JPH0765112A - Character recognition device - Google Patents

Abstract

PURPOSE:To correct many misread characters together at a time and to improve the efficiency of correcting operation by correcting a recognition result in a specified range according to a specified character kind. CONSTITUTION:A character segmentation means 3 segments character patterns from a document image. A character recognizing means 4 detects character candidates for the input patterns and a display means 9 displays the character candidates. A character kind and a correction range corresponding to the misread characters are specified through an input means 8 and a character kind specification actuating means 11 is actuated with a correction start button. The character kind specification actuating means 11 actuates a correcting means 6 with the specified correction range and specified character kind. The correcting means 6 reregisters only character candidates, corresponding to the specified character kind among the character candidates in a recognition result storage means 5, in the storage means 5 and deletes the character candidates other than the specified character kind from the contents of the storage means 5. Then the candidate character of 1st order in the storage means 5 is displayed by the display means 9.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device capable of easily correcting characters.

[0002]

2. Description of the Related Art As a conventional character recognition apparatus for easily correcting a character, there is, for example, Japanese Patent Laid-Open No. 4-19.
It is shown in Japanese Patent Publication No. 9483. The configuration of a conventional character recognition device will be described with reference to FIG. Figure 24
FIG. 7 is a block diagram showing a configuration of a conventional character recognition device.

In FIG. 24, 1 is an image inputting means for inputting a document image, 2 is an input image storing means for storing the input document image, 3 is a character cutting out means for cutting out a character pattern from the document image, and 4 is a character cutting out means. Character recognition means for recognizing the character pattern obtained in 3.
Is a recognition result storing means for storing the candidate character obtained as a result of character recognition and the similarity, 6 is a correcting means for correcting the recognition result,
Reference numeral 7 is a search means for searching a combination of specific candidate characters from the recognition result.

Reference numeral 8 is an input means composed of a mouse, a keyboard, etc., 9 is a CRT for displaying the recognition result, a display means such as a liquid crystal, and 10 is a control means such as a CPU for controlling the above means.

Next, the operation of the above-mentioned conventional character recognition apparatus will be described with reference to FIGS. 25, 26, 27 and 28. FIG. 25 is a diagram showing a document image "a certain condition is satisfied between AB in this problem and the next problem" input by the conventional character recognition device.

FIG. 26 is a diagram showing the recognition result of the conventional character recognition device, and 20, 21, and 22 are diagrams showing the same character as the recognition result. Furthermore, FIG.
7 is a diagram showing candidate characters in the recognition result storage means of the conventional character recognition device, and 23, 24 and 25 are recognition results 2
It is a figure which shows the candidate character of 0-22. Furthermore, FIG.
FIG. 7 is a diagram showing a recognition result after correction of a conventional character recognition device.

FIG. 25 shows an example of a document image input by the image input means 1. This document image is stored in the input image storage means 2. The character cutout unit 3 cuts out each character pattern from the document image, and the character recognition unit 4 recognizes the cutout character pattern and outputs the candidate character and the degree of similarity. The output candidate characters and the similarity are stored in the recognition result storage means 5.

FIG. 26 shows an example of the recognition result of the document image shown in FIG. 25, showing a display example by the display means 9. 20 to 22 are recognition results, 20 is a character “question”, 21 is a character “between”, and 22 is a recognition result of the character “question”. In the recognition results 20 and 22, the character "question" is misread as "between", and the recognition result 21 is correctly recognized.

FIG. 27 shows candidate characters in the recognition result storage means 5, 23 is a candidate character corresponding to the recognition result 20, 24 is a candidate character corresponding to the recognition result 21, and 2 is a candidate character.
Reference numeral 5 is a candidate character corresponding to the recognition result 22.

The recognition result 20 displayed by the display means 9
Is incorrect, the operator uses the input means 8 such as a mouse.
Then, “question” is selected from the candidate characters shown in 23 of FIG. 27, and the misread character “between” is corrected to “question”. Next, corresponding to the corrected character, the searching means 7 searches the uncorrected character for a character that includes the same candidate character as the corrected character 20, and the correcting means 6 results in a search. When a character pattern including the same candidate character exists, the recognition result corresponding to the character pattern is automatically corrected.

For example, in the example shown in FIG. 26, the recognition result "between" 22 having the same candidate characters as the candidate characters of the recognition result 20 corrected by the operator is corrected to "question". The recognition result 21 is not corrected because the candidate character is different. The above process is repeated until there are no characters including the same candidate. By this processing, the recognition result is corrected correctly as shown in FIG.

[0012]

Since the conventional character recognition apparatus as described above is configured to collectively correct misread characters including the same candidate character, misreading including the same candidate character may occur. When a large number are generated, the number of corrections can be reduced, but the probability of misreading including the same candidate character is low, and correction takes time. That is, when recognizing handwritten characters, even if the same writer writes the same character multiple times, the shapes will be slightly different, and the candidate characters will differ accordingly. There was a problem that the efficiency of the correction was not improved.

Similarly, when recognizing a print character, the state (quality) of the document image, for example, character deviation /
There is a problem that the candidate characters of the same character are not all the same due to scratches or the like, and the efficiency of character correction is not improved.

Further, there is a problem that the efficiency is not improved in the conventional method when the same misread is less.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to obtain a character recognition device capable of easily and efficiently correcting many misread characters that do not include the same candidate character. To aim.

[0016]

A character recognition apparatus according to claim 1 of the present invention comprises the following means. [1] Image storage means for inputting and storing a document image. [2] Character cutting means for cutting a character pattern from the stored document image. [3] Character recognition means for recognizing the character pattern and generating candidate characters and similarity as a recognition result. [4] Recognition result storage means for storing the candidate character and the similarity. [5] Display means for displaying the recognition result and the correction result. [6] Character type designation starting means for designating a character type to be read and a correction target area. [7] Correction means that is activated by the character type designation activation means and that corrects candidate characters in the correction target area based on the designated character type.

A character recognition device according to claim 2 of the present invention comprises the following means, and when there is no character of the designated character type among the candidate characters, the character type designation activation means designates the character type. The character recognition means is activated, and the character recognition means again recognizes only the designated character type as a recognition target. [1] Image storage means for inputting and storing a document image. [2] Character cutting means for cutting a character pattern from the stored document image. [3] Character recognition means for recognizing the character pattern and generating candidate characters and similarity as a recognition result. [4] Recognition result storage means for storing the candidate character and the similarity. [5] Display means for displaying the recognition result and the correction result. [6] Character type designation starting means for designating a character type to be read and a correction target area. [7] Correction means that is activated by the character type designation activation means and that corrects candidate characters in the correction target area based on the designated character type.

A character recognition device according to a third aspect of the present invention comprises the following means, and the character type designation activation means activates the character recognition means by designating a character type and a correction target area, and the character recognition means. Is for recognizing a character pattern again by a character recognition method corresponding to the designated character type. [1] Image storage means for inputting and storing a document image. [2] Character cutting means for cutting a character pattern from the stored document image. [3] Character recognition means for recognizing the character pattern and generating candidate characters and similarity as a recognition result. [4] Recognition result storage means for storing the candidate character and the similarity. [5] Display means for displaying the recognition result and the correction result. [6] Character type designation starting means for designating a character type to be read and a correction target area. [7] Correction means that is activated by the character type designation activation means and that corrects candidate characters in the correction target area based on the designated character type.

A character recognition apparatus according to a fourth aspect of the present invention comprises the following means, and the character type designation activation means activates the character cutout means and the character recognition means by designating the character type and the correction target area, The character cutout means recuts out a character pattern by a character cutout method corresponding to the designated character type, and the character recognition means rerecognizes a character pattern by a character recognition method corresponding to the designated character type. [1] Image storage means for inputting and storing a document image. [2] Character cutting means for cutting a character pattern from the stored document image. [3] Character recognition means for recognizing the character pattern and generating candidate characters and similarity as a recognition result. [4] Recognition result storage means for storing the candidate character and the similarity. [5] Display means for displaying the recognition result and the correction result. [6] Character type designation starting means for designating a character type to be read and a correction target area. [7] Correction means that is activated by the character type designation activation means and that corrects candidate characters in the correction target area based on the designated character type.

[0020]

In the character recognition apparatus according to the first aspect of the present invention, the image storage means inputs and stores the document image. Further, the character cutout unit cuts out a character pattern from the stored document image, the character recognition unit recognizes the character pattern, and a candidate character and a similarity are generated as a recognition result. Further, the recognition result storage means stores the candidate character and the degree of similarity, and the display means displays the recognition result and the correction result. Then, a character type and a correction target area to be read are designated by the character type designation activation means, and the candidate character in the correction target area is designated based on the designated character type by the correction means activated by the character type designation activation means. Will be fixed.

In the character recognizing apparatus according to the second aspect of the present invention, the image storing means inputs and stores the document image. Also, by the character cutting means,
A character pattern is cut out from the stored document image, the character recognition unit recognizes the character pattern, and a candidate character and a similarity are generated as a recognition result. Further, the recognition result storage means stores the candidate character and the degree of similarity, and the display means displays the recognition result and the correction result. Further, by the character type designation activation means, the character type to be read and the correction target area are designated,
The correction means activated by the character type designation activation means corrects the candidate character in the correction target area based on the designated character type. When there is no character of the designated character type in the candidate characters, the character type designation activation means designates the character type and activates the character recognition means, and the character recognition means performs the designation. The recognition is performed again with only the character type as the recognition target.

In the character recognition device according to the third aspect of the present invention, the document image is input and stored by the image storage means. Also, by the character cutting means,
A character pattern is cut out from the stored document image, the character recognition unit recognizes the character pattern, and a candidate character and a similarity are generated as a recognition result. Further, the recognition result storage means stores the candidate character and the degree of similarity, and the display means displays the recognition result and the correction result. Further, by the character type designation activation means, the character type to be read and the correction target area are designated,
The correction means activated by the character type designation activation means corrects the candidate character in the correction target area based on the designated character type. Then, the character type designation activation means designates the character type and the correction target area to activate the character recognition means, and the character recognition means reactivates a character pattern by a character recognition method corresponding to the designated character type. Be recognized.

In the character recognition apparatus according to the fourth aspect of the present invention, the document image is input and stored by the image storage means. Also, by the character cutting means,
A character pattern is cut out from the stored document image, the character recognition unit recognizes the character pattern, and a candidate character and a similarity are generated as a recognition result. Further, the recognition result storage means stores the candidate character and the degree of similarity, and the display means displays the recognition result and the correction result. Further, by the character type designation activation means, the character type to be read and the correction target area are designated,
The correction means activated by the character type designation activation means corrects the candidate character in the correction target area based on the designated character type. Then, by the character type designation activation means, the character type and the correction target area are designated, the character cutout means and the character recognition means are activated, and by the character cutout means, in a character cutout method corresponding to the specified charactertype. The character pattern is cut out again, and the character recognition means re-recognizes the character pattern by a character recognition method corresponding to the designated character type.

[0024]

【Example】

Example 1. The configuration of the first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing the configuration of the first embodiment of the present invention, and image input means 1-correction means 6 and input means 8-control means 10 are the same as those of the conventional apparatus described above. In each figure, the same reference numerals indicate the same or corresponding parts.

In FIG. 1, reference numeral 11 is a character type designation activation means for activating the other means by designating the character type to be read and the correction range.

By the way, the image storing means according to claim 1 of the present invention comprises the image inputting means 1 and the input image storing means 2 in the embodiment 1, and the character cutting means according to claim 1 of the present invention is In the first embodiment, it corresponds to the character cut-out means 3, the character recognition means according to claim 1 of the present invention corresponds to the character recognition means 4 in the first embodiment, and the recognition result storage according to claim 1 of the present invention is stored. The means corresponds to the recognition result storage means 5 in the first embodiment, the display means according to claim 1 of the present invention corresponds to the display means 9 in the first embodiment, and the character according to claim 1 of the present invention. The seed designation starting means is
In the first embodiment, it corresponds to the character type designation activation means 11, and the correction means according to claim 1 of the present invention corresponds to the correction means 6 in the first embodiment.

Next, the operation of the above-described first embodiment will be described with reference to FIGS. 2 and 3 to 12. Figure 2
3 is a flowchart showing the operation of the first embodiment of the present invention. 3 is a diagram showing an input document image in the first embodiment of the present invention.

FIGS. 4 to 6 show the first embodiment of the present invention.
It is a figure for demonstrating operation | movement of the character cutting-out means.
FIG. 4 is a diagram showing a projection histogram obtained by counting the number of black dots in the input document image shown in FIG. 3 in the character string direction and the vertical direction. Further, FIG. 5 is a diagram showing a basic rectangle obtained from the projection histogram. Further, FIG. 6 shows a character pattern obtained by using the basic rectangle, and 40 is a cut-out character pattern.

FIG. 7 is a diagram showing a display example of the display means 9 for displaying the recognition result according to the first embodiment of the present invention, and 41 shows the recognition result of the document image of FIG. FIG. 8 is a diagram for explaining the character type selection by the operator.
Reference numeral 2 is a character type selection button, 43 is a correction start button, and 44 is a character type selection menu. Further, FIG. 9 is a diagram showing an example in which the operator selects a correction range by the input means 8 according to the first embodiment of the present invention, where 45 is the correction range instructed by the operator, and 46 is the cut-out character pattern "2". 40
The recognition result of is shown.

FIG. 10 is a diagram showing the contents of the recognition result storage means 5 before correction and activation according to the first embodiment of the present invention, 47-.
Reference numeral 51 is a candidate character corresponding to the character pattern in the correction range 45, 52 is the first candidate character of the first character in the correction range 45, and 53 is the second candidate character. In addition, FIG.
It is a figure which shows the recognition result in the recognition result storage means 5 after correction of Example 1 of this invention, 54 shows the candidate character after correction of the candidate character 47. Further, FIG. 12 is a diagram showing a display example by the display means 9 for displaying the corrected recognition result according to the first embodiment of the present invention.

In steps 30 to 31, the document image of FIG. 3 is input by the image input means 1 and stored in the input image storage means 2. The character cutout unit 3 cuts out a character pattern from a document image by a conventional method disclosed in, for example, Japanese Patent Publication No. 4-67674.

Specifically, the number of black dots is counted in the character string direction and the vertical direction from the document image of FIG. 3 to obtain a projection histogram as shown in FIG. Next, when the histogram exceeds a predetermined threshold value, the information is converted into a binary image of "1", and otherwise "0", and the area where the image information of the value "1" is continuous is set as the width of the basic rectangle. The basic rectangle shown in is detected. Furthermore, when scanning from the upper end to the lower end of the document image, the coordinates when the pixel that shows the character portion is first encountered are the upper end coordinates of the basic rectangle, and the coordinates when the pixel is last encountered are the lower end of the basic rectangle. To detect as. This basic rectangle is the smallest unit that constitutes a character. A character evaluation value that quantitatively indicates whether or not the basic rectangle and a combined rectangle obtained by combining a plurality of basic rectangles are established as a single character, and the combination of rectangles having the maximum character evaluation value is shown in FIG. Cut out as a character pattern as shown.

In step 32, the character recognition means 4
Recognizes a character pattern by the conventional method disclosed in, for example, Japanese Patent Publication No. 2-6113.

Specifically, the character pattern is thinned, characteristic points such as end points, branch points, and inflection points are extracted, a line segment connecting these is obtained, and the line points are connected to the characteristic points other than the end points. The straight line components are extracted by combining pairs of line segments that have the same direction, and the similarity between the middle point pattern by length of the straight line component and the middle point pattern by length of the straight line component of the reference character pattern is used. Detects the input pattern candidate characters,
The obtained candidate character and the corresponding similarity are output. The candidate characters are detected by sorting the similarity with each reference pattern in descending order of similarity and extracting from the upper order by a predetermined maximum number of candidate characters. In this Example 1,
The following description will be made assuming that the maximum number of candidate characters is 4.

In steps 33 to 34, the first-ranked candidate character thus obtained is displayed as a recognition result by the display means 9 to the operator as shown in FIG. 7, for example. Next, the operator designates the character type corresponding to the misread character, here a numeral, by the input means 8. In specifying the character type, for example, as shown in FIG. 8, the character type selection button 42 is selected to display the character type selection menu 44, and the character type desired to be specified is selected from the character type selection menu 44.

In steps 35 to 37, the operator uses the input means 8 as shown in FIG.
Is designated, and further, the correction activation button 43 activates the character type designation activation means 11 to correct the candidate character. The character type designation activation means 11 activates the correction means 6 with the designated correction range 45 and the designated character type, and the correction means 6 makes the candidate characters in the recognition result storage means 5 candidates corresponding to the designated character type, that is, a numeral. Only the characters are registered again in the recognition result storage means 5, and the candidate characters other than the designated character type are deleted from the recognition result storage means 5.

FIG. 10 shows candidate characters of the characters within the correction range 45. In FIG. 10, the character pattern "2" 4
The candidate characters of 0 are the first candidate character “Z” 52 and the second candidate character “2” 53, but the first candidate character “Z” 52 is an alphabetic character and the second candidate character “2” 53 is a numeral. Only the first rank candidate character “Z” 52 is not the designated character type. Therefore, the first-ranked candidate character “Z” 52 is the recognition result 4 in the recognition result storage means 5.
7, the candidate character “2” 53 is newly added as the first candidate character. In this way, the candidate characters in the recognition result storage means 5 are sequentially corrected until the correction of the designated correction range is completed, and finally are corrected as shown in FIG.

In steps 38 to 39, thereafter, the first rank candidate character in the recognition result storage means 5 is displayed by the display means 9, as shown in FIG. As shown in FIG. 12, the recognition result is correctly corrected by specifying the character type. If there is another area to be modified, steps 34 to 38 are repeated until the modification is completed.

As described above, by correcting the candidate characters within the specified range corresponding to the specified character type, the operator can easily correct many misread characters at one time only by making a simple specification. it can.

The first embodiment of the present invention, as described above,
An image input unit 1 for inputting a document image, an input image storage unit 2 for storing the document image input by the image input unit 1, and a character cutout unit 3 for cutting out a character pattern from the document image in the input image storage unit 2. , Recognizes the character pattern obtained by the character cutting means 3,
Character recognition means 4 for outputting the candidate character and the similarity, recognition result storage means 5 for storing the similarity and the candidate character output from the character recognition means 4, and correction means 6 for correcting the recognition result.
A character type designation activation means 11 for designating a character type to be read and a correction target area, an input means 8 for accepting an input of the character type and a correction target area from an operator, and a display means 9 for displaying a recognition result. Control means 1 for controlling each of the above means
0, the character type designation activation unit 11 activates the correction unit 6 by designating the character type and the area to be corrected, and the correction unit 6 sets the candidate character in the recognition result storage unit 5 to the designated character type. It will be corrected accordingly.

In the character correction method according to the first embodiment, the character type designation activation means 11 activates the correction means 6 by using the correction range and the character type information inputted by the operator from the input means 8, and the correction means 6 designates. Since the recognition result within the range is corrected according to the specified character type, the correct characters are included in the upper candidate characters by excluding the candidate characters other than the specified character type, making it easy to correct a large number of misread characters collectively. can do.

That is, the first embodiment aims to improve the efficiency of the correction work of the misread character in the character recognition device. Therefore, when correcting misread characters,
The correction unit 6 corrects the candidate character according to the designated character type. Further, it is provided with a character type designation activation means 11 for designating a character type and activating the correction means 6. As a result, when correcting the recognition result, you can automatically correct many misread characters at once by specifying the character type and the correction range, which can greatly improve the efficiency of the correction work. it can.

Example 2. In the second embodiment of the present invention, when the designated character type does not exist in the candidate characters, the target character type is limited to the designated character type and recognition is performed again. The configuration of the second embodiment is similar to that of the first embodiment described above.

Regarding the operation of the second embodiment, referring to FIGS.
This will be described with reference to FIG. FIG. 13 is a flow chart showing the operation of the second embodiment of the present invention. Also,
FIG. 14 is a diagram showing candidate characters in the recognition result storage means 5 according to the second embodiment of the present invention. Further, FIG. 15 shows candidate characters in the recognition result storage means 5 after correction according to the second embodiment of the present invention. It is to be noted that, except for FIGS.

Steps 60 to 65 and step 6 in FIG.
The operations of 8 to 71 are the same as the operations of steps 30 to 39 in FIG. Similar to the first embodiment, the document image shown in FIG. 3 is input (step 60), characters are cut out (step 61), and each character pattern is recognized (step 62). The steps 61 and 62 may use the same method as that of the first embodiment, and the description thereof will be omitted here. Next, the recognition result obtained by recognizing the character pattern is displayed as shown in FIG. 7 (step 63). Here, the operator designates the character type as a numeral (step 64) because the misread character is a numeral in FIG. 7, and designates the correction range 45 as shown in FIG. 9 (step 65).

Next, when it is detected that the operator has selected the correction start button 43, it is judged whether or not the designated character type exists in the candidate characters in the recognition result in the recognition result storage means 5 (step 66), and the designation is made. If the character type exists, the correction means 6 is activated by the character type designation activation means 11 in the same manner as in the first embodiment to correct the recognition result in the recognition result storage means 5 (step 68).

Here, as shown in FIG. 14, the correction range 4
If the designated character type, that is, a number does not exist, such as a candidate character obtained by recognizing the second and fifth character patterns of 5, the recognition target character type is set to the designated character type (number), and the character type The designated activation means 11 activates the character recognition means 4 again to recognize the corresponding character pattern (step 67).

Since the designated character types are limited to numbers, the candidate character in the corrected recognition result storage means 5 is the category "7" for the second character in the correction range 45 as shown in FIG. It is the first candidate character, and the category “9” is the first candidate character for the fifth character as well, and both are corrected correctly. This is because in the recognition in step 62, the category “7” has the fifth highest similarity to the pattern of the second character of the correction range 45, and the category “7” has the category “5” to the pattern of the fifth character. 9 "is 5
This is because it has the second highest similarity, and since the maximum number of candidate characters is 4 here, both categories were not included in the candidate characters of the first recognition result.
It has been correctly recognized by step 67. In this way, steps 66, 67 and 68 are repeatedly executed until the correction of the designated area is completed (step 6
9) Then, the corrected recognition result is displayed as shown in FIG. 12 (step 70). The corrected result is correctly corrected as shown in FIG.

As described above, when the designated character type is not included in the candidate characters, only the designated character type is recognized again as the recognition target character type, and the character pattern in the designated area is recognized again. In the recognition process, the correct correction can be made even when the correct category is not included in the candidate characters.

That is, in the second embodiment, when there is no character of the designated character type among the candidate characters in the recognition result storage means 5, the character type designation activation means 11 sets the recognition target character type to the designated character type. The character recognition means 4 is activated for a limited time, and the character recognition means 4 is activated.
Recognizes only the specified character type as a recognition target and performs recognition again. Therefore, misread characters that do not include the correct category in the candidate characters can be corrected correctly.

Example 3. In the third embodiment of the present invention, the character recognizing means 4 recognizes a character pattern within the correction range again from the character type and the correction range specified at the time of correction by a recognition method corresponding to the specified character type. In addition, this Example 3
The configuration is similar to that of the first embodiment described above.

The operation of the third embodiment will be described below with reference to FIG.
Will be described with reference to. FIG. 16 is a flow chart showing the operation of the third embodiment of the present invention. Note that, except for FIG. 16, it is the same as the first embodiment.

Steps 80 to 85 and steps 87 and 88 shown in FIG. 16 are steps 30 to 35 and 3 shown in FIG.
The same as 8 and 39. First, as in the first embodiment, a document image is input (step 80), and a character pattern is cut out from the document image (step 81).

Next, the character recognizing means 4 recognizes by the recognition method shown in the first embodiment (step 82) and outputs the candidate character and the similarity. The output candidate character and the similarity are stored in the recognition result storage means 5, and the first candidate character is displayed on the display means 9 as the recognition result (step 83).

Here, it is assumed that the recognition result obtained by inputting the document image shown in FIG. 3 is as shown in FIG. Therefore, the operator selects a numeral as the target character type (step 84), inputs the correction range 45 as shown in FIG. 9 (step 85), and presses the correction start button 43 in the same procedure as in the first embodiment. Select and start correction start. Accordingly, when the character type designation activation means 11 activates the character recognition means 4 with the designated character type, the character recognition means 4 causes the character pattern within the correction range 45 designated by the recognition method corresponding to the designated character type. Is recognized (step 86).

In the third embodiment, since a numeral is designated as the character type, a method corresponding to the numeral, for example, Japanese Patent Publication No. 2-
As disclosed in Japanese Patent No. 43221, a method of recognizing a character pattern by tracing the contour of the character pattern to obtain end points, convex line segments, and concave line segments is used. When the recognition target character type is a numeral, the shape of the pattern is relatively simple and the number of straight line components is small. Therefore, the character recognition method using the straight line component of the pattern shown in the first embodiment is more effective than the character recognition method. The method described in Japanese Examined Patent Publication No. 2-43221 using features is
The features of the pattern can be more easily grasped, and better recognition accuracy can be obtained. The candidate character which is the result of recognition here and the degree of similarity are stored in the corresponding portion of the recognition result storage means 5, and the first-ranked candidate character in the recognition result storage means 5 is displayed on the display means 9 as shown in FIG. (Step 87).

As described above, by selecting and re-recognizing a method for which higher recognition accuracy is desired according to the designated character type, it is possible to reduce the intervention of the operator and significantly improve the efficiency of the correction work. .

That is, in the third embodiment, the character type designation activation means 11 limits the target character type to the designated character type, activates the character recognition means 4, and the character recognition means 4 causes the characters corresponding to the designated character type. Since the recognition method is applied, the character recognition rate at the time of correction can be improved, and the efficiency of the correction work required for correction can be improved.

Example 4. In the fourth embodiment of the present invention, the character cutting means 3 is selected from the character type and the correction range designated at the time of correction.
The character recognition means 4 cuts out and recognizes the image within the correction range by a method corresponding to the designated character type. The configuration of the fourth embodiment is the same as that of the first embodiment described above.
The configuration is the same.

The operation of the fourth embodiment will be described below with reference to FIG.
7, and will be described with reference to FIGS. FIG. 17
9 is a flow chart showing the operation of Embodiment 4 of the present invention. 18 is a diagram showing an input document image in the fourth embodiment of the present invention. Further, FIG. 19 is a diagram showing a basic rectangle obtained from a projection histogram in which the number of black dots of the document image of FIG. 18 is counted in the character string direction and the vertical direction. FIG. 20 is a diagram showing a character pattern cut out by the character cutout unit 3 using the basic rectangle shown in FIG. FIG. 21 is a diagram showing a display example of the recognition result before correction. FIG. 22 is a diagram showing a correction range designated by the operator. FIG. 23 is a diagram showing a display example of the recognition result after correction.

Steps 90 to 95, 98 shown in FIG.
Since step 99 is similar to steps 30 to 35, 38, and 39 shown in FIG. 2, description thereof will be omitted. Similar to the first embodiment described above, first, the document image shown in FIG. 18 is input (step 90), and the character pattern is cut out (step 9).
1). In the character segmentation in step 91, the basic rectangle shown in FIG. 19 is obtained in the same procedure as in the first embodiment, and individual character patterns are segmented using this.

Here, as a result of the character extraction, the patterns shown in FIG. 20 are each extracted as a character pattern. In step 91, individual patterns are cut out by a cutout method for Japanese sentences composed of double-byte characters such as kanji. Therefore, in FIG. 20, the character string “positiv
The pattern "e" is actually two half-width characters such as "po", "si", "ti", and "ve", and is incorrectly cut out as one full-width character.

Next, character recognition is performed on each of the cut out patterns in the same manner as in the first embodiment (step 92).
The recognition result is displayed as 1 (step 93). Figure 2
Since the alphabetic character portion is not correctly recognized in No. 1, the operator selects the alphabetic character as the target character type this time by the same procedure as in the first embodiment (step 94) and sets the correction range 55 as shown in FIG. Input (step 95). In FIG. 22, 55 is a correction range designated by the operator,
Reference numeral 56 is a cutout / recognition activation button for instructing to perform cutout and recognition again.

After that, when the operator selects the cut-out / recognition start button 56, the character-type designation start-up means 1 is selected accordingly.
First, the character type and the correction target area are designated to start the character cutting means 3. The character clipping means 3 performs character clipping (step 96) on the character string image within the designated correction range 55 by a method corresponding to the designated character type.

Here, since the designated character type is an alphabetic character, a method corresponding to the alphabetic character, for example, a method of cutting out each basic rectangle shown in FIG. 19 as one character is used. Here, since the alphabetic character pattern does not have a separating character composed of a plurality of basic rectangles such as katakana “H”, this method that does not consider a combined rectangle can cut out more correctly.

Next, the character type designation activation means 11 activates the character recognition means 4 with the designated character type. The character recognition means 4 recognizes the newly cut out character pattern by the character recognition method corresponding to the designated character type as in the third embodiment (step 97), and resembles the candidate character as the recognition result. The degree is stored in the corresponding portion of the recognition result storage unit 5. After that, the first-ranked candidate character in the recognition result storage means 5 is displayed on the display means 9 as shown in FIG. 23 (step 98).
As shown in FIG. 23, the misread character string "positive"
Is corrected correctly.

As described above, the character pattern is cut out and recognized again for the image within the specified range by using the cutout method and the recognition method corresponding to the specified character type, so that the recognition result can be obtained with high accuracy. Since it can be obtained, it becomes possible to easily make corrections including cutting out errors,
The efficiency of the correction work can be significantly increased.

In the first to fourth embodiments described above, the character type is always specified before the area to be corrected is specified. In the modification, by using the held designated character type, the step of inputting the character type may be omitted.
Further, although the correction target area is input after the character type is input, the character type input may be performed after the correction target area is input.

[0069]

As described above, the character recognition device according to the first aspect of the present invention includes the image storing means for inputting and storing the document image, and the character cutting means for cutting out the character pattern from the stored document image. And a character recognition unit that recognizes the character pattern and generates a candidate character and similarity as a recognition result, a recognition result storage unit that stores the candidate character and similarity, and a display that displays the recognition result and the correction result. Means, a character type designation activation means for designating a character type to be read and a correction target area, and a correction means which is activated by the character type specification activation means and corrects a candidate character in the correction target area based on the designated character type. Since it is equipped with and, many misread characters can be corrected automatically at once and the efficiency of the correction work can be greatly improved. That.

As described above, the character recognition device according to the second aspect of the present invention includes the image storing means for inputting and storing the document image, and the character cutting means for cutting out the character pattern from the stored document image. Character recognition means for recognizing the character pattern and generating a candidate character and similarity as a recognition result; recognition result storage means for storing the candidate character and similarity; and display means for displaying the recognition result and the correction result. A character type designation activation means for designating a character type to be read and a correction target area, and a correction means activated by the character type specification activation means for correcting a candidate character in the correction target area based on the designated character type. When there is no character of the designated character type in the candidate characters, the character type designation activation means activates the character recognition means by designating the character type, Since the recognition means re-recognizes only the specified character type as the recognition target, it is possible to automatically correct a lot of misread characters at once, and it is possible to greatly improve the efficiency of the correction work. Produce an effect.

As described above, the character recognition apparatus according to the third aspect of the present invention includes the image storing means for inputting and storing the document image, and the character cutting means for cutting out the character pattern from the stored document image. Character recognition means for recognizing the character pattern and generating a candidate character and similarity as a recognition result; recognition result storage means for storing the candidate character and similarity; and display means for displaying the recognition result and the correction result. A character type designation activation means for designating a character type to be read and a correction target area, and a correction means activated by the character type specification activation means for correcting a candidate character in the correction target area based on the designated character type. The character type designation activation means activates the character recognition means by designating the character type and the correction target area, and the character recognition means corresponds to the designated character type. Because reaffirmed character pattern in character recognition system, so many misread characters can be automatically corrected collectively once, an effect that the efficiency of the correction operation can be greatly improved.

As described above, the character recognition apparatus according to the fourth aspect of the present invention includes an image storing means for inputting and storing a document image, and a character cutting means for cutting out a character pattern from the stored document image, Character recognition means for recognizing the character pattern and generating a candidate character and similarity as a recognition result; recognition result storage means for storing the candidate character and similarity; and display means for displaying the recognition result and the correction result. A character type designation activation means for designating a character type to be read and a correction target area, and a correction means activated by the character type specification activation means for correcting a candidate character in the correction target area based on the designated character type. The character type designating and activating means designates the character type and the correction target area and activates the character slicing means and the character recognizing means, and the character slicing means. The character pattern is re-cut out by the character cutting method corresponding to the specified character type, and the character recognition means re-recognizes the character pattern by the character recognition method corresponding to the specified character type. As a result, it is possible to make automatic correction, and it is possible to significantly improve the efficiency of the correction work. Further, even for misread characters caused by an error in character cutout, by inputting a corresponding character type, the character cutout / recognition method can be appropriately switched, which enables efficient correction work. Produce an effect.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the first embodiment of the present invention.

FIG. 3 is a diagram showing a document image according to the first embodiment of the present invention.

FIG. 4 is a diagram showing a projection histogram according to the first embodiment of the present invention.

FIG. 5 is a diagram showing a basic rectangle according to the first embodiment of the present invention.

FIG. 6 is a diagram showing a result of extracting characters according to the first embodiment of the present invention.

FIG. 7 is a diagram showing a display example of a recognition result according to the first embodiment of the present invention.

FIG. 8 is a diagram showing character type selection according to the first embodiment of the present invention.

FIG. 9 is a diagram showing correction range selection according to the first embodiment of the present invention.

FIG. 10 is a diagram showing stored contents of a recognition result storage means according to the first embodiment of the present invention.

FIG. 11 is a diagram showing an operation of the correcting means according to the first embodiment of the present invention.

FIG. 12 is a diagram showing a display example after correction of a recognition result according to the first embodiment of the present invention.

FIG. 13 is a flowchart showing the operation of the second embodiment of the present invention.

FIG. 14 is a diagram showing the operation of the correction means according to the second embodiment of the present invention.

FIG. 15 is a diagram showing stored contents after correction of a recognition result according to the second embodiment of the present invention.

FIG. 16 is a flowchart showing the operation of the third embodiment of the present invention.

FIG. 17 is a flowchart showing the operation of the fourth embodiment of the present invention.

FIG. 18 is a diagram showing a document image according to the fourth embodiment of the present invention.

FIG. 19 is a diagram for explaining character cutout according to the fourth embodiment of the present invention.

FIG. 20 is a diagram showing a character segmentation result according to the fourth embodiment of the present invention.

FIG. 21 is a diagram showing a display example of a recognition result according to the fourth embodiment of the present invention.

FIG. 22 is a diagram showing correction range selection according to the fourth embodiment of the present invention.

FIG. 23 is a diagram showing a display example after correction of the recognition result according to the fourth embodiment of the present invention.

FIG. 24 is a block diagram showing a configuration of a conventional character recognition device.

FIG. 25 is a diagram showing a document image in a conventional character recognition device.

FIG. 26 is a diagram showing a recognition result of a conventional character recognition device.

FIG. 27 is a diagram showing contents in a recognition result storage unit of a conventional character recognition device.

FIG. 28 is a diagram showing a display example of contents in the recognition result storage means after correction of the conventional character recognition device.

[Explanation of symbols]

 1 image input means 2 input image storage means 3 character cut-out means 4 character recognition means 5 recognition result storage means 6 correction means 8 input means 9 display means 10 control means 11 character type designation activation means

Claims (4)

[Claims] 1. An image storage unit for inputting and storing a document image, a character segmenting unit for segmenting a character pattern from the stored document image, recognition of the character pattern, and generation of candidate characters and similarity as a recognition result. Character recognition means, recognition result storage means for storing the candidate characters and the degree of similarity, display means for displaying the recognition results and correction results, character type designation activation means for specifying the character type to be read and the correction target area, and the character type A character recognition device comprising: a correction unit that is started by a specified starting unit and corrects a candidate character in the correction target area based on the specified character type. 2. An image storage unit for inputting and storing a document image, a character segmenting unit for segmenting a character pattern from the stored document image, recognition of the character pattern, and generation of candidate characters and similarity as a recognition result. Character recognition means, recognition result storage means for storing the candidate characters and the degree of similarity, display means for displaying recognition results and correction results, character type designation activation means for specifying a character type to be read and a correction target area, and the character type The correction means is activated by the designated activation means and corrects the candidate character in the correction target area based on the designated character type. When there is no character of the designated character type in the candidate character, the character type designation is performed. The activating means activates the character recognizing means by designating the character type, and the character recognizing means again recognizes only the designated character type as a recognition target. Character recognition device. 3. An image storage unit for inputting and storing a document image, a character segmenting unit for segmenting a character pattern from the stored document image, recognition of the character pattern, and generation of candidate characters and similarity as a recognition result. Character recognition means, recognition result storage means for storing the candidate characters and the degree of similarity, display means for displaying the recognition results and correction results, character type designation activation means for specifying the character type to be read and the correction target area, and the character type It is provided with a correction means which is activated by the designated activation means and corrects the candidate character in the correction target area based on the designated character type, and the character type designation activation means designates the character type and the correction target area to generate the character. A character recognition device characterized in that a recognition means is activated, and the character recognition means recognizes a character pattern again by a character recognition method corresponding to the designated character type. 4. An image storage unit for inputting and storing a document image, a character segmenting unit for segmenting a character pattern from the stored document image, recognition of the character pattern, and generation of candidate characters and similarity as a recognition result. Character recognition means, recognition result storage means for storing the candidate characters and the degree of similarity, display means for displaying the recognition results and correction results, character type designation activation means for specifying the character type to be read and the correction target area, and the character type The correction means is activated by the designated activation means, and corrects the candidate character in the correction target area based on the designated character type. The character type designation activation means specifies the character type and the correction target area to specify the character. The slicing means and the character recognizing means are activated, and the character slicing means re-strips a character pattern by a character slicing method corresponding to the designated character type. The character recognition device is characterized in that the character recognition means re-recognizes a character pattern by a character recognition method corresponding to the designated character type.