JPH06266893A - Optical character reader - Google Patents

Abstract

PURPOSE:To reduce unread or erroneous reading status in an optical character reader. CONSTITUTION:A character frame calculating part 2 calculates a rough position at every read character inputted from a picture input part 1 by using previously inputted information such as the leading position of a character frame, a character pitch and the number of characters. A continued character judging part 3 inspects the presence/absence of image data on the position on the character frame calculated by the calculating part 2 and judges whether or not it is a continued character based upon the inspected result. When the judging part 3 judges the character as a continued character, an unnecessary range detecting part 4 detects the range of an unnecessary lateral bar-like image due to the continued character and a deleting part 5 deletes the unnecessary range detected by te detecting part 4. A character segmenting part 6 determines the accurate segmenting position at every read character in the range of the character frame calculated by the calculating part 2 or in the range for which the unnecessary range is deleted by the deleting part 5 and a character recognizing part 7 recognizes a read character segmented from the accurate segmenting position.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reader, and more particularly to an optical character reader for reading handwritten characters.

[0002]

2. Description of the Related Art Conventionally, in an optical character reading apparatus, as shown in FIG. 7, an image input section 1 for inputting characters written on a sheet as image data and a character frame for calculating the character frame of the characters. The calculation unit 2 includes a character cutout unit 6 that cuts out characters in the character frame, and a character recognition unit 7 that performs character recognition for each cut out character.

When the image input unit 1 takes in the image data of the characters written on the paper into a memory (not shown), the character frame calculation unit 2 preliminarily inputs information on the start position of the character frame, the character pitch, and the number of characters. Is used to calculate the rough position for each read character.

That is, the character frame calculation unit 2 sets Wa (i) = Xst + i · P (i =, where Xst is the start position of the specified character frame, P is the character pitch, and n is the number of characters.
The character frame position Wa (i) is calculated by the expression (0 to n). As a result, the character box calculation unit 2 causes the character box positions Wa (0), Wa (1), Wa (2), and Wa (2), as shown in FIG.
Wa (3) can be calculated.

The character cutout unit 6 projects the read character vertically and horizontally within the range of the rough position calculated by the character frame calculation unit 2, and determines an accurate cutout position for each read character. The character recognition unit 7 recognizes the read character cut out from the correct cutting position.

That is, as shown in FIG. 8, the character cut-out portion 6 has a left character frame position Wa (1) and a right character frame position Wa.
The vertical projection is performed within the range of (2) to determine the character positions Xa and Xb in the X direction. Next, the character cutout unit 6 projects in the horizontal direction within the range of the character positions Xa and Xb to determine the character positions Ya and Yb in the Y direction. The character recognition unit 7 performs character recognition on the image data for one character surrounded by the character positions Xa, Xb, Ya, Yb.

[0007]

In the above-mentioned conventional optical character reading device, when there are consecutive characters in horizontal writing as shown in FIG. 2, that is, one character to be read and at least one of the characters on both sides of the character to be read. When the two are in contact with each other with a horizontal bar-shaped image, there is a drawback that a horizontal bar-shaped image connecting the characters remains when cutting out the characters, resulting in non-reading or misreading.

Here, FIG. 2 shows a case where the horizontal bar from "0" of the first character is in contact with "6" of the second character.

Therefore, an object of the present invention is to solve the above drawbacks and to provide an optical character reading apparatus capable of reducing unread or misreading by deleting unnecessary horizontal bar-shaped images of continuous characters. is there.

[0010]

An optical character reading device according to the present invention is an optical character reading device for recognizing a character in a preset character frame, which is determined according to the presence or absence of image data on the character frame. Determining means for determining whether or not a character in the character frame is a continuous character, and determining an intersection of the unnecessary image data including image data on the character frame and the character when the determining means determines the continuous character. Detection means for detecting,
The deletion means deletes the unnecessary image data based on the intersection detected by the detection means and the coordinates of the image data on the character frame.

[0011]

An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, when the image input unit 1 takes in the image data of the characters written on the paper into a memory (not shown), the character frame calculation unit 2 inputs the information of the leading position of the character frame, the character pitch, and the number of characters which are input in advance. Is used to calculate the rough position for each read character.

The continuous character determining unit 3 inspects the presence or absence of image data at the position on the character frame calculated by the character frame calculating unit 2,
Based on the inspection result, it is determined whether the character is continuous. When the continuous character determining unit 3 determines that the character is a continuous character, the unnecessary range detecting unit 4 performs the process. When the continuous character is not determined, the process of the character cutting unit 6 is performed as in the conventional example. .

The unnecessary range detecting section 4 detects a range of an unnecessary horizontal bar image formed by continuous characters. The deleting unit 5 deletes the unnecessary range detected by the unnecessary range detecting unit 4.

The character cutout unit 6 projects the read characters vertically and horizontally in the range of the rough position calculated by the character frame calculation unit 2 or in the range in which the unnecessary range is deleted by the deletion unit 5, and reads each read character. Determine the exact cutout position of. The character recognition unit 7 recognizes the read character cut out from the accurate cutting position determined by the character cutting unit 6.

FIG. 2 is a diagram showing the relationship between the read image and the frame position, FIG. 3 is a diagram showing the process of detecting continuous characters by the continuous character determining unit 3 in FIG. 1, and FIG. 4 is unnecessary in FIG. FIG. 6 is a diagram showing a process of detecting an unnecessary range by the range detector 4.
FIG. 6 is a diagram showing a deletion process by the deletion unit 5 in FIG. 1.

A character recognition processing operation according to an embodiment of the present invention will be described with reference to FIGS. Less than,
As shown in FIG. 2, a case where the horizontal bar from the first character “0” is in contact with the second character “6” when recognizing the character string “067” will be described.

In this case, since the start position Xst of the character frame, the character pitch P, and the number of characters n are input to the character frame calculation unit 2, the character frame calculation unit 2 calculates Wa (i) = Xst + i · P.
The character frame position Wa according to the formula (i = 0 to n)
(0), Wa (1), Wa (2), Wa (3) are calculated.

The continuous character determining unit 3 determines the character frame positions Wa (0), Wa (1), Wa (2), calculated by the character frame calculating unit 2,
For each Wa (3), the presence or absence of image data is inspected at the position on the character frame, and it is determined whether or not there is a continuous character based on the inspection result. In the example shown in FIG. 2, since the horizontal bar from the first character “0” is in contact with the second character “6”, the continuous character determination unit 3 determines that the character frame position Wa (1) is Y. It is detected that there is image data on the axis, and it is determined that the first character "0" and the second character "6" are consecutive characters.

In the unnecessary area detecting section 4, the continuous character judging section 3 is the first.
When it is determined that the character "0" and the second character "6" are continuous characters, the unnecessary ranges of the first character "0" and the second character "6" are detected. Here, detection of the unnecessary range of the second character "6" will be described.

First, the unnecessary range detection unit 4 deletes the upper end coordinates (Xd, Xd, where the horizontal bar-shaped image intersects the second character "6").
Yd) and the deletion bottom coordinate (Xf, Yf) are obtained. The unnecessary range detection unit 4 firstly detects the left character frame position Wa (1).
Of the left frame intersection [Wa (1), Ywu] where the Y axis and the image intersect and the bottom frame intersection lower coordinate [Wa
(1), Ywl] is detected.

Next, the unnecessary range detection unit 4 makes a search in the axial direction orthogonal to the Y-axis by using the left frame intersection upper end coordinates [Wa (1), Ywu] as the search start point, and searches for that axis and the second character. The point where the eye "6" intersects is detected. The coordinates detected at this time are defined as deleted upper end coordinates (Xd, Yd).

Subsequently, the unnecessary range detecting unit 4 sets the lower end coordinates [Wa (1), Ywl] of the left frame intersection as the search start point, and determines the Y
The search is performed in the axial direction orthogonal to the axis, and the point where the axis intersects the second character "6" is detected. The coordinates detected at this time are defined as deleted bottom coordinates (Xf, Yf).

The unnecessary range detecting section 4 is a range surrounded by the deleted upper end coordinates (Xd, Yd) and the deleted lower end coordinates (Xf, Yf) and the Y axis of the left character frame position Wa (1) detected by the above-mentioned processing. Is an unnecessary range (see FIG. 3).

A method for obtaining the above-described deleted upper end coordinates (Xd, Yd) and deleted lower end coordinates (Xf, Yf) will be described in detail below with reference to FIG.

When detecting the deletion upper end coordinates (Xd, Yd), the unnecessary range detecting section 4 firstly sets the left frame crossing upper end coordinates [Wa].
(1), Ywu] as the search start point, and the black-and-white change point coordinates (Xc, Yc) at which the image data changes from white to black in the axial direction orthogonal to the Y-axis are searched.

When the black-and-white changing point coordinates (Xc, Yc) (hatched portion in the figure) are detected, the unnecessary range detecting section 4 detects the coordinates (Xc, Yc) on one dot line of the black-and-white changing point coordinates (Xc, Yc).
3 dots near (Yc-1) (Xc, Yc-1), (Xc
+1, Yc-1), (Xc + 2, Yc-1) (inside the thick line frame in the figure) are used as the inspection dots, and it is inspected whether or not a black dot is included in these inspection dots.

In this case, since black dots are included in these inspection dots, the unnecessary range detecting unit 4 makes the image data near the black-and-white changing point coordinates (Xc, Yc) an "upward key" image. Judge that there is. This "upward key shape"
When the image is detected, the unnecessary range detecting unit 4 sets the deletion upper end coordinates (Xd, Yd) to the black and white change point coordinates (Xc, Y).
c).

When detecting the deletion bottom end coordinates (Xf, Yf), the unnecessary range detecting unit 4 sets the left frame intersection bottom end coordinates [Wa].
(1), Ywl] as the search start point, and the black-and-white change point coordinates (Xe, Ye) at which the image data changes from white to black in the axial direction orthogonal to the Y-axis are searched.

When the black-and-white changing point coordinates (Xe, Ye) (hatched portion in the figure) are detected, the unnecessary range detecting section 4 detects the coordinates (Xe, Ye) one dot line below the black-and-white changing point coordinates (Xe, Ye).
3 dots near (Ye + 1) (Xe, Ye + 1), (Xe
+1, Ye + 1), (Xe + 2, Ye + 1) (inside the thick line frame in the drawing) are used as the inspection dots, and it is inspected whether or not a black dot is included in these inspection dots.

In this case, since black dots are not included in these inspection dots, the unnecessary range detection unit 4 moves the Y coordinate downward by one line and sets (Xe, Ye + 1) as the next search start point, Similar to the above process, the black and white change point coordinates (Xe ', Ye') at which the image data changes from white to black
To search.

The black and white change point coordinates (Xe ', Ye')
When the [hatched portion in the drawing, the coordinates are (Xe + 3, Ye + 1)] is detected, the unnecessary range detection unit 4 detects the coordinates (X) one dot line below the black-and-white change point coordinates (Xe ′, Ye ′).
3 dots (Xe ', Ye' +) in the vicinity of e ', Ye' + 1)
1), (Xe '+ 1, Ye' + 1), (Xe '+ 2, Y
e ′ + 1) (inside the thick line frame in the figure) is used as an inspection dot, and it is inspected whether or not a black dot is included in these inspection dots.

In this case, since black dots are included in these inspection dots, the unnecessary range detecting unit 4 determines that the image data near the black-and-white changing point coordinates (Xe ', Ye') is "downward key shape". Judge as an image. When this "downward key shape" image is detected, the unnecessary range detection unit 4 sets the deletion lower end coordinates (Xf, Yf) to the black and white change point coordinates (X
e ', Ye').

As described above, the unnecessary range detecting section 4 is a range surrounded by the deleted upper end coordinates (Xd, Yd) and the deleted lower end coordinates (Xf, Yf) and the Y axis of the left character frame position Wa (1), that is, unnecessary. When the range is detected, the deleting unit 5 deletes the image data within the unnecessary range.

In this case, the deleting section 5 firstly deletes the displacement position Z in the X direction for each line movement [the upper end coordinate of the deletion (Xd, Y
d) and the deletion bottom coordinate (Xf, Yf), the slope of a straight line] is deleted top coordinate (Xd, Yd) and deletion bottom coordinate (X
f, Yf) and Z = (Xf-Xd) / (Yf-Y
It is obtained from the equation of d).

In FIG. 5, since Xd = Xf + 2 and Yd = Yf-4, Z = (Xf−Xd) / (Yf−Yd) = [Xf− (Xf + 2)] / [Yf− (Yf-4)] = -2 / 4 = -0.5.

The deleting unit 5 deletes the upper end coordinates (Xd, Yd) and the lower end coordinates (Xf, Yf) and the left character frame position Wa.
Image data within the unnecessary range is deleted using the Y axis and the deletion position displacement Z of (1).

Therefore, as shown in FIG. 5, the deletion unit 5 deletes the deletion first line to the deletion fifth line. Here, the deletion start point of the deletion first line is the coordinate [Wa (1),
Yd], the deletion end point is the coordinate [Xd, Yd], and the deletion start point of the second deletion line is the coordinate [Wa (1), Yd +
1], the deletion end point is the coordinate [Xd + Z, Yd + 1].

The deletion start point of the third deletion line is the coordinate [Wa (1), Yd + 2], and the deletion end point is the coordinate [Xd.
+ 2Z, Yd + 2], the deletion start point of the fourth deletion line is coordinate [Wa (1), Yd + 3], and the deletion end point is coordinate [Xd + 3Z, Yd + 3].

Furthermore, the deletion start point of the fifth deletion line, which is the final line, is the coordinate [Wa (1), Yd + 4], and the deletion end point is the coordinate [Xd + 4Z, Yd + 4]. However,
When the deletion position displacement Z in the X direction contains a decimal,
The X coordinate to be deleted is determined by rounding off and converting to an integer.

Therefore, the first line to be deleted is [Wa
(1), Yd] to [Xd, Yd], and the deleted second line is [Wa (1), Yd + 1] to [Xd-1, Yd + 1],
The third line to be deleted is [Wa (1), Yd + 2] to [Xd-
1, Yd + 2], and the deleted fourth line is [Wa (1), Yd
+3] to [Xd-2, Yd + 3], and the fifth deleted line becomes [Wa (1), Yd + 4] to [Xd-2, Yd + 4].

FIG. 6 is a diagram showing a process of detecting an unnecessary range by the unnecessary range detecting section 4 of FIG. In FIG. 6, processing is performed when the unnecessary range detection unit 4 cannot detect the "upward key shape" image.

A method of obtaining the upper delete coordinate (Xd, Yd) and the lower delete coordinate (Xf, Yf) in that case will be described in detail below with reference to FIG.

When detecting the deletion upper end coordinates (Xd, Yd), the unnecessary range detecting unit 4 firstly sets the left frame crossing upper end coordinates [Wa].
(1), Ywu] as a search start point, and a black-and-white change point at which the image data changes from white to black in the axial direction orthogonal to the Y-axis is searched. At this time, since there is no black-and-white change point, the unnecessary range detection unit 4 determines that the "upward key-shaped" image cannot be detected.

Next, when detecting the deleted lower end coordinates (Xf, Yf), the unnecessary range detection unit 4 sets the left frame crossing lower end coordinates [Wa].
(1), Ywl] as the search start point, and the black-and-white change point coordinates (Xe, Ye) at which the image data changes from white to black in the axial direction orthogonal to the Y-axis are searched.

When the black-and-white changing point coordinates (Xe, Ye) (hatched portion in the figure) are detected, the unnecessary range detecting unit 4 detects the coordinates (Xe, Ye) one dot line below the black-and-white changing point coordinates (Xe, Ye).
3 dots near (Ye + 1) (Xe, Ye + 1), (Xe
+1, Ye + 1), (Xe + 2, Ye + 1) (inside the thick line frame in the drawing) are used as the inspection dots, and it is inspected whether or not a black dot is included in these inspection dots.

In this case, since black dots are included in these inspection dots, the unnecessary range detection unit 4 makes the image data in the vicinity of the black-and-white change point coordinates (Xe, Ye) a "downward key shape" image. Judge that there is. This "downward key shape"
When the image of is detected, the unnecessary range detection unit 4 sets the deletion lower end coordinates (Xf, Yf) to the black and white change point coordinates (Xe, Y).
e).

In the above processing, since the "upward key-shaped" image could not be detected, the deletion upper end coordinates (Xd, Y
d) remains undecided. Therefore, the unnecessary range detection unit 4
Is the upper edge coordinate [Wa (1), Ywu] intersecting the left frame and the black and white change point coordinate (Xe, Ye) as the lower edge coordinate (Xf, Yf) determined in the above process.
d, Yd) is set to (Xe, Ywu).

The deleting unit 5 deletes the upper end coordinates (Xd, Yd) and the deleting lower end coordinates (Xf, Y) detected by the unnecessary range detecting unit 4.
The image data within the range surrounded by f) and the Y axis of the left character frame position Wa (1) is deleted.

The above process is a process when there is an unnecessary horizontal bar-shaped image on the left side of "0".
Even if there is an unnecessary horizontal bar-shaped image on the right side of "0" like the character "0", the unnecessary image can be deleted in the same manner as the above processing.

As described above, the continuous character determining unit 3 detects that the image data exists on the Y axis of the character frame position calculated by the character frame calculating unit 2 and determines that the characters in the character frame are continuous characters. At this time, the unnecessary range detection unit 4 sets the two points of the upper and lower coordinates of the deletion where the horizontal bar-shaped image intersects with the character as the search start point with the coordinates of the image data on the Y axis of the character frame position as the search start point. It is detected by searching the black-and-white change point coordinates where the image data changes from white to black in the axis direction orthogonal to the axis, and the detected delete upper end coordinates and delete lower end coordinates, the Y axis of the character frame position, and the delete upper end. The deletion unit 5 using the coordinates and the deletion position displacement calculated from the deletion bottom coordinates
By deleting the image data in the unnecessary range, it becomes possible to correctly recognize the unread or misread characters in the unnecessary horizontal bar-shaped image of continuous characters.
Therefore, it is possible to reduce misreading and erroneous reading in the optical character reading device.

Although one embodiment of the present invention has described the deletion of unnecessary horizontal bar-shaped images due to continuous characters in horizontally written handwritten characters, unnecessary vertical bar-shaped images due to continuous characters in vertically written handwritten characters have been described. It is obvious that the present invention can be applied to, but is not limited to.

[0053]

As described above, according to the present invention, when the image data on the preset character frame is detected and it is determined that the characters in the character frame are continuous characters, the character frame Detects the conversion point from white dots to black dots on the axis orthogonal to the character frame based on the image data coordinates, and based on the conversion point, the axis and the character orthogonal to the image data coordinates on the character frame And a second intersection between the unnecessary image data including the image data on the character frame and the character, and the coordinates of the image data on the character frame and the first and second intersections. By deleting the unnecessary image data based on the gradient between the first and second intersections, there is an effect that it is possible to reduce non-reading and erroneous reading.

[Brief description of drawings]

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

FIG. 2 is a diagram showing a relationship between a read image and a frame position.

FIG. 3 is a diagram illustrating a process of detecting a continuous character by a continuous character determining unit in FIG.

FIG. 4 is a diagram showing a process of detecting an unnecessary range by an unnecessary range detection unit in FIG.

5 is a diagram showing a deletion process by a deletion unit in FIG.

FIG. 6 is a diagram showing a process of detecting an unnecessary range by an unnecessary range detecting unit in FIG.

FIG. 7 is a block diagram showing a configuration of a conventional example.

FIG. 8 is a diagram showing a clipping process performed by the character clipping unit of FIG. 7.

[Explanation of symbols]

 2 Character box calculation unit 3 Continuation character determination unit 4 Unnecessary range detection unit 5 Deletion unit 6 Character cutout unit

Claims (4)

[Claims] 1. An optical character reading device for recognizing a character within a preset character frame, wherein whether or not the character within the character frame is a continuous character is determined according to the presence or absence of image data on the character frame. Determination means for determining, determination means for detecting an intersection of the unnecessary image data including image data on the character frame and the character when the determination means determines the continuous character, and the detection means An optical character reading device comprising: a deleting unit that deletes the unnecessary image data based on the intersection and the coordinates of the image data on the character frame. 2. A detecting means detects a conversion point from white dot data to black dot data on an axis orthogonal to the character frame based on the coordinates of the image data on the character frame, and based on the conversion point. The first intersection of the character and the orthogonal axis passing through the coordinates of the image data on the character frame and the second intersection of the unnecessary image data and the character are detected. The optical character reader according to claim 1. 3. The deleting means calculates a gradient between the first and second intersections, and based on the coordinates of the image data on the character frame, the first and second intersections, and the gradient. The optical character reader according to claim 2, wherein the unnecessary image data is deleted. 4. When the first intersection is not detected,
The optical character reader according to claim 2 or 3, wherein the first intersection is set based on the second intersection.