JP2581809B2 - Character extraction device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character extracting apparatus for extracting a character from a slanted character line with a ruled line so as not to include the ruled line.

(Prior Art) In recent years, in order to improve the processing efficiency of a large number of documents, there is an increasing demand for mechanically inputting information described in the documents to a computer. The character recognition device used for this mechanical input separates an image of one character from a printed or handwritten character group image in units of one character, and performs character recognition for one character.

Generally used documents are often provided with ruled lines in addition to characters. For character recognition, it is necessary to separate an image for one character so as not to include the ruled lines.

The prior art for this image separation is disclosed in
Japanese Patent No. 160487 discloses an optical character reader, and Japanese Patent Application Laid-Open No. 62-2173.
No. 85, the character separation system can be mentioned.

In the former conventional technique, a row direction and a column direction are virtually set on an image memory storing image data from a reading unit, and image data of a document is scanned in a row direction. A peripheral distribution is created by calculating the number of pixels, an area where the cumulative number of black pixels of the peripheral distribution in the row direction is equal to or greater than a predetermined value is detected as an underline part, and an area where the cumulative black pixel number is less than the predetermined value is detected as a character part. Are detected as the character start and end positions in the column direction. Furthermore, only the character portion is scanned except for the underline portion, and a peripheral distribution is created by calculating the cumulative number of black pixels in the column direction by this scanning, and from the change in the peripheral distribution in the column direction, the character start and end positions in the row direction are determined. To detect.

In this prior art, when a ruled line is added along the row direction, in order to accurately detect the character start and end positions in the column direction, the true row direction of the image matches the predetermined row direction. It is necessary to obtain such an image and create a marginal distribution such that characters and ruled lines do not overlap virtually in a predetermined line direction.

In the latter conventional technique, a plurality of local regions are set on an image including a character line with a straight line such as an underline,
The images of these local regions are symbolized into predetermined direction elements by pattern matching, and the direction elements are further classified. Then, a line segment having a predetermined condition (for example, length or existing area) is selected from line segments composed of direction line elements (for example, horizontal straight line components) having the same direction as a line segment unnecessary for character recognition such as an underline. Is detected, and the detected line segment is removed from the image as an unnecessary line segment. After obtaining an image from which unnecessary line segments have been removed in this way, this image is scanned to create a marginal distribution in the row and column directions, and from the created marginal distribution, the character start and end positions in predetermined row and column directions. Was detected.

(Problems to be Solved by the Invention) However, in the former conventional technique described above, the deviation between the predetermined line direction virtually set in the image memory and the true line direction of the image becomes large, and the characters in the predetermined line direction become large. If the ruled line overlaps or the character and the ruled line are in contact with each other, black pixels of the character and the ruled line are mixed on the same scanning line at the time of creating the marginal distribution, so that the position of the character and the ruled line cannot be detected with high accuracy.

Also, in the former conventional technique, since the character start and end positions in the column direction are the same for all characters on the same character line, when the predetermined line direction is deviated from the true line direction, the character In some cases, incomplete characters that are missing are cut out and characters cannot be recognized. The degree of missing characters becomes worse nearer the end of the line.

In the latter conventional technique described above, the read image is divided into a plurality of local areas, the read image of each local area is compared with a predetermined plurality of basic patterns by matching, and the local area is oriented based on the comparison result. Convert to line elements. Such a local matching process is complicated, and therefore, the configuration of the device employing the conventional technique is complicated, which causes a problem that the size of the device is increased and the cost of the device is increased.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-described conventional problems, and to accurately remove characters based on a change in peripheral distribution even when a predetermined line direction is deviated from a true line direction. An object of the present invention is to provide a character cutout position that can detect the cutout position of a character.

(Means for Solving the Problems) In order to achieve this object, a character cutout device according to the present invention includes: an image output unit that outputs image data of a character line with a ruled line; and an image storage that stores image data. And the image data of the character detection area set for each character of the character row is scanned in the column direction, the cumulative number of black pixels is detected for each scanning line, and the first peripheral distribution is created. A character interval detection unit that detects the position of a character in a character detection region from a change in the peripheral distribution, sets the position of a ruled line detection region adjacent to the character based on the position of the character, and sets the character cutout position in the line direction. Scanning the image data of the ruled line detection area in the row direction, detecting the cumulative number of black pixels for each scanning line, creating a second peripheral distribution, and detecting the position of the ruled line from the change in the second peripheral distribution. , Character and ruled line separation position based on the position of the ruled line A character rule separation unit that sets and sets the character cutout position in the column direction based on the separation position, and a position correction in the column direction for setting the character detection area and rule detection area for the second and subsequent characters in the character line It is characterized by comprising a correction value determining unit for setting a value based on the position of a character cut out before that, and a control unit for outputting a character cutout position for each character.

(Function) According to the character extracting device having such a configuration, the position correction value in the column direction for setting the character detection area and the ruled line detection area of the second and subsequent characters of the character line is extracted before the character. Set based on the position of.

The position correction value indicates the amount of displacement or the range of displacement of the character and the ruled line in the column direction. Therefore, even if the character line is inclined, the character detection area is detected so as to include the character to be cut out. The area can be set with high accuracy, and the ruled line detection area can be set with high accuracy so as to include the ruled line adjacent to the character to be cut out.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. It should be noted that the drawings are only schematically shown to the extent that the present invention can be understood. It is not limited to.

First Embodiment FIG. 1 is a functional block diagram schematically showing a configuration of a first embodiment of the present invention.

The character extracting apparatus according to this embodiment includes an image output unit 10 that outputs image data of a character line with a ruled line, an image storage unit 12 that stores image data, and a character set for each character in the character line. Scanning the image data of the detection area in the column direction, detecting the cumulative number of black pixels for each scanning line, creating the first peripheral distribution, and detecting the position of the character in the character detection area from the change in the first peripheral distribution Then, based on the position of the character, the position of the ruled line detection area adjacent to the character is set, and the character interval detection unit 14 that sets the character cutout position in the line direction is scanned. Detecting the cumulative number of black pixels for each scanning line, creating a second peripheral distribution, detecting the position of the ruled line from the change in the second peripheral distribution, setting the separation position of the character and the ruled line based on the position of the ruled line, Character that sets character extraction position in column direction based on separation position A line separation unit 16 and a correction value for setting a position correction value in the column direction for setting a character detection area and a ruled line detection area of the second and subsequent characters of a character line based on the position of a character cut out earlier. It comprises a determination unit 17 and a control unit 18 that outputs a character cutout position for each character.

 Hereinafter, this embodiment will be described in more detail.

(Image Output Unit) The image output unit 10 of this embodiment scans a recording medium such as an original on which a character line with a ruled line is recorded, and converts the reflected light P from this medium into a black and white binary quantized electric signal. Signal (image data)
And outputs the image data of the character line with the ruled line in pixel units.

(Image Storage Unit) In this embodiment, the image storage unit 12 is configured using an image memory that stores image data from the image output unit 10. An XY coordinate system is set on this memory. For example, assuming a horizontally written character line, the X-axis direction is virtually set as a row direction and the Y-axis direction is set as a column direction.

In order to extract characters without missing characters, it is most preferable to match the true line direction of the image with the X-axis direction. However, even if the true line direction is deviated from the X-axis direction, missing characters may occur. Can be cut out without reducing the number of characters as compared with the related art or without missing characters.

(Correction Value Determining Unit) The correction value deciding unit 17 of this embodiment inputs, from a line position information detecting unit (not shown), line position information indicating a character detection area position and a ruled line detection area position in the column direction of the first character.

The line position information detection unit sets a region (may or may not include the first character) near the first character expected by the format information, scans this region, and scans the cumulative black on the scanning line in the line direction. The number of pixels is obtained for each scanning line, and the position in the column direction of the region where the scanning line having the cumulative number of black pixels equal to or greater than the predetermined threshold value is detected as the column direction position of the ruled line in the vicinity of the first character. Based on the ruled line position, the line position information detecting unit sets the line position information indicating the position of the character detection area in the column direction of the first character so that the character detection area includes the first character and detects the ruled line in the column direction. The line position information indicating the region position is set so that the ruled line detection region includes a ruled line adjacent to the first character.

Further, the correction value determination unit 17 obtains a position correction value in the column direction for setting the character detection area and the ruled line detection area of the second and subsequent characters based on the positions of the characters cut out earlier.

In this embodiment, the character extraction position is displaced in the column direction from the character detection area position using the character detection area position in the column direction of the first character and the character extraction area position in the column direction of the character that has been cut out after the first character. Then, a position correction value is obtained from the displacement amount.

Then, the correction value determination unit 17 uses the position correction value and the character detection area position and ruled line detection area position in the column direction of the first character in the column direction of the next character to be cut out (hereinafter also referred to as the next character). The character detection area position and the ruled line detection area position are set. The detection area position of the next character in the column direction is set so that the character detection area includes the next character, and the ruled line detection area position of the next character in the column direction is a ruled line where the ruled line detection area is adjacent to the next character. Is set to include

(Character Detector) The character detector 14 of this embodiment is a first marginal distribution generator.
20, a first peripheral distribution storage unit 22, a black block detection unit 24, and a ruled line detection area position determination unit 26.

The first marginal distribution creating unit 20 sets the character detection area in the line direction so that the character detection area includes the cutout character and at least a part of the character to be cut out after the character based on the character position in the line direction. Set the position.

For the first character of the character line,
The position of the character detection area in the line direction is set based on the position in the line direction of the first character detected by 18.

For the characters following the first character, if the character detection area is set as described above, the black block detection unit 24 detects not only the position of the cut character in the line direction but also the position of the next character in the line direction. The character detection area position in the line direction is set based on the position of the character in the line direction.

Then, the first marginal distribution creating unit 20 scans the image data of the character detection area in the column direction, detects the cumulative number of black pixels for each scanning line, and generates a column including the cumulative number of black pixels for each scanning line. Create a marginal distribution in the direction (first marginal distribution).

The first peripheral distribution storage unit 22 stores the cumulative number of black pixels of the first peripheral distribution for each scanning position.

The black block detection unit 24 detects the character start and end positions in the row direction from the change in the cumulative number of black pixels for each scanning position in the first peripheral distribution, and the ruled line detection area position determination unit 26 determines the character start and end in the line direction. Based on the end position, the start and end positions of the ruled line detection area in the row direction are set so that the ruled line detection area includes the ruled line adjacent to the cutout character.

(Character Ruled Line Separation Unit) The character ruled line separation unit 16 of this embodiment includes a second marginal distribution creation unit 28, a second marginal distribution storage unit 30, a ruled line detection unit 32, and a ruled line separation position determination unit 34.

The second peripheral distribution creating unit 28 scans the image data of the ruled line detection area in the row direction, detects the cumulative number of black pixels for each scanning line, and detects the cumulative number of black pixels for each scanning line. Is created (second marginal distribution).

The second peripheral distribution storage unit 30 stores the cumulative number of black pixels of the second peripheral distribution for each scanning position.

The ruled line detection unit 32 detects the start and end positions of the ruled line in the column direction based on the change in the cumulative number of black pixels for each scanning position of the second peripheral distribution, and the ruled line separation position determination unit 34 starts and ends the ruled line in the column direction. A separation position for separating characters and ruled lines in the column direction is set based on the position, and a character cutout position in the column direction is determined from the separation position.

(Control Unit) The control unit 18 of this embodiment sets the start position of the temporary character detection area in the line direction of the first character so as to include the first character and the margin near the first character. Then, the image data in the character detection area is line-sequentially scanned from the start end position, the number of cumulative black pixels on the scanning line in the column direction is obtained for each scanning line, and the number of cumulative black pixels exceeding a predetermined threshold is detected. The position of the line in the line direction is detected as the start position of the first character.

Next, in order to deepen the understanding of the present invention, an operation flow of this embodiment will be described with an example. First, FIGS. 2 to 3
This will be described with reference to the drawings.

FIG. 2 is an operation flow chart showing an example of the operation flow of this embodiment. FIGS. 3 (A) and 4 (A) are examples of image data when one ruled line exists above and below horizontal writing characters. FIGS. 3 (B) and 4 (B) show the first example. And the second
It is a figure showing the example of the circumference distribution of the character detection field of the character of the character. In FIGS. 3 and 4A, the X and Y axes are coordinate axes set on the image memory, and the X axis direction is virtually the row direction and the Y axis direction is the column direction. The horizontal axis in FIGS. 3 and 4B is the horizontal axis in FIGS. 3 and 4A.
The coordinate axis corresponding to the X-axis and the vertical axis represent the cumulative number of black pixels CH (X) on the scanning line at the sub-scanning position X.

The cutout apparatus of this embodiment starts operation when a processing start signal is input. (START).

When the operation is started, the image output unit 10 scans a predetermined reading range A of a document or the like, and stores monochrome binary image data in the image storage unit.
12 outputs. On the other hand, the control unit 18 inputs line position information indicating the position of the character detection area and the ruled line detection area in the column direction of the first character from a line position information detection unit (not shown). (S1) Here, with reference to FIG. 3, the operation of the row position information detecting unit (not shown) will be described.

The line position information detector detects the start position in the line direction for the vicinity of the first character based on the format information such as the character description range.
X _a and terminal position X _b and column direction starting end position Y _a and terminal position Y _b
And. It is sufficient that the vicinity area mouth includes the ruled line near the first character, and may include the first character in addition to the ruled line near the first character. In the illustrated example, the vicinity area opening when only the ruled line is included is shown.

Next, the row position information detection unit scans the inside of the neighboring area with the row direction as the main scanning direction, finds the number of cumulative black pixels on the scanning line at each sub-scanning position Y, and determines the area where the cumulative number of black pixels is equal to or greater than a predetermined threshold. LYT1, LYB1 and positions LYT2, LYB2 in the column direction are detected. LYT1 and LYT2 represent the position of the upper ruled line near the first character in the column direction, and LYB1 and LYB2 represent the position of the lower ruled line near the first character in the column direction.

Then, the line position information detecting section includes line position information LJT2 and LJB2 representing the character detection region positions YT2 (1) and YB2 (1) in the column direction of the first character, and the ruled line detection region position YT1 in the column direction of the first character. (1) and row position information LJT1 representing YB1 (1) and
Find LJB1. The row position information LJT2, LJB2, LJT1, and LJB1 are defined by, for example, the following equation (1).

LJT2 = LYT2 + α1 LJB2 = LYB2-α1 LJT1 = LYT1-α2 LJB1 = LYB1 + α2 (1) α1 in the expression (1) is a character to be cut out even if the character line is inclined within a predetermined inclination range. Is a constant value set so that the area LJT2 ≦ Y ≦ LJB2 does not include the ruled line attached to the first character and at least a part of the ruled line attached to the character following the character. Set to 12.

Α2 in the expression (1) is a ruled line attached to the first character to be a cut-out character and a character following the character even if the character line is inclined within the range of the predetermined amount of inclination. A constant value is set so that at least a part of the ruled line is included in the area LJT1 ≦ Y ≦ LJB1, for example, α2 = 12.

Subsequent to S1, the control unit 18 sets information “0” indicating the incomplete extraction processing in the extraction end flag (S2).

Next, the control unit 18 determines whether or not the extraction process has been completed with reference to the extraction end flag (S3).

Here, for example, “1” and “0” are used as information indicating the end and the non-end of the extraction process. The extraction device of this embodiment ends the operation (END) if the extraction end flag = 1 in S3, and performs the following processing S4 if the extraction end flag = 0.

In S4, the correction value determination unit 17 determines the n-th character (n is a natural number)
Character detection area start end position YT2 (n) in column direction for each character
And the end position YB2 (n) and the start position YT1 (n) and end position YB1 (n) of the ruled line detection area in the column direction.

In the case of S4a: n = 1, the correction value determination unit 17 sends the row position information LYT1, LY
T2, LYB1 and LYB2 are input, and the position YT2 (1) and YB (1) of the character detection area in the column direction and the position YT1 (1) and YB1 (1) of the ruled line detection area in the column direction are input for the first character. , Set in accordance with the following equation (2), and these positions YT2 (1), YB2
(1) Outputs YT1 (1) and YB1 (1) to the control unit 18.

YT2 (1) = LJT2 YB2 (1) = LJB2 YT1 (1) = LJT1 YB1 (1) = LJB1 (2) In the case of S4b: n ≠ 1, the correction value determination unit 17 determines the first character and the subsequent characters. n-th (n ≧
For the character 2), for example, positions YT2 (n), YB2 (n), YT1 (n) and YB1 (n) defined by the following equation (3).
Are set, and these positions are output to the control unit 18.

YT2 (n) = YT2 (1) + HENI YB2 (n) = YB2 (1) + HENI YT1 (n) = YT1 (1) + HENI YB1 (n) = YB1 (1) + HENI (3) where HENI is corrected This is the position correction value obtained by the value determination unit 17 in S12 described below.

Subsequent to S4, the character cutout apparatus of this embodiment performs a process for setting the start position HS (n) and the end position HE (n) of the character detection area in the line direction (S5).

S5a: n = 1 First, the control unit 18 sets a temporary character detection area start end position in the line direction of the first character. Starting end position of the temporary, as the character detection region encompasses the first character and a blank area in the first character vicinity on the basis of the format information such as character stated range, is set, for example, the start position of the temporary and X _b I do.

The control unit 18 scans the first character detection area while sequentially moving the sub-scanning position X from the tentative start position to the character side, and obtains the cumulative number of black pixels on the scanning line in the column direction for each scanning line. . And detecting a sub-scanning position X of scanning lines cumulative number of black pixels exceeds a predetermined threshold value as the first character in the row direction of the starting end position B _ls.

Next, the first marginal distribution creating unit 20 detects the detected start position B
Based on _ls , for example, a character detection area start position HS (1) and end position HE (1) in the line direction of the first character defined by the following equation (4) are set.

HS (1) = B _1S −β HE (1) = HS (1) + HX (4) In the case of S5b: n ≠ 1, the first marginal distribution creating unit 20 outputs the n-th character to be extracted Based on the starting position B _ns in the row direction, the starting position HS (n) and the ending position in the row direction defined by the following equation (5), for example.
Set HE (n).

HS (n) = B _ns −β HE (1) = HS (n) + HX (5) In Expressions (4) and (5), β and HX are constant values that are arbitrarily and suitably set. In the embodiment, β and HX are set such that the character detection region of the nth character includes at least a part of the character of the nth character and at least a part of the character of the (n + 1) th character.
It was set. In setting β and HX, β and HX may be set so that the character detection area of the nth character includes at least the nth character and a blank area near the character. For example, when the maximum width of the character width is 4 mm and the resolution of the image sensor is 16 pixels / mm, β = 32 and H
X may be set to 128.

Subsequent to S5, the first marginal distribution creating unit 20 outputs the character line "ABCD"
It is determined whether or not character extraction from is completed (S6).

For example, when the position HE (n) is larger or smaller than the sum of the position _Xb and the length in the line direction of the character description surface defined by the format information, it is regarded that the character extraction is completed or not completed.

If the character extraction is completed, the first marginal distribution creating unit 20 rewrites the processing end flag held in the register to 1 (S
7) Then, the first marginal distribution creation unit 20 and the black block detection unit 24 perform processing for detecting the position of the character in the character detection area in the line direction (S8). If character extraction has not been completed, the process of S8 is performed after S6.

In S8, the first marginal distribution creating unit 20 sets the HS (n) and HE (n) set by the creating unit 20 and the YT2 from the control unit 18.
(N), based on YB2 (n), the character detection area of the n-th character (HS (n) ≦ X ≦ HE (n) and YT2 (n) ≦ Y ≦ Y
The image data in the area of B2 (n) is scanned to create a first marginal distribution. In the creation of this peripheral distribution, the cumulative number of black pixels CH (X) on the scanning line obtained for each peripheral sub-scanning position X
Is stored in the first marginal distribution storage unit 22. Cumulative black pixel count CH
(X) is represented by the following equation (6).

However, when the image data of the pixel at the position of the coordinates (X, Y) is a black pixel, P (X, Y) = 1, and when it is a white pixel, P (X, Y)
(X, Y) = 0.

After completing the creation of the first marginal distribution in S8, next, the black block detection unit 24 calculates the cumulative number of black pixels in the first marginal distribution storage unit 22.
CH (X) is read out and compared with a threshold value THL B for each sub-scanning position X, for example, THL B = 1, and CH (X) satisfying CH (X) ≧ THL B
The start position and end position in the line direction of the area where is present are the character start position B _ns and end position B in the line direction of the n-th character.
_Detected as _nE .

A scan line satisfying CH (X) ≧ THLB is defined as a black scan line and CH
If the scanning line satisfying (X) <THLB is called a white scanning line, the sub-scanning position X of the black scanning line when the black scanning line follows the white scanning line is the character start position in the row direction. The sub-scanning position X of the black scanning line when the white scanning line follows the B _ns and the black scanning line may be _{set as} the character end position B _{nE in the} row direction.

In this embodiment, the black block detection unit 24 determines the character start position B _ns and the end position B _nE of the n-th character, and the character start position B _{(n + 1) s in} the line direction of the next (n + 1) -th character. And outputs to the control unit 18 a character start position B _{(n + 1) s} used for setting a character detection area of the (n + 1) th character to be cut out next.

Then, the ruled-line detection area position determining unit 26 on the basis of the n-th character starting position B _ns and end position B _nE, determines the character extraction position of the n th character, it outputs the extraction position to the control unit 18 I do. For example, a row direction of the character start position B _ns and end position B _nE character in the line direction character segmentation start position KXL
(N) and the end position KXR (n).

Subsequent to S8, the ruled line detection area position determination unit 26 receives the character start end position B _ns and the end position B _{ns in} the row direction from the black block detection unit 24.
_Based on _nE , the ruled line detection area position in the row direction is set to include the ruled line adjacent to the n-th character (S
9) The position of the ruled line detection area in the row direction is output to the control unit 18.

In this embodiment, to set the second ruled line detection area adjacent to the first ruled line detection region and the end position B _nE adjacent to the n-th character of the starting end position B _ns.

Start position BX in the row direction of the first ruled line detection area of the n-th character
LL (n) and end position BXLR (n), and start and end positions BXRL (n) and BXRR (n) in the row direction of the second ruled line detection area.
Are defined, for example, as shown in the following equation (7).

When n = 1 or KXL (n) -1-β> KXR (n-1) +1 BXLL (n) = KXL (n) -1-βn ≧ 2 and KXL (n) -1-β ≦ KXR ( When n−1) +1 BXLL (n) = KXR (n−1) +1 BXLR (n) = KXL (n) −1 BXRL (n) = KXR (n) +1 n = n _L (n _L is a character line Number of characters in one line) or KXR
(N) + 1 + β <KXL (n + 1) −1 BXRR (n) = KXR (n) + 1 + β n ≦ n _L −1 and KXR (n) + 1 + β ≧ KXL (n + 1) −1
BXRR (n) = KXL (n + 1) -1 (7) Next, a description will be given with reference to FIG. 2 and FIG. FIG. 5A is a diagram showing image data of almost one character of the n-th character, and FIGS. 5B and 5C are diagrams showing peripheral distribution of the ruled line detection area of the n-th character. In FIG. 5A, the X and Y axes are coordinate axes set on the image memory,
The vertical axes in FIGS. (B) and (C) are the coordinate axes corresponding to the Y axis in FIG. 5 (A) and the horizontal axes are the cumulative number of black pixels BLH (Y) and BRH (Y) on the scanning line at the sub-scanning position Y. Represents

Subsequent to S9, the second margin distribution creating unit 28 and the ruled line detection unit 32 perform processing for detecting the position of the ruled line in the column direction in the first and second ruled line detection areas (S10).

In S10, the second marginal distribution creating unit 28 determines the positions BXLL (n), BXLR (n), and BXLL (n) of the first ruled line detection area 2 input from the control unit 18.
YT1 (n) and YB1 (n) and the position BX of the second ruled line detection area E
Based on RL (n), BXRR (n), YT1 (n) and YB1 (n), the first ruled line detection area d (BXLL (n) of the n-th character
≤X≤BXLR (n) and YT1 (n) ≤Y≤YB1 (n)) and the second ruled line detection region e (BXRL (n) ≤X≤BXRR
(N) and the area of YT1 (n) ≦ Y ≦ YB1 (n) is scanned,
The cumulative number of black pixels BLH (Y) and BRH (Y) on the scanning line at each sub-scanning position Y is obtained, and the cumulative black pixel number B for each sub-scanning position Y is calculated.
LH (Y) and BRH (Y) are stored in the second marginal distribution storage unit 30, and a second marginal distribution is created.

The cumulative number of black pixels BLH (Y) and BRH (Y) is defined, for example, by the following equation (8).

Next, the ruled line detection unit 32 detects a ruled line position on a side closer to the character of the ruled line adjacent to the n-th character. The ruled line detection unit 32 stores the cumulative black pixel number BLH in the second marginal distribution storage unit 30.
(Y) is read out from YT2 (n) in the order of the sub-scanning position Y, and compared with the threshold value THLL for each sub-scanning position Y.
(Y) The sub-scanning position Y where ≧ THLL is set to the ruled line position YL in the column direction
Detected as T. In addition, the cumulative number of black pixels BLH (Y) is YB2
From (n), the sub-scanning position Y is read in ascending order and compared with the threshold value THLL for each sub-scanning position Y. First, BLH (Y) ≧ THLL
Is detected as the ruled line position YLB in the column direction. For the first ruled line detection area, these ruled line positions YLT
And YLB are detected.

The threshold value THLL is a constant that is arbitrarily and suitably set, and is defined, for example, by the following equation (9).

THLL = {(BXLR (n) -BXLL (n) +1) / 2} +1 ...
(9) At the same time, the ruled line detection unit 32 reads the cumulative black pixel number BRH (Y) of the second peripheral distribution storage unit 30 from YT2 (n) in descending order of the sub-scanning position Y, and sets a threshold value THLR for each sub-scanning position Y. First, the sub-scanning position Y where BRH (Y) ≧ THLR is detected as the ruled line position YRT in the column direction. Also, the cumulative number of black pixels BRH
(Y) is read from YB2 (n) in ascending order of the sub-scanning position Y, and is compared with the threshold value THLR for each sub-scanning position Y.
(Y) The sub-scanning position Y where ≧ THLR is set to the ruled line position YR in the column direction
Detect as B. As for the second ruled line detection area, these ruled line positions YRT and YRB are detected.

The threshold value THLR is a constant that is arbitrarily and suitably set, and is defined, for example, by the following equation (10).

THLR = {(BXRR (n) -BRRL (n) +1) / 2} +1 ...
(10) Subsequent to S10, the ruled line separation position determination unit 34 separates the nth character from the ruled line adjacent to the nth character in the column direction, and separates the nth character YT (n), YB. (N) is set (S11).

The separation positions YT (n) and YB (n) are defined, for example, by the following equation (11).

When YLT ≧ YRT YT (n) = YLT When YLT ≧ YRT YT (n) = YRT When YLB ≧ YRB YB (n) = YRB When YLB <YRB YB (n) = YLB ... (11) and The ruled line separation position determination unit 34 sets a character extraction start position and an end position in the column direction based on these separation positions, and outputs these extraction positions to the control unit 18. For example, the separation positions YT (n) and YB (n) are set as the character extraction start position and end position in the column direction, and the control unit 18 determines the extraction position KXL (n), KXR ( n)
And YT (n) and YB (n) are output to the next-stage device (not shown).

After S11, the correction value determination unit 17 obtains a position correction value (S1
2).

The correction value determination unit 17 obtains a position correction value HENI required to set the positions of the character detection area and the ruled line detection area in S4 for the n-th character after the first character, for example, as described below. The (n-1) th character extraction start position YT (n-1) and end position YB (n-1) in the column direction are input from the ruled line separation position determination unit 34, and these extraction positions YT (n
-1), YB (n-1) and the previously set character detection area positions YT2 (1) and YB2 (1) in the column direction of the first character are used to calculate HENI defined by the following equation (12). Ask. After S12, the process of S3 is performed.

| YT (n-1) -YT2 (1) | ≥ | YB (n-1) -YB2
(1) | HENI = YB (n-1) -YB2 (1) | YT (n-1) -YT2 (1) | <| YB (n-1) -YB2
(1) When | HENI = YT (n-1) -YT2 (1) (12) As shown in Expression (3), the character detection area position YT2 (1) in the column direction of the first character is obtained. YB2 (1) and ruled line detection area position
YT1 (1) and YB1 (1) are corrected by the position correction value HENI. These corrected positions are the character detection area positions YT2 (n) and YB2 (n) and the ruled line detection area position YT1 in the column direction of the n-th character.
(N) and YB1 (n) are input to the control unit 18.

The position of the character detection area and the position of the ruled line detection area in the column direction are corrected in the column direction for each character.

Second Embodiment FIG. 6 is a functional block diagram schematically showing a configuration of a second embodiment of the present invention. The same components as those of the first embodiment are denoted by the same reference numerals.

The character extracting apparatus according to this embodiment includes an image output unit 10, an image storage unit 12, a character interval detection unit 14, a character ruled line separation unit 16, a correction value determination unit 36, and a control unit 18.

The second embodiment is the same as the first embodiment except that the configuration and operation of the correction value determining unit 36 are different from those of the first embodiment. Hereinafter, points different from the first embodiment will be mainly described, and detailed description of the same points as the first embodiment will be omitted.

 Hereinafter, this embodiment will be described in more detail.

(Correction Value Determining Unit) The correction value deciding unit 36 of this embodiment is a gradient calculating unit that calculates the inclination of a character line for each character up to the (N-1) th character.
36a, and for the characters before the (N-1) th character, the position correction value of the character whose inclination is obtained is obtained using the inclination of the character line obtained for each character, and the characters after the Nth character And a correction value calculation unit 36b for calculating a position correction value using the average value (average inclination) of the inclinations of the character lines of the characters up to the (N-1) th character.

Note that the correction value calculation unit 36b sets the character detection area position and ruled line detection area position of the first character in the column direction, and the character detection area position and ruled line detection area position of the next character to be cut out in the column direction.

In this embodiment, the distance a in the line direction from the arbitrarily set line direction reference position to the center position of the character obtained based on the already detected line direction character cutout position is arbitrarily set. From the reference position in the column direction to the ruled line position in the column direction or the distance b in the column direction that has already been detected, the inclination of the character line is determined, and the average inclination is determined from the inclination. Further, a distance c in the line direction from a reference position in the line direction set arbitrarily and suitably to a center position in the line direction of the next character to be cut out obtained using format information such as a character width and a character pitch; Using the degrees, predict the displacement or range of displacement in the column direction of the character to be extracted next,
Further, a displacement or a range of displacement of the ruled line in the column direction is predicted, and a character detection area position and a ruled line detection area position in the column direction are set based on the displacement or the range of displacement.

For example, the above-described distances a and c are obtained by setting the reference position in the row direction as the center position of the first character vicinity area in the row direction. Further, the reference position in the column direction is set as the row position information LYT2 or LYB2, and the above distance b is obtained.

(Character space detecting unit) The character space detecting unit 14 of this embodiment calculates the character center position in the line direction for obtaining the distance a from the character cutout position in the line direction.

Next, in order to deepen the understanding of the present invention, an operation flow of this embodiment will be described by way of an example.

First, a description will be given with reference to FIGS. 7 and 3. FIG. 7 is an operation flow chart showing an example of the operation flow of this embodiment.

The cutout apparatus of this embodiment starts operation when a processing start signal is input. (START).

When the operation is started, the image output unit 10 scans a predetermined reading range A of a document or the like, and stores monochrome binary image data in the image storage unit.
Output to 12. On the other hand, the control unit 18 has line position information LJT2 and LJB2 representing the character detection region position of the first character in the column direction, and line position information LJT representing the ruled line detection region position of the first character in the column direction.
T1 and LJB1, line position information LYT2 and LYB2 indicating a ruled line position near the first character, and a line direction reference position XKC for obtaining the distance a are input from a line position information detection unit (not shown). (* S1) The row position information detector detects the row position information LJT2, LJB2, LJT1,
LJB1, LYT2 and LYB2 are detected in the same manner as in the first embodiment. The row position information detection unit sets a center position in the X direction of the ruled line at the reference position XKC and to for example, an area X _a ≦ X ≦ X _b for determining the distance a above. This reference position XKC
Is represented by the following equation.

XKC = X _a + (X _b −X _a ) / 2 (13) After * S 1, the control unit 18 sets information “0” indicating the end of the extraction processing in the extraction end flag (* S 2). .

Next, the control unit 18 sets 1 to a character number counter that counts the number of characters for which a character whose position is to be extracted is detected in the character line where character extraction is performed (* S3).

Next, the correction value calculation unit 36b calculates the start position YT2 (n) and the end position YB2 of the character detection area in the column direction for the nth character.
(N) and the ruled line detection area start end position YT1 (n) and end position YB1 (n) in the column direction are set (* S4).

• * S4a: When n = 1 The correction value calculator 36b sends the row position information LJT2, LJ
Input JB2, LJT1 and LJB1, and set the character detection area position YT2 (1), YB2 (1) and the ruled line detection area position YT1 (1), YB1 (1) of the first character according to the following equation (2). I do.

YT2 (1) = LJT2 YB2 (1) = LJB2 YT1 (1) = LJT1 YB1 (1) = LJB1 (2) ** S4b: n ≠ 1 When the correction value calculation unit 36b is used, The n-th (n
For the characters of ≧ 2), for example, character detection area positions YT2 (n) and YB2 (n) and ruled line detection area positions YT1 (n) and YB1 (n) defined by the following equation (3) are set.

YT2 (n) = YT2 (n-1) + HENI YB2 (n) = YB2 (n-1) + HENI YT1 (n) = YT1 (n-1) + HENI YB1 (n) = YB1 (n-1) + HENI ... (3) However, HENI is calculated by the correction value calculation unit 36b in * S19 or * S2 described later.
This is the position correction value obtained with 0. Using the position correction value HENI, the positions YT2 (n), YB2 (n), YT1 (n) and YB1 (n) in the column direction are corrected for each character according to the inclination of the character line.

Subsequent to * S4, the character cutout apparatus of this embodiment performs processing for setting the start position HS (n) and the end position HE (n) of the character detection area in the line direction (* S5).

* When S5a: n = 1 First, the control unit 18 sets a temporary starting position in the line direction for the character detection area of the first character, and sequentially moves the sub-scanning position X from the temporary starting position to the character side. While scanning, the first character detection area is scanned, and the cumulative number of black pixels on the scanning line in the column direction is obtained for each scanning line. And detecting a sub-scanning position X of scanning lines cumulative number of black pixels exceeds a predetermined threshold value as the first character in the row direction of the starting end position B _ls.

Next, the first marginal distribution creating unit 20 calculates the detected start position
Based on B _ls , for example, the start position HS (1) and the end position HE (1) in the line direction of the first character defined by the following expression (4) are set, and these positions HS (1), HE (1)
Is input to the control unit 18.

HS (1) = B _ls −β HE (1) = HS (1) + HX (4) In the case of * S5b: n ≠ 1, the first marginal distribution creating unit 20 attempts to cut out the n-th character based on the starting end position B _ns of the row direction, for example, the following equation (5) the row direction of the starting end position HS, defined by (n) and the end position HE
(N) is set, and these positions HS (n) and HE (n) are input to the control unit 18.

HS (n) = B _ns −β HE (n) = HS (n) + HX (5) * After S5, the first marginal distribution creating unit 20 sets the character line “ABC
D ”or not (* S
6).

If the character extraction is completed, the first marginal distribution creation unit 20 rewrites the end of the processing held in the register to 1 (* S7), and then the first marginal distribution creation unit 20 and the black block detection unit 24 change the character detection area. A process for detecting the position of the character in the line direction is performed (* S8). If the character extraction has not been completed, * S6 is followed by * S8.

* In S8, the first marginal distribution creating unit 20 sets the character detection area C of the n-th character (HS (n) ≦ X ≦ HE (n) and YT2 (n) ≦
The image data of Y ≦ YB2 (n) is scanned to create a first marginal distribution. In the generation of this peripheral distribution, the cumulative number of black pixels CH on the scanning line obtained for each peripheral sub-scanning position X
(X) is stored in the first marginal distribution storage unit 22. The cumulative number of black pixels CH (X) is represented by the following equation (6).

After completing the creation of the first marginal distribution, the black block detection unit 24 reads the cumulative number of black pixels CH (X) in the first marginal distribution storage unit 22 and compares it with the threshold value THLB for each sub-scanning position X. CH
(X) ≧ THL B become CH (X) detects the starting end position and end position in the row direction of the area which exists as a character start position B _ns and end position B _nE of the n-th character in the row direction.

In this embodiment, the black block detection unit 24 determines the character start position B _ns and the end position B _nE of the n-th character, and the character start position B _{(n + 1) s in} the line direction of the next (n + 1) -th character. And outputs to the control unit 18 a character start position B _{(n + 1) s} used for setting a character detection area of the (n + 1) th character to be cut out next.

Then, the ruled-line detection area position determining unit 26 on the basis of the n-th character starting position B _ns and end position B _nE, determines the character extraction position of the n-th character, and outputs the extraction position to the control unit 18 . For example, a row direction of the character start position B _ns and end position B _nE of the first character in the row direction character segmentation start position KXL
(N) and the end position KXR (n).

* After S8, the ruled line detection area position determination unit 26 determines the line direction so as to include the ruled line adjacent to the n-th character based on the character start position B _{ns and} end position B _nE in the line direction. The ruled line detection area position is set (* S9).

In this embodiment, to set the second ruled line detection area adjacent to the first ruled line detection region and the end position B _nE adjacent to the n-th character of the starting end position B _ns.

Start position BX in the row direction of the first ruled line detection area of the n-th character
LL (n) and end position BXLR (n), and start and end positions BXRL (n) and BXRR (n) in the row direction of the second ruled line detection area.
Are calculated according to, for example, the following equation (7).
(N), BXLR (n), BXRL (n) and BXRR (n) are output to the control unit 18.

When n = 1 or KXL (n) -1-β> KXR (n-1) +1 BXLL (n) = KXL (n) -1-βn ≧ 2 and KXL (n) -1-β ≦ KXR ( When n−1) +1 BXLL (n) = KXR (n−1) +1 BXLR (n) = KXL (n) −1 BXRL (n) = KXR (n) +1 n = n _L (n _L is a character line BXRR (n) = KXR (n) + 1 + βn ≦ n _L −1 and KXR (n) + 1 + β ≧ KXL (n + 1) −1 when KXR (n) + 1 + β <KXL (n + 1) −1
BXRR (n) = KXL (n + 1) -1 (7) Furthermore, the ruled line detection area position determination unit 26 sets the ruled line detection area position in the row direction and sets the distance a for the nth character. Character center position XMC in line direction to find
(N) is calculated, and the position XMC (n) is output to the control unit 18. The center position XMC (n) is represented by the following equation (14).

XMC (n) = KXL (n) + (KXR (n) -KXL (n)) / 2 ...
(14) Next, a description will be given with reference to FIGS. 7 and 5.

* After S9, the second marginal distribution creating unit 28 and the ruled line detecting unit 32
Performs processing for detecting the position of the ruled line in the column direction within the first and second ruled line detection areas (* S10).

* In S10, the second marginal distribution creating unit 28 inputs the positions of the first ruled line detection area D and the second ruled line detection area E of the n-th character from the control unit 18 and scans these ruled line detection areas D and E. And the cumulative number of black pixels BLH on the scanning line at each sub-scanning position Y
(Y) and BRH (Y) are obtained, and the cumulative number of black pixels BLH (Y) and BRH (Y) are stored in the second peripheral distribution storage unit for each sub-scanning position Y.
Stored at 30 to create a second marginal distribution.

The cumulative number of black pixels BLH (Y) and BRH (Y) is defined, for example, by the following equation (8).

At S10, the position of the ruled line adjacent to the character of the ruled line adjacent to the n-th character is detected. For this reason, the ruled line detection unit 32 calculates the cumulative black pixel number BLH in the second marginal distribution storage unit 30.
(Y) is read out from YT2 (n) in the order of the sub-scanning position Y, and compared with the threshold value THLL for each sub-scanning position Y.
(Y) The sub-scanning position Y where ≧ THLL is set to the ruled line position YL in the column direction
Detected as T. In addition, the cumulative number of black pixels BLH (Y) is YB2
From (n), the sub-scanning position Y is read in ascending order and compared with the threshold value THLL for each sub-scanning position Y. First, BLH (Y) ≧ THLL
Is detected as the ruled line position YLB in the column direction. For the first ruled line detection area, these ruled line positions YLT
And YLB are detected.

The threshold value THLL is a constant that is arbitrarily and suitably set, and is defined, for example, by the following equation (9).

THLL = {(BXLR (n) -BXLL (n) +1) / 2} +1 ...
(9) At the same time, the ruled line detection unit 32 reads the cumulative black pixel number BRH (Y) of the second peripheral distribution storage unit 39 from YT2 (n) in descending order of the sub-scanning position Y, and sets a threshold value THLR for each sub-scanning position Y. First, the sub-scanning position Y where BRH (Y) ≧ THLR is detected as the ruled line position YRT in the column direction. Also, the cumulative number of black pixels BRH
(Y) is read from YB2 (n) in ascending order of the sub-scanning position Y, and is compared with the threshold value THLR for each sub-scanning position Y.
(Y) The sub-scanning position Y where ≧ THLR is set to the ruled line position YR in the column direction
Detect as B. As for the second ruled line detection area, these ruled line positions YRT and YRB are detected.

The threshold THLR is a constant that is arbitrarily and suitably set, and is defined by, for example, the following equation (10).

THLR = {(BXRR (n) -BXRL (n) +1) / 2} + ......
(10) After * S10, the ruled line separation position determination unit 34 separates the n-th character from the ruled line adjacent to the n-th character in the column direction by a separation position YT (n) of the n-th character. Set YB (n) (* S11).

The separation positions YT (n) and YB (n) are defined, for example, by the following equation (11).

When YLT ≧ YRT YT (n) = YLT When YLT <YRT YT (n) = YRT When YLB ≧ YRB YB (n) = YRB When YLB <YRB YB (n) = YLB ... (11) and The ruled line separation position determination unit 34 sets a character extraction start position and an end position in the column direction based on these separation positions, and determines these extraction positions by the control unit 18 and the inclination calculation unit 36a.
Output to For example, the separation positions YT (n) and YB (n)
The character extraction start position and end position in the column direction
18 is the cutout position KXL in the row and column directions of the nth character
(N), KXR (n) and YT (n), YB (n) are output to the next-stage device (not shown).

* After S11, the control unit 18 refers to the extraction end flag and determines whether or not the extraction processing has been completed. (* S12).

At S12, if the cutout end flag = 1, the operation is terminated (END), and if the cutout end flag = 0, the inclination calculating section 36 sets the count value n of the character number counter to N (N Is a natural number of an arrangement of 2 ≦ N ≦ n _L ) (* S13). If n <N in * S13, the character of the n-th character is the character before the (N-1) -th character. Therefore, the gradient calculating unit 36a calculates the gradient K (n ) Is calculated (* S14).

At S14, the inclination calculating unit 36a inputs the cutout positions YT (n) and YB (n) of the nth character in the column direction from the ruled line separation position determining unit 34, and furthermore, the center position XMC of the nth character. (N) and the row position information LYT2, LYB2, XKC,
The slope K (n) defined in 5) is calculated. The inclination calculating unit 36a stores the calculated K (n) and outputs it to the correction value calculating unit 36b.

When | YT (n) −LYT2 | ≧ | YB (n) −LYB2 | K (n) = CS × (YB (n) −LYB2) / (XMC (n) −XK
C) When | YT (n) −LYT2 | <| YB (n) −LYB2 | K (n) = CS × (YT (n) −LYT2) / (XMC (n) −XK
C) (15) Here, CS is a constant that is arbitrarily set in order to accurately calculate the inclination K (n), and is set to, for example, CS = 128.

In this embodiment, the difference between the reference position XKC in the line direction and the character center position XMC (n) in the line direction is the distance a described above. Further, the row position information LYT2, LYB2 is used as the reference position in the column direction for obtaining the distance b, and the reference position LYT2 or LYT2 in the column direction is used.
LYB2 and cutout position (ruled line position) YT (n) or YB in the column direction
The difference from (n) is the distance b described above.

* After S13, the correction value calculation unit 36b next outputs the (n
+1) To calculate the position correction value HENI of the character * S1
Perform step 5.

Also, if n ≧ N in * S13, the inclination calculator 36a calculates n =
It is determined whether or not N (* S16).

* If n = N in S16, it indicates that the character of the nth character is the character of the Nth character. Therefore, the inclination calculating unit 36a calculates the position correction value of the character from the Nth character onward. The average inclination AK to be used is calculated (* S17). The average inclination AK is represented by the following equation (16).

* After S17, the correction value calculation unit 36b outputs the next (n
+1) To determine the position correction value HENI of the character *
Perform the process of S15.

* If n ≠ N in S16, the n-th character is the (N +
1) Since it indicates that it is the character after the character, the correction value calculation unit 36b next to * S16 determines the position correction value HENI of the (n + 1) th character to be extracted next. Perform processing.

* In S15, the correction value calculation unit 36b checks whether or not n <N.
The inclination K is used to calculate the position correction value HENI of the character to be extracted next.
It is determined whether to use (n) or the average inclination AK.

* If n <N in S15, the (n + 1) th to be cut out next
Since the character of the character is the character before the Nth character, the correction value calculation unit 36b obtains the position correction value HENI of the character to be cut out next using the inclination K (n) (* S18). Correction value calculator 3
6b, the inclination K (n) is input from the inclination calculator 36a, and the reference position XKC and the starting position KXL in the line direction of the next character to be extracted.
(N + 1) is input from the control unit 18, and the position correction value HENI is calculated according to, for example, the following equation (17).

HENI = (KXL (n + 1) + CWH-XKC) × K (n) / CS ...
(17) However, CWH is the width in the X direction of the character obtained from the format information.
For example, when the character width in the X direction is 4 mm and the resolution of the image sensor is 16 pixels / mm, CWH may be set to 32. In this embodiment, XKC is a reference position for obtaining the distance c, KXL (n + 1) + CWH is the center position in the line direction of the next character to be extracted using format information, and c = c KXL (n + 1) + CWH-XKC is calculated.

* After S18, the control unit 18 adds 1 to the count value of the character counter.
The addition is performed (* S19), and then the correction value calculation unit 36b performs the processing of * S4.

* If n ≧ N in S15, the (n + 1) th to be cut out next
Since the character of the character is the character after the N-th character, the correction value calculation unit 36b obtains the position correction value HENI of the character to be extracted next using the average inclination AK (* S20). Correction value calculator 36b
Inputs the average inclination AK from the inclination calculator 36a, and obtains the position correction value HENI according to, for example, the following equation (18).

HENI = (KXL (n + 1) + CWH-XKC) .times.AK / CS (18) After * S20, the control unit 18 performs * S19, and then the correction value calculation unit 36b performs * S4.

According to the first and second embodiments described above, the position on the character side of the ruled line adjacent to the character to be extracted is set as a character extraction position, and therefore, a character is accurately extracted for each character. Characters can be cut out separately from the ruled lines.

Further, according to the second embodiment, the character detection area and the ruled line detection area are set using the positions and inclinations of the extracted characters in the column direction and the row direction which have already been detected. It is possible to provide a high-performance character recognition device capable of performing accurate extraction and performing accurate character recognition by accurately extracting characters even if the character pitch is indefinite when this embodiment is applied to a character recognition device. Further, according to the second embodiment, the average value of the inclination of each character obtained for the (N-1) th character or earlier in the character line is used for cutting out the character after the Nth character. Can be faster.

The present invention is not limited only to the above-described embodiment, and the operation of each component, the operation flow, the input / output signal, the numerical condition of each parameter, the relative positional relationship, the number of arrangements, and other conditions Any suitable changes can be made.

For example, in the above-described embodiment, the character detection area does not include the ruled line but includes only the character. However, the character detection area may be set to include the character and the ruled line. Even if the character detection area includes characters and ruled lines, the character position in the line direction can be detected by arbitrarily setting the threshold value THLB.

In the above-described embodiment, the character detection area is set so as to include the cut-out character and the character next to the character, and the start position of the next character detected from the change in the marginal distribution is used to determine the character in the line direction of the next character. Although the detection area position and the ruled line detection area position are detected, the character detection area is set so as to include only the cut-out character and the margin area adjacent to the character, and the cut-out position in the line direction is detected. The character detection area position and the ruled line position in the line direction of the character to be extracted next may be set using the extraction position and the character pitch, character width, and other format information.

Also, in the above-described embodiment, an example has been described in which characters are cut out from a character line with a horizontally written ruled line.
The present invention can also be applied to a case where characters are cut out from a character line with vertical ruled lines as shown in the figure.

Further, in the above-described embodiment, an example in which one ruled line is provided on one side and one side of one character line has been described. However, the present invention is applied to a case where there is a ruled line only on one side of the character line. Can also be applied. If there is a ruled line on only one side of the character line, the ruled line position near the first character is first detected in the same manner as described above, and the ruled line position, the width of the line in the column direction, the line spacing, etc. May be used to set the character detection area position and ruled line detection area position in the column direction of the first character.

In addition, the inclination amount of the character line and the position correction value for setting the character detection area and the ruled line detection area are not limited to those in the above-described embodiment, but can be arbitrarily and suitably changed. For example, in the second embodiment, the inclination K (n) is set to the row position information LYT.
2, LYB2 was determined as the reference position, but these L
YT2, LYB2 in place of the first character in the column direction YT
(1) The inclination K (n) may be used with YB (1) as a reference position. Further, in the equation (15) of the second embodiment, the cutting start position in the column direction is set as the inclination K (n).
The method of minimizing the gradient obtained by using YT (n) and the gradient obtained by using the cutout end position YB (n) in the column direction is used, but these positions YT (n) and YB (n) are used. The average value of the gradients obtained using (n) may be used, or the position correction value HENI may be obtained by using the gradient obtained using the position YT (n) as the gradient K (n). YT1 (n) and YT2 (n) are set based on the position correction value HENI, and the position correction value HENI is obtained as the inclination K (n) obtained using the cutout end position YB (n) in the column direction. YB1 (n) and YB2 (n) may be set based on HENI. In the second embodiment, instead of obtaining the inclination K (n) from the character cutout position in the column direction, the inclination may be obtained from the position of the ruled line.

(Effects of the Invention) As is clear from the above description, according to the character extracting apparatus of the present invention, the character detection area and the ruled line detection area for the second and subsequent characters of the character line are set in the column direction. The position correction value is set based on the position of the character cut out earlier.

The position correction value indicates the amount of displacement or the range of displacement of the character and the ruled line in the column direction. Therefore, even if the character line is inclined, the character detection area includes the character to be cut out and is adjacent to the character. The character detection area can be set accurately so as not to include the characters of the character line, and the ruled line detection area can be set accurately so that the ruled line detection area includes the ruled line adjacent to the character to be cut out. .

Therefore, since the continuous position change of the character in the inclined character line can be propagated to the character detection area and the ruled line detection area, when the character of the character line is cut out from the inclined character line with the ruled line. The character of the character line can be accurately extracted without extracting the character of the character line adjacent to the character line.

In addition, since the character detection area and the ruled line detection area can be set as described above, the peripheral distribution in these areas is created, and the created peripheral distribution is compared with a predetermined threshold value to check the change in the peripheral distribution. The character can be accurately and stably cut out by a simple process of detecting the position of the character and the ruled line based on the. Since characters can be cut out by simple processing, the device configuration can be simplified.

In addition, since the character detection area and the ruled line detection area can be set as described above, the position of the character to be cut out and the position of the ruled line adjacent to the character can be accurately detected, and thus the character and the ruled line are in contact. The character can be cut out by accurately separating the character and the ruled line.

Therefore, if the character extraction device of the present invention is applied to a character recognition device, a document in which another character line exists near a character line with a ruled line is scanned in a direction shifted from the true document line direction of the document. Even if the image data is obtained or the character line with ruled lines is printed on the document with a slant, the character is accurately cut out from the change of the surrounding distribution and the character recognition can be performed accurately. Can be made small and inexpensive.

[Brief description of the drawings]

FIG. 1 is a functional block diagram schematically showing a configuration of a first embodiment of the present invention, FIG. 2 is a diagram showing an example of an operation flow of the first embodiment, and FIGS. 3 (A) and 3 (B). FIGS. 4A and 4B are diagrams showing examples of peripheral distributions of image data and character detection regions, and FIGS. 5A and 5B are diagrams showing examples of peripheral distributions of image data and character detection regions. FIGS. 6B and 6C are diagrams showing examples of peripheral distribution of image data and ruled line detection areas, FIG. 6 is a functional block diagram schematically showing a configuration of a second embodiment of the present invention, and FIG. FIG. 8 is a diagram showing an example of the operation flow of the embodiment, and FIG. 8 is a diagram showing an example of a character line with a vertically written ruled line. 10: Image output unit, 12: Image storage unit 14: Character interval detection unit, 16: Character ruled line separation unit 17, 36: Correction value determination unit 18: Control unit.

Claims (1)

(57) [Claims] 1. An image output unit for outputting image data of a character line with a ruled line, an image recording unit for storing the image data, and an image of a character detection area set for each character of the character line Scanning the data in the column direction, detecting the cumulative number of black pixels for each scanning line, creating a first peripheral distribution, detecting the position of a character in a character detection area from a change in the first peripheral distribution, A character interval detection unit that sets the position of a ruled line detection area adjacent to the character based on the position of the character and sets a character cutout position in the line direction; and scans the image data of the ruled line detection area in the line direction to perform each scan. Detecting the cumulative number of black pixels for each line to create a second peripheral distribution, detecting the position of the ruled line from a change in the second peripheral distribution,
A character rule separation unit that sets a character and rule separation position based on the position of the rule and sets a character cutout position in a column direction based on the separation position; A correction value determination unit that sets a position correction value in the column direction for setting the area and the ruled line detection area based on the position of a character cut out before that; and a control unit that outputs a character cutout position for each character. A character extracting device, comprising: