units of rows or columns is measured, and when the matching ratio is equal to or less than a predetermined value, the recognition result of the other optical character recognition device for the row or column is determined by majority. A character information recognition method characterized by performing a majority decision excluding from a reference. 3. The character according to claim 1, wherein the image information and the character information are displayed in a hierarchical display by superimposing the result of the majority decision by color for each character. Information recognition method. 4. The character information recognizing method according to claim 3, wherein a color classification for each character as a result of the majority decision is displayed by changing a color gradation for each of the coincidence rates. 5. A means for recognizing character information from image information; a plurality of other means for recognizing character information from the image information; and a recognition result of character information of the one means and the other plurality of means. Regarding the majority decision of the recognition result of the character information of the means, the number of characters recognized in one line or one column of the recognition result is obtained, and the recognition result of the character information of the plurality of other means is the character of the one means. If the information does not match the recognition result, the majority of the row or column is excluded by referring to the recognition result of the other character recognition unit from the majority decision, and the image information and the character information are converted to the result of the majority decision. A character information recognizing device comprising: means for color-coding each character to display in a hierarchical display in a superimposed manner; and means for enabling the character pattern to be corrected. 6. A means for recognizing character information from image information, a plurality of other means for recognizing character information from the image information, a recognition result of character information of the one means and the other plurality of means. Regarding the majority decision of the recognition result of the character information of the means, the match rate of the recognition result of the character information of the plurality of other means and the recognition result of the character information of the one means is obtained in one row or one column unit, When the matching rate is equal to or less than a predetermined value, a unit that performs a majority decision by excluding the recognition result of the other character recognition unit regarding the row or column from the majority decision reference, and the image information and the character information are used as a result of the majority decision. A character information recognizing device comprising: means for color-coding each character to display in a hierarchical display in a superimposed manner; and means for enabling the character pattern to be corrected.