JP2000057261A - Character segmenting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character segmenting device which improves a segmenting method for a character segmentation object pattern. SOLUTION: Image data Si supplied from a scanner, etc., are stored in an image storage part 1 and the image data S1 are outputted from the image storage part 1. Further, feature information S3 set by an input part or a control part, etc., is already stored in a feature storage part 3 and outputted to a character segmentation part 2. The character segmentation part 2 cuts characters out of the image data S1, one by one, through character segmentation corresponding to the feature information S3. The cutting result is outputted as the character cutting result S2 to the outside of the character segmenting device.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character extracting apparatus having an improved method for extracting a character extracting target pattern composed of characters or symbols.

[0002]

2. Description of the Related Art In a conventional character extracting device, characters are extracted from an image input from a reading device such as a scanner. FIG. 2 is a diagram showing an example of character extraction processing of horizontally written characters in a conventional character extraction device, and FIG. 3 is a flowchart for explaining the acquisition of character extraction candidate points in FIG. The character extraction processing (a) to (d) in the conventional character extraction device will be described with reference to FIGS.

(A) A projection distribution of pixels (for example, black dots) of a character pattern in the vertical direction is measured, and a middle point where the value of the projection distribution becomes 0 is a character extraction candidate point 1.
~ 8. In this case, as shown in FIG. 2, the projection distribution is measured by the number of black dots perpendicular to the row (step ST1). The start point coordinates are deleted (step ST2), and the processing coordinates X are adjusted to the left end of the row (step ST3). The processing coordinate X is moved to the right until the value of the projection distribution is no longer 0 (step ST4). It is determined whether or not the value of the projection distribution at the processing coordinates X is 0 (step ST5). If not, it is determined whether or not the starting point coordinates are stored (step ST8). If the start point coordinates are not stored, the processing coordinates X are moved to the right by one dot (step ST).
6). It is determined whether or not the processing coordinate X is at the right end of the row. If the processing coordinate X is at the right end, the process is terminated. If not, the process returns to step ST5 (step ST7).

If the value of the projection distribution is 0 in step ST5, the process proceeds to step ST12. It is determined whether or not the start point coordinates are stored (step ST1).
2) If it is stored, the process proceeds to step ST6. If it is not stored, the current processing coordinate X is stored as the start point coordinate, and the process proceeds to step ST6 (step ST1).
1). If the start point coordinates are stored in step ST8, the process proceeds to step ST9. ｛(Processing coordinates X
-1) + Start point coordinate｝ / 2 is set as a character extraction candidate point (step ST9). The start point coordinates are deleted and the step ST
The process proceeds to step 6 (step ST10).

(B) For example, the maximum value of the interval between the adjacent character extraction candidate points 1 to 8 (in FIG. 2, the character extraction candidate point 5
And the character separation candidate point 6) as a reference interval W. (C) Character extraction candidate points having an interval equal to or less than a fixed value (for example, 1/3 or less of the reference interval W) from the reference interval W (FIG.
Then, character extraction candidate points 4, 8) are excluded from the candidates. (D) Remaining character extraction candidate points 1, 2, 3, 5, 6,
7 to the final character cutout positions A, B, C, D, E, F
And In FIG. 2, the maximum value of the intervals between the character extraction candidate points 1 to 8 is set as the reference interval W. However, this is an example of a method for determining the reference interval. The value may be set as the reference interval W.

[0006]

However, the conventional character extracting apparatus has the following problems. FIG.
(A), (b), (c) is a figure which shows the example of character extraction which becomes a problem. In FIG. 4A, when “old” and “1 day” in the row La are compared, the projected distribution of black dots in the vertical direction is similar, but “old” is cut out as one character, and “1” is cut out.
It is necessary to determine that the date is cut out as two characters.
In FIG. 4B, it is necessary to determine whether a minute pattern in the line Lb is cut out as a character (for example, a period) or not as garbage. In FIG. 4C, it is necessary to determine that the pattern on the lower side in the line Lc is cut out as a character (for example, a period) or processed as garbage. However, it is difficult to appropriately perform these determinations on an arbitrary image, and one character is mistakenly replaced by two.
There has been a problem that the character extraction rate is reduced by extracting as characters, or unnecessary characters are extracted by extracting garbage as characters.

[0007]

According to a first aspect of the present invention, there is provided a character extracting apparatus, comprising: Image storage means for storing image data, characteristic information storage means for storing characteristic information of the pattern for character extraction, and character extraction means for extracting characters from the stored image data based on the characteristic information. ing. By adopting such a configuration, the image data of the form is stored in the image storage unit, and the characteristic information of the character extraction target pattern is stored in the characteristic information storage unit. In a character extracting unit, a character is extracted from the image data stored in the image storage unit based on the characteristic information.

In the invention according to a second aspect, in the character extracting apparatus according to the first aspect, a target character storage unit for storing a character extraction target pattern, and a character characteristic for inputting the character extraction target pattern and outputting the characteristic information. Determining means for storing the characteristic information output by the character characteristic determining means in the characteristic information storage means. By adopting such a configuration, the character extraction target pattern is stored in the target character storage unit, and the character extraction target pattern is input to the character feature determination unit and the characteristic information is output. The feature information output from the character feature determination unit is stored in a feature information storage unit. According to a third aspect of the present invention, in the character extracting device according to the first or second aspect, the characteristic information represents a characteristic unique to the character extraction target pattern. According to a fourth aspect of the present invention, in the character extracting apparatus according to any one of the first to third aspects, the character extracting unit extracts a character based on the characteristic information output from the characteristic information storing unit.

According to a fifth aspect of the present invention, in the character extracting device according to the second, third or fourth aspect, the target character storage means stores a character code of a character extraction target pattern. In the invention according to claim 6, claim 2,
In any of the character extraction devices of 3, 4, or 5, the character characteristic determination unit includes a characteristic information table storage unit that stores the characteristic information previously associated with a character code of the character extraction target pattern. In the invention according to claim 7, in the character extracting apparatus according to claim 6, the characteristic information table storage means stores the character code and the characteristic information of the character code in association with each other. In the invention according to claim 8, in the character cutout apparatus according to claim 6 or 7, the character feature determination means stores the feature information corresponding to the character code output from the target character storage means in the feature information table storage means. And outputs it to the feature information storage means.

[0010] In the invention according to claim 9, claims 1, 2,
In the character extracting device of 3, 4, 5, 6, 7 or 8,
The unique features of the character extraction target pattern include the presence or absence of left / right separable characters, the presence / absence of upper / lower separable characters, the presence / absence of small characters, the presence / absence of subscripts, the presence / absence of superscript characters, and the number It indicates at least one of the presence or absence of a character other than. In the invention according to claim 10, claims 1, 2, 3, 4, 5, 6, 7, 8
Or, in the character extracting device according to 9, when the character information output by the characteristic information storage means does not include left-right separable characters, a vertical projection distribution value is set in horizontal writing character extraction. Cut out all sections where the value is less than the value as characters, or if there is no upper / lower separable character, cut out all sections where the value of the horizontal projection distribution is less than or equal to the set value in vertical writing character extraction. I have to.

[0011] In the invention according to claim 11, claim 1,
In the character cutout device of 2, 3, 4, 5, 6, 7, 8 or 9, the character cutout means may include a case where a minute character smaller than a normal character does not exist in the feature information output by the feature information storage means. If the size of the pattern is not cut out as a character below the predetermined threshold, or if there is no subscript, the upper end of the pattern is below the predetermined threshold below the line Is not cut out as a character, or when there is no superscript character, a character whose lower end of the pattern is equal to or greater than a predetermined threshold value above the line is not cut out as a character. In the invention according to claim 12, claims 1, 2, 3, 4, 5, 6,
In the character extracting device according to 7, 8, or 9, the character extracting means is characterized in that the characteristic information output from the characteristic information storage means does not include any character other than a numeral, and the interval between characters is equal to or longer than a reference interval. In this case, the character interval is cut out in accordance with the length of the character interval.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment FIG. 1 is a block diagram of a character extracting device according to a first embodiment of the present invention. The character extracting apparatus includes an image storage unit (for example, an image storage unit) 1 for storing image data Si of a form in which a character extracting target pattern composed of characters or symbols is written in a horizontal or vertical direction. The image storage unit 1 is configured by a memory capable of storing the image data Si in a format in which the two-dimensional coordinates of the character extraction target pattern on the form can be reproduced. In addition, the character extraction device is provided with a storage unit (for example, a characteristic information storage unit) 3 for characteristic information unique to the character extraction pattern. The feature information storage unit 3
Feature information S3 representing features unique to the character extraction target pattern
Is stored in the memory. On the output side of the feature information storage unit 3, a character extracting unit (for example, a character extracting unit) 2 is connected. The character cutout unit 2 has a function of cutting out characters by inputting the image data S1 and the feature information S3 output from the image storage unit 1, and cutting out characters and outputting a character cutout result S2 outside the character cutout device. Is what you do.

Next, the operation of FIG. 1 will be described. Image data Si supplied from a scanner or the like (not shown) is stored in the image storage unit 1, and the image storage unit 1 outputs image data S1. Further, feature information S3 set by an input unit or a control unit (not shown) is stored in the feature storage unit 3, and the feature information S3 is output to the character cutout unit 2. The character cutout unit 2 cuts out characters one by one from the image data S1 by character cutout corresponding to the characteristic information S3. The extracted result is output to the outside of the character extraction device as a character extraction result S2. Here, the content of the feature information S3 is
For example, there are the following (a) to (f). (A) There are no separable characters on the left and right. (B) There are no upper and lower separable characters. (C) There is no minute character smaller than a normal character. (D) There is no superscript. (E) There is no subscript. (F) Only numbers (0-9) are present.

FIGS. 5 (a), 5 (b) and 5 (c) are diagrams for explaining characters which can be separated to the left and right and those which cannot be set in the setting contents (a). As shown in FIG. 5A, a character that can be separated to the left and right is a character in which there is a portion where the value of the projection distribution of the black dot in the vertical direction is 0 in one character. The non-separable character on the left and right is a character having no portion where the value of the projection distribution becomes 0, as shown in FIG. Also, as shown in FIG. 5 (c), depending on the font, the font may not be separable left and right, but any font that has a font separable left and right is regarded as a character separable left and right. Similarly, FIGS. 6A and 6B are diagrams for explaining characters that can be vertically separated and characters that cannot be separated. FIG. 6A shows characters that can be separated up and down.
(B) shows characters that cannot be separated vertically. As in the case of FIG. 5, all fonts that have a vertically separable font are regarded as vertically separable characters.

The small characters in the setting content (c) are, for example, commas (,) and single quotes (').
A character whose vertical and horizontal size is smaller than other general characters (for example, zero “0”) (for example, each vertical and horizontal size is 以下 or less of the general character size). . The superscript in the set answer (d) is, for example, a single quotation mark (') or a semi-voiced mark (゜)
The lower end of the character is a character above the line (for example, higher than the center of the line). The subscript in the set answer (e) is a lowercase letter “tsu” representing a comma (,)
The upper end of the character, such as "@", is the character below the line (for example, lower than the center of the line).

The operation of the character extracting section 2 based on the setting contents (a) to (f) is as shown in the following (1-1) to (1-6). (1-1) Operation Based on Setting Content (a) FIG. 7 is a diagram illustrating an operation of the character cutout unit 2 when there is no separable character on the left and right in the line L. When the setting content (a) is stored in the feature storage unit 3 and output to the character cutout unit 2 as the feature information S3, the character cutout unit 2 controls so that all of the character cutout candidates are set as character cutout positions. Therefore, at “February 11” in the row L in FIG. 7, the character extraction positions are character extraction candidates,..., And “2”, “month”, “1”, “1”, “day” It is cut out exactly. On the other hand, in the conventional method, the reference interval W is between the character extraction candidate and the character extraction candidate, and the interval between the character extraction candidate and the character extraction candidate is less than half of the reference interval W. Therefore, the character extraction candidate is excluded from the extraction candidates. . Therefore,
“One day” is erroneously cut out as one character (ie, “old”).

(1-2) Operation Based on Setting Content (b) FIG. 8 is a diagram for explaining the operation of the character cutout unit 2 when there is no vertically separable character in the line L. When the setting content (b) is stored in the feature storage unit 3 and output to the character extraction unit 2 as the characteristic information S3, the character extraction unit 2 sets (1)
As in the case of -1), control is performed so that all character extraction candidates are set as character extraction positions. Therefore, in "Are, that" in the line L in FIG. 8, the character extraction positions are character extraction candidates,,, and "A", "RE", ","
"So" and "re" are cut out exactly. On the other hand, in the conventional method, the reference interval W is between the reference interval and the character cutout candidate,
Since the interval with the character extraction candidate is not more than half of the reference interval W, the character extraction candidate 3 is excluded from the extraction candidates. Therefore, ", so" is erroneously cut out as one character.

(1-3) Operation Based on Setting Content (c) FIG. 9 is a diagram for explaining the operation of the character cutout unit 2 when a minute character does not exist in the line L. When the setting content (c) is stored in the feature storage unit 3 and output to the character extraction unit 2 as the characteristic information S3, the character extraction unit 2 does not extract a character having a pattern size equal to or smaller than a preset threshold value as a character.
If the threshold value is, for example, の of the height of the line, the dusts A to E in FIG. 9 are not cut out as characters. Therefore, the character extraction accuracy is improved. (1-4) Operation Based on Setting Content (d) FIG. 10 is a diagram illustrating an operation of the character cutout unit 2 when a superscript character does not exist in the line L. In this figure, numerals, commas (,) and garbage A to C are mixed. When the feature storage unit 3 stores the setting content (d) and outputs it to the character extraction unit 2 as the characteristic information S3, the character extraction unit 2 sets the lower end of the pattern above the line L to be equal to or greater than a predetermined threshold (for example, Those existing above the center of the line L) are not cut out as characters. In the conventional method, the garbage A to C existing in the line L is cut out as a character, but in the present embodiment, the garbage A is not cut out as a character because the lower end of the garbage A is located higher than the center of the line L. . Therefore, the character extraction accuracy is improved.

(1-5) Operation Based on Setting Content (e) FIG. 11 is a diagram for explaining the operation of the character cutout unit 2 when a subscript character does not exist in the line L. In this figure, numerals, single quotes ('), double quotes ("), and garbage A are mixed. The setting contents (e) are stored in the feature storage unit 3 and output to the character extraction unit 2 as feature information S3. In this case, the character cutout unit 2 does not cut out a character whose upper end of the pattern is below a preset threshold below the line L (for example, below the center of the line L) as a character. Then, the trash A existing in the row L is cut out as a character.
Is located at a position lower than the center of the line L, and is not cut out as a character. Therefore, the character extraction accuracy is improved.

(1-6) Operation Based on Setting Contents (f) FIG. 12 is a diagram for explaining the operation of the character cutout unit 2 when only a numeral is present in the row L. When the setting content (f) is stored in the feature storage unit 3 and output to the character extraction unit 2 as the characteristic information S3, the character extraction unit 2 performs the following (i) to (i).
Cut out characters as in v). (I) The projection distribution of black dots in all directions is measured, and character cutout candidates are obtained. (Ii) Find the minimum value min of the interval between adjacent character extraction candidates among the character extraction candidates. (iii) When the interval between adjacent character extraction candidates is at least n times the minimum value min, character extraction candidates are created equally in the interval between character extraction candidates according to the value of n. For example,
If the value of n is n <1.5, no division is performed, and 1.5 ≦ n <
In the case of 2.5, a process of dividing into two is performed, and in the case of 2.5 ≦ n, a process of dividing into three is performed.

(Iv) All the created character extraction candidates are set as character extraction positions. As shown in FIG. 12, when two “4” and “4” are in contact with each other,
Character extraction candidate points are acquired, and no character extraction is performed at one place. On the other hand, in the present embodiment,
When the minimum value min of the interval between the character extraction candidate points is between the character extraction candidate point and the character extraction candidate point, the distance between the character extraction candidate point and the character extraction candidate point is about twice the minimum value min. Not more than 5 times. Therefore, a character extraction candidate is created by dividing the space between the character extraction candidate point and the character extraction candidate point into two equal parts, and final character extraction positions A, B, C, D, E, and F are obtained. Therefore, even if the character touches, the character is accurately cut out.

Second Embodiment FIG. 13 is a block diagram of a character extracting apparatus according to a second embodiment of the present invention, and is common to the elements in FIG. Are given. In this character extracting apparatus, a target character storage unit 5 and a character feature determination unit 4 having a feature information table storage unit 4A are provided. In FIG. 13, the target character storage unit 5 stores a character code (for example, a JIS code) S5 to be extracted.
And outputs the character code S5 to the character feature determination unit 4. The characteristic information table storage unit 4A is a table in which character information is previously associated with a character code to be extracted. The character feature determination section 4 outputs feature information S4 corresponding to the character code S5. The feature information S4 is input to the feature information storage unit 3 and stored as the feature information S3. Other configurations are the same as those in FIG.

The operation of this character extracting apparatus differs from the character extracting apparatus shown in FIG. 1 in the following points. The target character storage unit 5 outputs the character code S5 to the character feature determination unit 4.
The character feature determination unit 4 extracts feature information corresponding to the character code S5 from the feature information table storage unit 4A, and outputs it to the feature information storage unit 3 as feature information S4. For example, if the character extraction target (character code) is a number “0” to “9” and a comma “,”, the character feature determination unit 4 is associated with the number “0”. It is a character that cannot be separated, (2) it is a character that cannot be separated vertically, (3) it is not a small character, (4) it is not a superscript, (5) it is not a subscript, and (6) it is a number. The information S4 is output to the feature information storage unit 3. The numbers "1" to "9" are similarly associated with "0" in the same manner as (1) a character that cannot be separated left and right, (2) a character that cannot be separated vertically.
(3) Not a small character, (4) Not a superscript, (5)
Characteristic information S4 such as not being a subscript or (6) being a number is output. Furthermore, (1) a character that cannot be separated horizontally, (2) a character that cannot be separated vertically, (3) a minute character, (4) a non-superscript character, 5) is a subscript, (6) is not a number,
Is output.

The characteristic information storage unit 3 integrates the above characteristic information S4, (A) there is no character that can be separated on the left and right, (B) there is no character that can be separated on the top and bottom, and (C) it is better than normal characters Characteristic information S3 is stored, such as small small characters, (D) no superscript, (E) subscript, and (F) characters other than numbers (0 to 9). (A),
(B) and (D) correspond to the feature information (a), (b), and (d) described in the first embodiment. Therefore, character extraction using the feature information S3 is performed in the first embodiment. Processing combining (1-1), (1-2), and (1-4) in
Alternatively, the processing is a combination of any one or more. As described above, in the second embodiment, the target character storage unit 5
Stores the character code S5, the character characteristic determination unit 4 outputs the characteristic information S4 converted by the characteristic information table storage unit 4A from the character code S5, and the characteristic information storage unit 3 stores the characteristic information S3 from the characteristic information S4. So, the first
In addition to the advantages of the embodiment, character extraction can be performed in a simple procedure by using a character code defined more uniquely than characteristic information defined by complicated information.
Note that the present invention is not limited to the above embodiment, and various modifications are possible. For example, the thresholds of the setting contents (c) to (e) in the first embodiment may be set based on the height of the line or by the number of dots of the character size.

[0025]

As described in detail above, according to the first aspect of the present invention, the characteristic information of the character extraction target pattern is stored in the characteristic information storage means, and based on the characteristic information,
Since the character is cut out from the image data stored in the image storage means by the character cutout means, the character cutout accuracy can be improved. According to the invention according to claim 2, the character extraction target pattern is stored in the target character storage means, and the characteristic information is output from the character extraction target pattern.
In addition to the effect of the invention according to claim 1, character extraction can be performed in a simpler procedure. According to the third aspect of the invention, the characteristic information represents a characteristic peculiar to the character extraction target pattern, so that the character extraction accuracy can be improved.
According to the fourth aspect of the present invention, the character cutout unit cuts out the character based on the feature information output from the feature information storage unit, so that the character cutout accuracy can be improved.

According to the fifth aspect of the present invention, since the target character storage means stores the character code of the character extraction target pattern, the character extraction accuracy can be improved. According to the invention according to claim 6, the character feature determination means includes the feature information table storage means for storing the feature information previously associated with the character code of the character extraction target pattern, so that the character extraction accuracy can be improved. . According to the invention according to claim 7, the characteristic information table storage means stores the character code and the characteristic information of the character code in association with each other, so that the character extraction accuracy can be improved. Claim 8
According to the invention, the character feature determination unit acquires the feature information corresponding to the character code output from the target character storage unit from the feature information table storage unit and outputs the acquired feature information to the feature information storage unit. Cutting accuracy can be improved. According to the ninth aspect of the present invention, the character extraction target pattern-specific features include a character extraction target pattern having left / right separable characters, a vertical character separable character, a minute character, a subscript character, Since at least one of the presence of a superscript character and the presence or absence of a character other than a number is indicated, the accuracy of character extraction can be improved.

According to the tenth aspect, the character extracting means sets the value of the vertical projection distribution in the horizontal character extracting when the character information which is output from the characteristic information storage means does not include a character which can be separated into right and left. All the sections where the value of the horizontal projection distribution is less than or equal to the set value in vertical character extraction are extracted as characters. Since the clipping is performed, the precision of character clipping can be improved. According to the eleventh aspect of the present invention, the character cutout means has a pattern size equal to or smaller than a predetermined threshold when there is no small character smaller than a normal character in the characteristic information output from the characteristic information storage means. Is not extracted as a character,
If there is no superscript character, do not cut out the character with the upper end of the pattern below the predetermined threshold below the line as a character, or if there is no superscript character, Since the pattern having the lower end of the pattern exceeding the predetermined threshold is not cut out as a character, the character cutout accuracy can be improved. According to the twelfth aspect of the invention, the character extracting unit is configured to output the characteristic information stored by the characteristic information storage unit when there is no character other than a numeral and the interval between the characters is equal to or longer than a reference interval. Since the character interval is cut out in accordance with the length of the character interval, the character extraction accuracy can be improved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a character cutout device according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a conventional character cutout process.

FIG. 3 is a flowchart in FIG. 2;

FIG. 4 is a diagram illustrating an example of extracting a character in question.

FIG. 5 is a diagram illustrating characters that can be separated into right and left and characters that cannot be separated.

FIG. 6 is a diagram illustrating characters that can be separated vertically and characters that cannot be separated.

FIG. 7 is a diagram illustrating an operation when there is no character that can be separated on the left and right.

FIG. 8 is a diagram illustrating an operation when there is no character that can be vertically separated.

FIG. 9 is a diagram for explaining an operation when a small character does not exist;

FIG. 10 is a diagram illustrating an operation when a superscript does not exist.

FIG. 11 is a diagram illustrating an operation when a subscript does not exist.

FIG. 12 is a diagram illustrating an operation when only a numeric string exists.

FIG. 13 is a configuration diagram of a character cutout device according to a second embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 image storage unit 2 character extraction unit 3 feature information storage unit 4 character feature determination unit 4A feature information table storage unit 5 target character storage unit

Claims (12)

[Claims] An image storage unit configured to store image data of a form in which a character extraction target pattern including a symbol or a character is written; a characteristic information storage unit configured to store characteristic information of the character extraction target pattern; Character extracting means for extracting a character from the stored image data based on the character data. 2. A target character storage unit that stores a character extraction target pattern, and a character characteristic determination unit that inputs the character extraction target pattern and outputs the characteristic information, wherein the character characteristic determination unit outputs 2. The character extracting apparatus according to claim 1, wherein characteristic information is stored in said characteristic information storage means. 3. The character extraction device according to claim 1, wherein the characteristic information represents a characteristic unique to the character extraction target pattern. 4. The character extracting device according to claim 1, wherein the character extracting unit extracts a character based on the characteristic information output from the characteristic information storage unit. 5. The character extraction device according to claim 2, wherein the target character storage unit stores a character code of a character extraction target pattern. 6. The character information determination unit according to claim 2, further comprising: a characteristic information table storage unit configured to store the characteristic information previously associated with a character code of the character extraction target pattern. 4 or 5
The character extraction device described. 7. The character extracting apparatus according to claim 6, wherein the characteristic information table storage unit stores the character code and the characteristic information of the character code in association with each other. 8. The method according to claim 8, wherein the character feature determination unit acquires the feature information corresponding to the character code output from the target character storage unit from the feature information table storage unit and outputs the acquired feature information to the feature information storage unit. 8. The character extracting device according to claim 6, wherein the character extracting device is a character extracting device. 9. The unique features of the character extraction target pattern include: the presence or absence of left / right separable characters, the presence / absence of vertical characters, the presence / absence of minute characters, the presence / absence of subscripts, and the superscript 9. The character extracting device according to claim 1, wherein the character extracting device indicates at least one of the presence or absence of a character and the presence or absence of a character other than a number. 10. The character extracting unit, when there is no left-right separable character in the characteristic information output by the characteristic information storage unit, the value of a vertical projection distribution in horizontal character extraction is set to a value equal to or less than a set value. Character section, or when there is no upper / lower separable character, all sections where the value of the horizontal projection distribution is equal to or less than the set value in vertical character cutout are cut out as characters. Claims 1, 2, 3, 4, 5, 6, 7,
10. The character extracting device according to 8 or 9. 11. The method according to claim 1, wherein the character information extracting unit outputs, when the feature information output from the feature information storing unit does not include a minute character smaller than a normal character, a pattern whose size is equal to or less than a predetermined threshold. If the subscript is not present, or if there is no subscript, the character with the upper end of the pattern below the predetermined threshold below the line is not extracted as a character, or the superscript does not exist 9. The method according to claim 1, wherein a character having a lower end of the pattern above a predetermined threshold value above the line is not cut out as a character.
Or the character extracting device according to 9. 12. The character extracting unit, when there is no character other than a number in the characteristic information output by the characteristic information storage unit and the interval between characters is equal to or longer than a reference interval, 4. The method according to claim 1, wherein the character space is cut out in accordance with the length of the space.
A character extracting device according to 4, 5, 6, 7, 8 or 9.