JP2003203204A - Character recognition method and character recognition device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a character recognition method and device capable of performing a character recognition having high precision and high reliability by preventing the recognition output of a recognition result for other than characters. <P>SOLUTION: An area on a document is divided to a character area in which characters are described, and two or more areas by attribute corresponding to other descriptions, character data are divided to each line unit within the character area (step S1101), and the average confidence degree of recognition result of characters is calculated in each line unit in the character area (steps S1102 and S1103). Thereafter, a character data first candidate with low confidence degree is substituted by a space (step S1115) based on the overlapping with a surface area to the line (step S1104), the overlapping with a drawing (step S1105), the ratio of the number of high-confidence degree characters to the number of characters in the line (step S1110) and the like to prevent the output of the recognition result of uncertain characters, whereby the character recognition precision and reliability are improved. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention performs character recognition, and more particularly, in a character recognition method for reading an image of a document and outputting character data, it is possible to prevent recognition output of a recognition result other than characters and to achieve higher accuracy. The present invention relates to a character recognition method and a character recognition device that can perform character recognition.

[0002]

2. Description of the Related Art As a conventional character recognition method, Japanese Patent No. 2 has been adopted.
There is a technique disclosed in Japanese Patent No. 991779, in which reliability information of a single character is used to evaluate a plurality of pieces of information at a character recognition stage, and a value corresponding to the reliability of the character is calculated and used.

As another method of utilizing the certainty factor of the character, as disclosed in Japanese Patent Laid-Open No. 5-182014, the character recognition result for which low certainty factor information is obtained is corrected by the user. There is known a method of performing a display prompting the user.

Further, the technique disclosed in Japanese Patent Laid-Open No. 7-220091 discloses not only the information on each character, but also the automatic division and identification of the area, and then the character recognition.
The result is used to re-determine the attribute.

Further, the technique disclosed in Japanese Patent Application Laid-Open No. 8-101880 calculates the certainty factor of a specific area from the certainty factor of a character and changes the display method according to the calculated result. Further, the technique disclosed in Japanese Patent Laid-Open No. 9-284416 is applied to reject determination of the entire result of the image by obtaining the certainty factor of the entire document, not limited to only the area. As described above, many methods using certainty factor for character recognition have been proposed.

[0006]

However, the method of judging rejection based on the certainty factor of the document itself based on each of the above-mentioned prior arts improves the accuracy of character recognition of the entire document as much as possible, and the recognition result of objects other than characters as much as possible. When aiming for the direction of not outputting, the expected effect cannot be obtained.

When the certainty factor is obtained for each region and some processing is performed, if the region is cut out as one region in which the non-character region and the character region are fused, the certainty factor of the non-character part naturally lowers. Depending on the mixture ratio of the non-character portion, the certainty factor of the area increases or decreases.

When the certainty factor of the area is low, it is possible that the obtained character recognition result will not be utilized if the processing for changing the character area into a figure is executed. on the other hand,
Reject (exclude) all items with low confidence in character units
Then, when an image with poor quality is recognized, the rejection result occupies a large number, and an unsightly and unusable character recognition result is output. As a poor quality image,
For example, in addition to an image with a lot of noise, there are an image with a high density and characters being crushed, and, conversely, an image with a light density and faintness.

By the way, as a method for automatically discriminating a character area, a circumscribing rectangle is extracted as in Japanese Patent Laid-Open No. 7-037036 by the present applicant, and the circumscribing rectangle is classified according to its size and internal information, and character There is a method of generating an area by integrating rectangles with character rectangles. like this,
There is also a method in which after the character area is obtained by the area division method, character recognition is performed to obtain the character code and coordinates and the certainty factor thereof. As a method for obtaining this certainty factor, a technique such as the above-mentioned Japanese Patent No. 2991779 is used. But,
In this case, the unrecognizable character recognition result is output in the same manner as above due to the influence of the accuracy of the character area discrimination.

In order to solve the above-mentioned problems of the prior art, the present invention provides a character recognition method and apparatus capable of preventing recognition output of recognition results other than characters and performing character recognition with higher accuracy and reliability. The purpose is to provide.

[0011]

[Means for Solving the Problems]
In order to achieve the object, a character recognition method according to the invention of claim 1 is a character recognition method for automatically discriminating a character area on a document and recognizing characters in the character area as character data. An area dividing step of dividing a character area in which a character is described and a plurality of attribute-specific areas corresponding to other descriptions, a character recognition step of recognizing character data for each line in the character area, and the character The certainty factor calculation step of obtaining certainty factor of the character recognition result for each line in the area and outputting the certainty factor indicating the certainty factor.

According to the first aspect of the present invention, since the certainty factor is obtained for each line, the recognition result of the non-character portion is not output, and the character recognition accuracy can be improved.

Further, in the character recognition method according to a second aspect of the present invention, in the invention according to the first aspect, the certainty factor calculation step is more likely to determine the certainty of the obtained character recognition result for each line. It is characterized in that it is determined to be uncertain, and when it is determined to be uncertain, all the character recognition results of the line are replaced with predetermined characters and output.

According to the second aspect of the present invention, when the character recognition result is uncertain, by replacing the character, the replaced character can be easily processed, and the processing work after the character recognition can be facilitated. You can

Further, in the character recognition method according to a third aspect of the present invention, in the invention according to the first aspect, the certainty factor calculation step is more likely to determine the certainty of the obtained character recognition result for each line. It is determined that the line is not a character region and the region attribute is changed to another region when it is determined to be uncertain.

According to the third aspect of the present invention, it is possible to determine that character recognition has been executed for a portion that is not a character area, and the accuracy of area division can be improved.

Further, in the character recognition method according to the invention of claim 4, in the invention of claim 3, the certainty factor calculation step recognizes the character of each line in every line in the character area. When it is determined that the result is uncertain, the character area is deleted.

According to the invention of claim 4, by deleting the character area, unnecessary data output for character recognition can be prevented and the reliability of the character recognition result of the original or the like including other area attributes can be improved. it can.

Further, in the character recognition method according to a fifth aspect of the present invention, in the invention according to the third aspect, the certainty factor calculation step is performed for a part of the lines in the character area to detect the characters of the line unit. When it is determined that the recognition result is uncertain, the size of the character area is changed so that the character area is formed only with certain lines other than the certain line.

According to the invention of claim 5, the size of the character area is changed so as not to include a part of the lines which are non-character areas. Therefore, the character recognition in the character area is highly accurate. Can be realized.

The character recognition method according to a sixth aspect of the present invention is the character recognition method according to any one of the first to fifth aspects, wherein the certainty factor calculation step uses as an information for obtaining the certainty factor of the line. Among the certainty factors of the character, the number of characters having the certainty factor equal to or higher than a predetermined threshold value is used.

According to the invention of claim 6, since the certainty factor of the line is obtained by using the number of characters having the certainty factor equal to or more than a predetermined threshold value, for example, the recognition result such that the certainty as a sentence is obtained. In some cases, there are many character results with a high degree of certainty. Therefore, by using this number of characters as the degree of certainty of a line, it is possible to improve the accuracy of character recognition.

The character recognition method according to a seventh aspect of the present invention is the character recognition method according to any one of the first to fifth aspects, wherein the certainty factor calculation step uses as an information for obtaining the certainty factor of the line It is characterized by using the ratio of the number of alphanumeric characters to the number of inline characters among the characters of.

According to the invention of claim 7, when the ratio of alphanumeric characters which is not effective for the language processing at the time of character recognition in a line is large, the judgment standard of the average certainty factor is lowered to obtain the kana kanji. It is possible to maintain a predetermined character recognition accuracy.

Further, the character recognition method according to the invention of claim 8 is the character recognition method according to any one of claims 1 to 5, wherein the certainty factor calculation step uses other information as information for obtaining the certainty factor of the line. The feature is that the degree of overlap with the area of the attribute of is used.

According to the eighth aspect of the invention, the degree of overlap between the coordinate value of the character line and the area of the different attribute makes it possible to obtain the certainty factor of the character and improve the accuracy of character recognition.

A character recognition device according to the invention of claim 9 is
In a character recognition device for automatically discriminating a character area on a manuscript and recognizing a character in the character area as character data, the area on the manuscript is described as a character area in which a character is described, and a plurality of areas corresponding to other descriptions. Area dividing means for dividing the area into attribute areas, character recognition means for recognizing character data line by line in the character area, and probability of character recognition result for each line in the character area, The certainty factor calculating means which outputs the certainty factor which shows likelihood is provided, It is characterized by the above-mentioned.

According to the ninth aspect of the invention, since the certainty factor is obtained for each line, it is possible to improve the character recognition accuracy without outputting the recognition result of the non-character portion.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of a character recognition method and a character recognition device according to the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the overall construction of the character recognition device of the present invention. The character recognition device 100 character-recognizes image data read by the scanner 101 and outputs character data such as text to a display 102 and a printing device 103 such as a printer.

The character recognition device 100 includes an image memory 104 for storing image data of the scanner 101 and an image memory 1.
CPU 105 for performing character recognition processing on image data 04, C
RO in which the character recognition processing program of PU 105 is stored
M106, RAM107 used as a work area of data at the time of character recognition processing of CPU105, CPU10
The dictionary (dictionary data storage unit) 108 referred to in the character recognition processing of No. 5 is configured.

The character recognition processing program of the ROM 106 is roughly constituted by a program for each function of an area dividing unit for dividing the image data into units for character recognition processing, an OCR (character recognition) unit, and a certainty factor processing unit. There is.

FIG. 2 is a flow chart showing an outline procedure of the character recognition processing executed by the character recognition device 100. C
The PU 105 stores image data of a document or the like read from the scanner 101 and stored in the image memory 104 in the ROM 10
Character recognition is performed by the character recognition processing program 6 of FIG.

First, the area dividing unit divides the image data into areas for character recognition processing (step S20).
1). Next, character recognition is performed for each area divided by the OCR section (step S202). Next, the certainty factor processing unit obtains a certainty factor that is "certainty" with respect to the character recognition result, and adjusts the output of the character recognition result for each certainty factor (step S203).

FIG. 3 is a flowchart showing a procedure for performing the line certainty factor process in the character recognition process executed by the character recognition device 100. First, the CPU 105 divides the image data into regions by line for character recognition processing by the region dividing unit (step S301). After that, the subsequent processing is executed for each of the divided lines (step S302). Next, the OCR unit executes a character recognition process for a character area in each of the divided areas (step S30).
3).

Next, the certainty factor processing unit obtains a certainty factor of the line which is the "probability" for the character recognition result of this line (step S304). Next, it is determined whether or not the character recognition processing has been completed for all the divided areas (step S305). If not yet completed (step S30
5: No), returning to step S302, the character recognition processing for the next area is executed. When the character recognition processing for all the areas is completed (step S305: Ye
s), the result of character recognition of each line and the result of line certainty are output.

Next, FIG. 4 is a flow chart showing the procedure of character / non-character determination using the average certainty factor, which is executed by the character recognition apparatus 100. First of all,
The result after the character recognition is input, and the character recognition result is divided into regions for each line (step S401). Then, assuming that the number of row data of each divided row is i, the following processing is continued until the number of row data i reaches the total number of rows in the area (step S402). First, the certainty factor processing unit calculates an average value of the line certainty factor (in-line average certainty factor), which is the “probability” for the character recognition result of this line (step S403). Next, the calculated in-row average certainty factor is compared with a comparison threshold Th1 (step S404).

If the in-line average certainty factor exceeds the threshold Th1 (step S404: Yes), it is determined that this line is a character (step S405). On the other hand, if the in-row average certainty factor does not exceed the threshold value Th1 (step S404: N
o), it is determined that this line is non-character (step S40)
6). After that, the process returns to step S402, character / non-character determination is performed for all lines, and the determination result of the line certainty factor is output (step S407).

Next, FIG. 5 is a flow chart showing the procedure of character / non-character determination using the average certainty factor, which is executed by the character recognition apparatus 100. First of all,
The result after character recognition is input, and the character recognition result is divided into regions on a line-by-line basis (step S501). Then, assuming that the number of row data of each divided row is i, the following processing is continued until the number of row data i reaches the total number of divided rows in the area (step S502). First, the certainty factor processing unit calculates an average value of the line certainty factor (in-line average certainty factor) which is the “probability” for the character recognition result of this line (step S50).
3). Next, the calculated in-row average certainty factor is compared with the comparison threshold Th1 (step S504).

If the in-line average certainty factor exceeds the threshold Th1 (step S504: Yes), it is determined that this line is a character (step S505). On the other hand, if the in-row average certainty factor does not exceed the threshold Th1 (step S504: N
o), it is determined that this line is a non-character (step S506),
This line is replaced with a predetermined reject character determined in advance (step S507). After this, step S50
Returning to step 2, character / non-character determination for all lines and reject processing for non-characters are performed, and the determination result of the line certainty factor is output (step S508).

As this non-character, a special character, for example, a character "=" or a character code which is not normally used in OCR (outside the specified range) is replaced with a reject character (reject code). Any character may be assigned as the reject character, and the process for the reject character can be performed in a later step after character recognition. If a space is assigned as a reject character, it is erroneously determined that there is no character before recognition, so it is desirable to use a character that displays something.

The certainty factor processing unit obtains the line certainty factor by comparing it with a threshold value and then determines the line certainty factor to be either probable or uncertain, and as a result, When it is determined that the line is not accurate, the line may be determined not to be the character area.

Next, FIG. 6 is a flow chart showing a procedure for changing the area type by the character / non-character determination executed by the character recognition apparatus 100. First, the character recognition result for each area is input to the area dividing unit. Number of areas I
The initial value of n increases from 0 to the number n of areas of the entire document. Next, the input character recognition result of a certain area is divided into areas for each line (step S601). Then, with the number of row data of each divided row as i, the following processing is continued until the number of row data i reaches the total number of divided rows in the area (step S602). First, the certainty factor processing unit calculates an average value of the line certainty factor (intra-line average certainty factor) which is the "probability" for the character recognition result of this line (step S).
603). Next, the calculated in-row average certainty factor is compared with a comparison threshold Th1 (step S604).

If the in-line average certainty factor exceeds the threshold Th1 (step S604: Yes), it is determined that this line is a character (step S605). On the other hand, if the in-row average certainty factor does not exceed the threshold Th1 (step S604: N
o), it is determined that this line is a non-character (step S606),
Assuming that this line is a non-character line, the number of non-character lines is incremented (In ⁺⁺ ) (step S607). After that, the process returns to step S602, character / non-character determination is performed for all lines, and the number of non-character lines is counted, and 1
Output the determination result of the line certainty factor in the area.

At this time, at the time of outputting the result, it is determined whether or not the number of non-character lines in one area has reached the limit value n based on the obtained line certainty factor (step S608). If the number of non-character lines has not reached the limit value n (step S608:
No), it is determined that this area is a character area, and the result of the line certainty factor is output. On the other hand, when the number of non-character lines reaches the limit value n (step S608: Yes), it is determined that the area being processed is not a character area, and the attribute type of this area is changed to another (step S609). ).

FIG. 7 is a diagram showing the attributes of each area on the document 700. As shown in the figure, when the area is divided, the area attributes are a character area 701, a figure area 702, and a table area 70.
3. Classified as an enclosed frame area 704, etc., and an attribute is added.
According to the layout on the document 700, the character area 701 is provided with character recognition area orders 1 to 4 (701a to 701e).

When the area attribute is changed in step S609, for example, when the character area 701e corresponds to this, the attribute type of this area 701e is changed. For example, the process shown in FIG. 6 is executed again. In addition, this area 7
01e itself is deleted. In addition to this, the area 701e may be changed to a drawing area.

Next, FIG. 8 is a flow chart showing a procedure for changing the size of the character area by the character / non-character determination executed by the character recognition apparatus 100. First, the character recognition result for each area is input to the area dividing unit.
The number of areas In increases from 0 as an initial value to the number of areas n of the entire document. Next, the input character recognition result of a certain area is divided into areas for each line (step S801). Then, assuming that the number of row data of each divided row is i, the following processing is continued until the number of row data i reaches the total number of divided rows in the area (step S802). First, the certainty factor processing unit calculates an average value of the line certainty factor (in-line average certainty factor) which is the “probability” for the character recognition result of this line (step S803). Next, the calculated in-row average certainty factor is compared with the comparison threshold Th1 (step S80).
4).

If the in-line average certainty factor exceeds the threshold Th1 (step S804: Yes), it is determined that this line is a character (step S805). On the other hand, if the in-row average certainty factor does not exceed the threshold Th1 (step S804: N
o), it is determined that this line is a non-character (step S806),
Assuming that this line is a non-character line, the number of non-character lines is incremented (In ⁺⁺ ) (step S807). After that, the process returns to step S802, character / non-character determination is performed for all lines, and the number of non-character lines is counted.
Output the determination result of the line certainty factor in the area.

At this time, at the time of outputting the result, it is determined whether or not the number of non-character lines in one area has reached the limit value n based on the obtained line certainty factor (step S808). If the number of non-character lines has not reached the limit value n (step S808:
No), it is determined that this area is a character area, and the result of the line certainty factor is output. On the other hand, when the number of non-character lines has reached the limit value n (step S808: Yes), it is determined that the area being processed is not a character area, and the size of this area is changed (step S809).

FIG. 9 is a diagram showing an example of changing the size of the character area. As shown in FIG. 9A, it is assumed that the high confidence row 901a to 901n is continuously determined in this character area 701, and the low confidence row 901x is determined at the bottom row.
In this case, as shown in FIG.
The area size is changed so that the character area 701 is formed only by the high confidence rows 901a to 901n excluding x.

FIG. 10 shows another example of changing the character area size. As shown in FIG. 10A, in this character area 701, high confidence line 901a to 901 that are continuous.
It is assumed that the low confidence factor row 901x is determined between d and 901e to 901g. In the illustrated example, the low confidence line 901x
Is multiple lines.

In this case, as shown in FIG. 9B, one group of high confidence line 9 except for the low confidence line 901x.
One character area 701A divided by 01a to 901d
To form. Also, the other group of high confidence line 901
The other character area 701B divided by e to 901g is formed.

As described above, the character area 701 itself can be divided when the area size is changed according to the position of the low confidence factor line 901x in the character area 701. On this occasion,
The region attribute of each line of the low confidence factor line 901x can be changed from the character region to the drawing region.

Next, FIG. 11 is a flowchart showing the processing contents of the low confidence processing according to the present invention. In this low confidence factor process, the confidence factor of the recognized character result is referred to, and the process of deleting the character recognition result is executed when the average confidence factor of the line or area unit is low. At this time, the processing conditions are finely divided to leave the correct characters as much as possible and delete the recognition result that does not look like a character.

All the low confidence processings described below are performed on a line-by-line basis. The following six features are combined and used as features for determining whether a character is a character or not in this low confidence processing.

[0057] 1) Intra-row average confidence 2) High confidence character number 3) Ratio of high confidence character number 4) Number of alphanumeric characters 5) User set threshold 6) Overlapping with figures, tables, etc.

In the actual character recognition result, the number of characters in one line cannot be known until the data is analyzed for one line (up to the position of the line feed code). Therefore, it can be considered that all of the above features have been collected by the time when the area coordinate values of the line are known by performing the feature extraction at the same time as finding the line feed position. This is why the feature calculation is performed before the process of excluding the character recognition line because it overlaps the table area in the flowchart shown in FIG.

First, the character recognition result for each area is input to the area dividing unit. Then, the input character recognition result of a certain area is divided into areas (step S1).
101). Then, with the number of row data of each divided row being i, the following processing is continued until the number of row data i reaches the total number of divided rows in the area (step S1102).
First, the certainty factor processing unit calculates an average value of the line certainty factor (intra-line average certainty factor) which is the “probability” for the character recognition result of this line (step S1103). At this time, the number of high-confidence characters, for example, the number of characters with a certainty factor of 80 or more is counted.

Next, it is judged whether or not this row is included in the table (character string inside the table) (step S110).
4). If the line is included (step S110)
4: Yes), low confidence processing is not performed (return to step S1102). The reason is that there is a high possibility that a numeric string will be entered inside the table, and such a character string tends to have a relatively low certainty factor, so numerical data with a low certainty factor is not deleted. .

If the line is not included in step S1104 (step S1104: No), it is then determined whether or not the line overlaps a figure which is somewhat small with respect to the entire image (step S1105). If they overlap (step S1105: Yes), the overlap flag is turned on (step S1106). If they do not overlap (step S1105: No), the overlap flag is turned off (step S1107).

Threshold processing is used for the determination of "small figure". The range in which the result area exists is determined, and the value of 1/2 of the smaller length in the vertical and horizontal directions is used as the threshold value. Then, depending on the overlap flag with this figure, the condition for deleting characters is changed.

In this processing, the threshold for the low confidence processing is set to 50.
This is an example of the case where the state when set to (Th1 = 50) is most effective. And the overlap flag is OFF
After (Step S1107), it is determined whether the average certainty factor exceeds 60 (Step S1108), and if it exceeds (Step S1108: Yes), Step S1
Return to 102. If not exceeded (step S11)
08: No), and proceeds to step S1100.

The overlap flag is ON (step S110
After 6), it is judged whether the average certainty factor exceeds 70 (step S1109), and if the average certainty factor exceeds 70 (step S1109).
1109: Yes), and the process returns to step S1102.
If not exceeded (step S1109: No), the process proceeds to step S1110.

In step S1110, the ratio of the high confidence character number to the in-line character number is determined. The determination uses the calculation formula of (the number of high confidence characters / the number of characters in a line> 40%).
If the ratio exceeds 40% (step S1110: Y
es), and returns to step S1102. If it does not exceed (step S1110: No), there is a high confidence character number, and the average confidence of this character is the threshold Th1 (5).
0) or more is determined (step S1111). If both are satisfied (step S1111: Ye
s) and returns to step S1102. If either one is not satisfied (step S1111: No), the process proceeds to step S1112.

Next, there are a predetermined number (for example, four) of alphanumeric characters in a line, and the average certainty factor of this character is the threshold value T.
It is determined whether or not h1 or more (step S1112). If both are satisfied (step S1112: Ye
s) and returns to step S1102. If either one is not satisfied (step S1112: No), then the state of the overlap flag with the drawing is determined for this row (step S1113).

If this line overlaps the figure (the overlap flag with the figure is ON) (step S1113: YES), the procedure moves to step S1115. On the other hand, if this line does not overlap with the figure (the overlap flag with the figure is OFF) (step S1113: No), the high confidence character in the line exceeds a predetermined ratio (for example, 10%), and It is determined whether the average certainty factor exceeds a threshold value (step S111).
4). If it exceeds (step S1114: Yes),
The procedure returns to step S1102. If either one is not satisfied (step S1114: No), step S
Moving to 1115. In step S1115, the character data first candidate in the line is replaced with a space, and step S1
Return to 102.

Language processing is often used for the information for obtaining the above certainty factor. The character line is cut out from the area that has become the character area by the area identification first. For example, after taking a projection and cutting out a line in a part with few black pixels,
Characters are extracted using the projection and the circumscribed rectangle of black pixels.
At this time, in Japanese, one character does not necessarily become one rectangle, so several kinds of cutting methods are used for one character, and the character is cut so that the best result is the final result.

After this, post-processing is executed. The cut out character string is cut out in word units by a method such as morphological analysis, and this word is matched with a word in the language dictionary. Then, a high degree of certainty is obtained when a recognition result that is grammatically consistent is obtained. In the case of such a recognition result in which the certainty is obtained as a sentence, many character results have a high degree of certainty. In such a case, as in the process of step S1110 described above, the number of characters that can be recognized more reliably than the average certainty factor is very effective.

In the post-processing, for example, "doorway"
Suppose there is a word, and it was also in the word dictionary. However, it is assumed that the person who was recognized was a "departure person" (and alone). The second candidates for recognition of "person" and "b" are "on" and "mouth", respectively. Therefore, if the second candidate and the first candidate are replaced with each other, a matching word can be created in the word dictionary, and it is determined that this is the correct answer. Such post-processing has an influence on the above certainty factor, and the recognition result matched with the word dictionary also has a higher certainty factor.

In the above low confidence processing, conversion into space is performed in units of lines. Therefore, if all the lines in a certain area are converted into spaces, it is useless to output a large number of spaces as recognition results using the area as a character area. Therefore, when all the lines in the area are replaced with spaces, the character area size changing process (the processing of step S809 in FIG. 8) is executed and the area itself is deleted from the result (see FIGS. 9 and 10). ). Further, the processing is not limited to deletion, and processing for changing the attribute of this area to a drawing or the like is also possible.

Further, in the above process, when deleting the in-line character, the space code is inserted as the first candidate, and the previous candidate characters are stored in the second and lower candidates in the form of lowering the candidate rank. Since it is not a process of replacing the first candidate itself with a space, by referring to the intermediate data generated for the recognition process and stored separately, it is possible to replace the candidate again in the subsequent process.

Further, according to the above processing, if the character line overlaps with the figure as a result (step S1105: Y
es), each processing executed below (step S110)
9. Since the value for comparison (average certainty factor) used in step S1114) is high, a large number of characters are deleted.

Further, in the above processing, the threshold value (Th1) of the low confidence processing is set to 50. However, when it is desired to set the threshold value to 50 or more, the threshold value of a desired value is set to Th1. The same processing can be performed. On the other hand, when the user-set threshold value is lower than 50, it is a request to output a character with a low certainty factor. Therefore, in this case, unlike the processes in the above flowchart, the average certainty factor in a line is compared with the user-set threshold value. Then, the process of determining whether it is high or low is executed.

Further, in the processing of step S1110, the proportion of high confidence characters in the number of characters in the entire line is determined. This prevents fluctuations in reliability when, for example, one line has 5 characters and 3 of them have high confidence, and 1 line has 40 characters and 3 of which have high confidence. it can.

Further, the process of comparing the average of the certainty factors of the characters in the line with the threshold value (Th1) set by the user (step S1).
111, step S1114), a simple processing procedure with as few parameters as possible can be performed, and the intention of the user can be easily reflected.

Further, a process of comparing the ratio of alphanumeric characters included in a line to the number of characters in a line (step S111)
By 2), the influence of alphanumeric characters on the certainty factor is reduced. More specifically, among alphanumeric characters, there are relatively many similar characters such as b and 6, q and 9, o and 0, and s and S, and almost no language processing is effective for numbers. Based on the above reasons, the certainty factor tends to be lower than that of Kana-Kanji. Therefore, when a large number of alphanumeric characters are present in a line, the process of lowering the criterion for determining the average certainty factor is effective and is used.

Further, when calculating the certainty factor within a line, the position overlap between the coordinate value of the character line and the area of another attribute is used instead of the certainty factor of the character. In particular, the determination of the overlap with the table area (step S1104) is used. It is effective not to execute all the subsequent low confidence processing on the character lines inside the table area. If there are many alphanumeric characters in the result of character recognition for the table area, it is likely that the result of recognizing the numerical table is high, so the processing that removes the character with low certainty in the subsequent processing is excluded. It is desirable to do.

Similarly, by determining the overlap with the drawing or the photograph (step S1105), if the average confidence factor of the entire one line in the character area overlapping the drawing is low, a part of the drawing is deleted. It is easy to determine that this is due to character recognition. Here, it is not limited to using all the drawing areas, and for example, a configuration may be adopted in which only the drawing that becomes the entire surface of the image when displayed in a rectangle is excluded from the determination target.

Characteristic of judging character or other by the low confidence processing described above 1) Average in-line confidence 2) High confidence character number 3) High confidence character number ratio 4) Alphanumeric character number 5) User set threshold value 6 ) For the overlap with the figure, the table, etc., it is possible to obtain the certainty factor of the row by using at least one piece of information of these features in combination.

Further, when the feature including the certainty factor information and the feature of the area information including the character line (overlapped) are used, the character line of the character line is changed depending on the type of the area including the character line. It is also possible to adopt a configuration in which the threshold value for comparing certainty factor information is changed.

In the character recognition method described in this embodiment, a program prepared in advance is personalized.
It can be realized by executing on a computer such as a computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, a floppy (R) disk, a CD-ROM, an MO, or a DVD, and is executed by being read from the recording medium by the computer.
The program can be distributed via the recording medium and a network such as the Internet.

[0083]

As described above, according to the first aspect of the invention, the character recognition method according to the first aspect of the invention automatically determines the character area on the original and determines the characters in the character area. In the character recognition method that recognizes as character data,
An area dividing step of dividing an area on the original document into a character area in which characters are described and a plurality of attribute-specific areas corresponding to other descriptions, and character recognition for recognizing character data for each line in the character area Since the step and the certainty factor calculation step of obtaining the certainty factor of the recognition result of the character for each line in the character area and outputting the certainty factor indicating the certainty factor, the recognition result of the non-character part is thereby obtained. It is possible to improve the character recognition accuracy without outputting.

According to the invention of claim 2, in the invention of claim 1, the certainty factor calculation step is:
The certainty of the obtained character recognition result for each line is judged to be probable or not certain, and when it is not confirmed, all the character recognition results of the line are predetermined. Since the output is performed by replacing the character with a predetermined character, it is possible to easily perform the processing on the replaced character and to facilitate the processing work after the character recognition.

According to the invention described in claim 3, in the invention described in claim 1, the certainty factor calculation step is:
The certainty of the obtained character recognition result for each line is judged to be probable or not certain. If not, it is judged that the line is not a character area, and Since the attribute is changed to another area, it is possible to determine that character recognition has been executed for a portion that is not a character area, and thus it is possible to improve the accuracy of area division.

According to the invention described in claim 4, in the invention described in claim 3, the certainty factor calculation step is:
When it is determined that the recognition result of the characters in each line is not accurate in all the lines in the character area, the character area is deleted, thereby preventing unnecessary character recognition data output. Thus, it is possible to improve the reliability of the character recognition result of a document or the like including other area attributes.

According to the invention of claim 5, in the invention of claim 3, the certainty factor calculation step is:
For some lines in the character area, when it is determined that the recognition result of the characters in each line is not accurate,
Since the size of the character area is changed so that the character area is formed only by the probable rows excluding the line, it is possible to improve the accuracy of character recognition in the character area.

According to a sixth aspect of the present invention, in the invention according to any one of the first to fifth aspects, the certainty factor calculation step uses as information for obtaining the certainty factor of the line, Among the certainty factors of characters, the number of characters having certainty factors equal to or higher than a predetermined threshold value is used, so for example, when the recognition result is such that certainty is obtained as a sentence, high certainty factor is obtained. Since the number of character results increases, the use of this number of characters for the certainty factor of the line has the effect of improving the accuracy of character recognition.

Further, according to the invention described in claim 7, in the invention described in any one of claims 1 to 5, the certainty factor calculation step is performed as information for obtaining the certainty factor of the line. Since the ratio of the number of alphanumeric characters to the number of characters in a line is used among characters, if the ratio of alphanumeric characters that is not effective for language processing at the time of character recognition in a line is large, lower the criterion for determining the average certainty level, etc. It is possible to maintain the predetermined character recognition accuracy obtained in step 1.

Further, according to the invention described in claim 8, in the invention described in any one of claims 1 to 5, the certainty factor calculation step uses other information as information for obtaining the certainty factor of the row. Since the degree of position overlap with the attribute area is used, depending on the degree of position overlap between the character line coordinate values and the area of another attribute,
It is possible to obtain the certainty factor of the character and to improve the accuracy of the character recognition.

According to the ninth aspect of the invention, in the character recognition device for automatically discriminating the character area on the original and recognizing the characters in the character area as character data, the area on the original is recognized as a character. Area dividing means for dividing into a plurality of attribute-based areas corresponding to other descriptions, character recognition means for recognizing character data on a line-by-line basis within the character area, and inside the character area Since a certainty factor of each character is obtained for each line and a certainty factor calculation unit for outputting a certainty factor indicating the certainty factor is output, the certainty factor of each line is obtained in this way, so that the non-character portion The character recognition accuracy can be improved without outputting the recognition result.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an outline procedure of character recognition processing of the character recognition device according to the embodiment of the present invention.

FIG. 3 is a flowchart showing a procedure for performing line certainty degree processing in the character recognition processing of the character recognition device according to the embodiment of the present invention.

FIG. 4 is a flowchart showing a procedure of character / non-character determination using the average certainty factor of the character recognition device according to the embodiment of the present invention.

FIG. 5 is a flowchart showing a procedure of character / non-character determination using the average certainty factor of the character recognition device according to the embodiment of the present invention.

FIG. 6 is a flowchart showing a procedure for changing an area type by character / non-character determination of the character recognition device according to the embodiment of the present invention.

FIG. 7 is a diagram showing attributes of respective areas on a document of the character recognition device according to the embodiment of the present invention.

FIG. 8 is a flowchart showing a procedure for changing the size of a character area by character / non-character determination of the character recognition device according to the embodiment of the present invention.

FIG. 9 is a diagram showing an example of changing the size of a character area of the character recognition device according to the embodiment of the present invention.

FIG. 10 is a diagram showing another modification of the size of the character area of the character recognition device according to the embodiment of the present invention.

FIG. 11 is a flowchart showing the processing contents of the low confidence degree processing of the character recognition device according to the embodiment of the present invention.

[Explanation of symbols]

100 Character Recognition Device 101 Scanner 102 Display 103 Printing Device 104 Image Memory 105 CPU 106 ROM 107 RAM 108 Dictionary 700 Original 701 (701a to 701e, 701A, 701B)
Character area 702 Figure area 703 Table area 704 Enclosing frame areas 901a to 901n High confidence line 901x Low confidence line

Claims (9)

[Claims] 1. A character recognition method for automatically recognizing a character area on a document and recognizing a character within the character area as character data, comprising: Area dividing step of dividing into a plurality of attribute-based areas corresponding to, character recognition step of recognizing character data in each line unit in the character area, and certainty of character recognition result in each line unit of the character area. And a confidence factor calculation step of outputting a confidence factor indicating the certainty factor, and a character recognition method comprising: 2. The certainty factor calculation step determines the certainty of the obtained recognition result of the characters in line units into two, which is likely and not certain, and when it is determined that the certainty is not certain. The character recognition method according to claim 1, wherein all the character recognition results of the line are replaced with predetermined characters and output. 3. The certainty factor calculation step determines the certainty of the obtained recognition result of the characters on a line-by-line basis to be probable and not probable. The character recognition method according to claim 1, wherein the line is determined not to be a character region, and the region attribute is changed to another region. 4. The certainty factor calculating step deletes the character area when it is determined that the recognition result of the character of each line is not accurate in all the lines in the character area. The character recognition method according to claim 3. 5. The certainty factor calculation step, if it is determined that the recognition result of the character of each line is uncertain for a part of the lines in the character region,
4. The character recognition method according to claim 3, wherein the size of the character area is changed so that the character area is formed only by a certain line excluding the line. 6. The certainty factor calculation step uses, as the information for obtaining the certainty factor of a line, the number of characters having the certainty factor of a predetermined threshold value or more among the certainty factors of the characters in the line. The character recognition method according to any one of claims 1 to 5. 7. The certainty factor calculation step uses a ratio of the number of alphanumeric characters in a line to the number of characters in the line as the information for obtaining the certainty factor of the line. Character recognition method described in. 8. The confidence factor calculation step uses the degree of overlap with a position of another attribute as the information for obtaining the confidence factor of a row, according to any one of claims 1 to 5. Character recognition method described. 9. A character recognition device for automatically discriminating a character area on a manuscript and recognizing characters in the character area as character data. Area dividing means for dividing into a plurality of attribute-based areas, character recognition means for recognizing character data in each line unit in the character area, and certainty of character recognition result in each line unit in the character area. And a certainty factor calculating means for outputting the certainty factor indicating the certainty factor, and the character recognition device.