JPH09114925A - Optical character reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize a void character without preparing a character dictionary for void. SOLUTION: An image input means 1 inputs a document image containing the void character to an image memory 2, and a line segmenting means 3 segments the document image in unit of line and inputs it to a line memory 4. A character pattern segmenting means 5 segments a character pattern unit of one character and inputs it to a character pattern memory 6. From the distribution value and width of the character pattern, a void character pattern judging means 7 judges whether that character pattern is a void pattern or not and when the character pattern is a void pattern, it is inputted to a void character pattern memory 8. From the distribution value and width of the character pattern and the distribution value and width of a character absent pattern distribution value memory 9, an inverted character pattern segmenting means 10 segments inverted character pattern and a character pattern white/ black inverting means 12 inverts white and black. A character recognizing means 13 performs character recognition.

Description

Detailed Description of the Invention

[0001]

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

[0002]

2. Description of the Related Art Conventionally, techniques in such a field include:
For example, some documents were described in the following documents. Reference: Japanese Unexamined Patent Publication No. 1-279385 In character recognition in a general optical character reading device, for example, black characters are printed on a white background paper, entered characters are cut out character by character, and the cut out character patterns are collated with a dictionary. And recognize the character. On the other hand, there is an apparatus disclosed in the above document as an apparatus for cutting out a character pattern from a character printed and written in white on a black background. FIG. 2 is a diagram showing an example of a document pattern. When reading white characters (“2”, “3”, etc.) drawn in a black outline as shown in FIG. 2, a general optical character recognition apparatus prepares a recognition dictionary for this pattern. There is a need. Further, the device of the above-mentioned document is a device for reading a reverse character like a white character written on a black paper, and cannot read an outlined character written on a white paper as shown in FIG.

[0003]

However, the conventional optical character recognizing device has the following problems when reading a white character as shown in FIG. When reading a white character as shown in FIG. 2, in a general optical character recognition device, as described above, it is necessary to prepare a recognition dictionary for this character pattern, and the capacity of the recognition dictionary becomes enormous. Further, as described above, the apparatus of the above-mentioned document cannot read white characters written on a white paper as shown in FIG.

[0004]

In order to solve the above problems, the first invention is image input means for inputting an image of a document written on a paper whose ground color is white or dropout color to an image memory. A character pattern cutout means for cutting out a character pattern of one character based on the image of the document input to the image memory, and a character pattern for recognizing the cutout character pattern of one character by referring to a recognition dictionary. In the optical character reading device provided with the identification means, the following means are provided. That is, a white character pattern determining means for determining whether or not the character pattern is a white character pattern drawn in a black contour of a predetermined shape, and a white character without a character in which a plurality of contours are all black. Characterless pattern distribution value memory in which pattern data is recorded in advance, and the characterless white character pattern and the white character pattern recorded in the characterless pattern distribution value memory are compared, and the white character is compared. Select a blank characterless pattern corresponding to the pattern, and based on the blank character pattern and the blank characterless pattern corresponding to the blank character pattern, cut the reverse character pattern of the area including the blank character pattern. A reverse character pattern cutout unit for outputting and a character pattern white / black reversal unit for reversing the image of the reversed character pattern white to black and black to white are provided. In the case of a white character pattern, the character pattern recognition / identification means is configured to perform character recognition from the character pattern reversed by the character pattern white / black reversing means. Therefore, the above problem can be solved.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION First Embodiment FIG. 1 is a functional block diagram of an optical character reading apparatus showing a first embodiment of the present invention. The optical character reading device according to the first embodiment is different from the conventional optical character reading device in that the character pattern is different in order to read white characters written on a paper whose background color is white or dropout color. White character pattern determining means 7 for determining whether or not a character is a white character, a white character pattern memory 8 for storing a white character pattern, a character with a black outline, and a character for recording data of a white character A pattern distribution value memory 9, an inverted character pattern cutout means 10 for cutting out an inverted character pattern, an inverted character pattern memory 11 for storing an inverted character pattern, and a character pattern white / black inversion means 12 for inverting an image from black to white. It is provided. As shown in FIG. 1, the optical character reader according to the first embodiment includes an image input unit 1, an image memory 2, a line cutout unit 3, a line memory 4, a character pattern cutout unit 5, a character pattern memory 6, Outline character pattern discrimination means 7, outline character pattern memory 8, no-character pattern distribution value memory 9,
The reverse character pattern cutout unit 10, the reverse character pattern memory 11, the character pattern white / black inversion unit 12, and the character pattern recognition unit 13 are included. Image input means 1
Is for inputting a character image written on a paper whose ground color is white or dropout color to the image memory 2.
The line cut-out means 3 cuts out each line and inputs the pattern of the line to the line memory 4.

The character pattern cutout means 5 cuts out a character pattern for each character in each line in the line memory 4 and inputs it to the character pattern memory 6. The white character pattern discriminating means 7 judges whether or not the character pattern of each character in the character pattern memory 6 is a white character, and if it is a white character, the white character pattern is the white character. The data is input to the pattern memory 8. The characterless pattern distribution value memory 9 is a memory for recording the data of a plurality of characterless white patterns, and the reverse character pattern cutting means 10 is provided.
Is to cut out a reversed character pattern including a blank pattern area and input the reversed character pattern into the reversed character pattern memory 11, and the character pattern white / black reversing means 12 is
The image of the reverse character pattern is reversed from white to black and from black to white. Hereinafter, the operation of FIG. 1 will be described. The image input means 1 binarizes a document image including white characters as shown in FIG. 2 into, for example, white “0” and black “1”,
Input to the image memory 2. The line cutout unit 3 cuts out the document image line by line and inputs it to the line memory 4. The character pattern cutout unit 5 cuts out a character pattern in units of one character from the document image stored in the line memory 4 in units of lines and inputs it to the character pattern memory 6.

As described above, (a) the process of inputting the document image of white characters to the image memory 2 by the image input means (b) the process of inputting the document image line by line by the line cutout means 3 into the cutout memory 4. (C) The character pattern cutting means 5 cuts out a character pattern for each character from an image for each line, and a character pattern memory 6 is provided.
The processes (a) to (c) of the process of inputting to can be implemented by conventional techniques. As an example of the conventional technique, there is JP-A 03-248288. FIG. 3 is a diagram showing an example of a character pattern cut out through (a) to (c), and a broken line in the figure indicates
The range of each extracted character is shown. This character pattern is input to the character pattern memory 6. In the following description, the black area is 1 and the white area is 0 on the character pattern memory 6.
Assume that The outline pattern determining means 7 projects the character pattern on the horizontal axis (hereinafter referred to as the X axis) and the vertical axis (hereinafter referred to as the Y axis) to represent the number of black pixels at each x and y. Calculate the distribution value. The distribution value at the point x is B
The distribution value at _X (x) and point y is defined as B _Y (y). Of x that satisfies B _X (x)> 0, the minimum value is x _s ,
The maximum value is defined as x _e . Further, of y satisfying B _Y (y)> 0, the maximum value is defined as y _s and the minimum value is defined as y _e .
The width of the distribution values in the X and Y directions is | x _e −x _s | =
It is defined as D _x , | y _s −y _e | = D _y .

FIG. 4 is a projection view of the outline character pattern for the outline character "2" on the X-axis and the Y-axis. The following two processes 1-1 and 1-2 are performed for each character pattern. Process 1-1: For appropriately set values D _xs and D _ys ,
Character patterns that satisfy the following conditions are defined as outline character candidates, and character patterns that do not satisfy the following are not outline characters. However, D _xs and D _ys are set so that the following condition 1 is always satisfied if the character is a white character, and the number of characters other than the white character satisfying the next condition 1 is the minimum. Condition 1-1: D _x > D _xs (≧ _may be satisfied) and D _y > D _ys
(It may be ≧ 0) FIGS. 5A to 5C are diagrams showing the processing contents of the outline character pattern determining means, and in particular, FIG. 5A shows the X axis of the outline character “2” and The Y-axis projection view, FIG. 6B is the X-axis and Y-axis projection view of “4”, and FIG. 6C is the X-axis and Y-axis projection view of “Q”.

For example, if the white character "2" shown in FIG. 5 (a) and the character pattern of "question" shown in FIG. 5 (c) satisfy the condition 1-1, the patterns shown in FIGS. The pattern of c) becomes a blank character candidate. As a result of processing 1-1, the following processing 1-
Perform Step 2. Process 1-2: For explanation, 1 on the X-axis on the character pattern memory
The bit width is defined as Δx, and the 1-bit width on the Y axis is defined as Δy. Character patterns that satisfy the following condition 2 for appropriately set values Nx, Ny, δx, and δy are white characters, and if they do not satisfy the following conditions, the white characters are not white characters. However, if Nx and Ny are blank characters, the condition must be met,
Set so that the number of fills other than the outline characters is the minimum. δx and δy are minute values determined in consideration of noise and the like at the time of image input. Condition 1-2: | B _X (x _s + n _x Δx) −B _X (x _e −n _xe Δx) | <δ x (n _x = 0, 1, ..., N _X ) (may be ≦) | _{BY (y s -n y Δy) -B} Y (y e + n y Δy) | <δy (n y = 0,1, ..., N y) ( good ≦ even) this condition 1-2, the white character The case is based on the fact that the contour is substantially left-right and vertically symmetrical. FIG.
In the patterns (a) and (c), if only the condition (a) satisfies the condition 2, the pattern (a) will be selected as a blank character.

In this way, for each character pattern,
By sequentially performing the processing 1-1 and the processing 1-2, it is possible to determine whether the pattern cut out by the character pattern cutting means 5 is a blank character. As a result of processing 1-1 and processing 1-2, if it is determined that the character pattern is not a blank character pattern, the pattern is input to the character pattern recognition means 13, and if it is determined that the character pattern is a blank character pattern, the pattern is input. Is input to the blank character pattern memory 8. Process 1-1 and process 1-2
Inversely, the process 1-2 may be performed first, and then the process 1-1 may be performed. Reverse character pattern cutout means 9
As shown in FIG. 4, a white character pattern is input from the white character pattern memory 8 and projected on the X and Y axes to obtain a distribution value. If there are no black pixels in the projected area, the distribution value will be zero. The distribution value at the point x is defined as B _X (x), and the distribution value at the point y is defined as B _Y (y). B _X (x)
Of x satisfying> 0, the minimum value is defined as x _s and the maximum value is defined as x _e . Of y that satisfies B _Y (y)> 0, the minimum value is defined as y _e and the maximum value is defined as y _s . Width of distribution value D _x = | x
_It is defined as _e −x _s |, D _y = | y _s −y _e |.

FIG. 6 is a diagram showing the contents of the characterless pattern distribution value memory. A medium on which white characters without characters (the outline is all black) is written in advance is read by the image input means 1, and the X-axis, as shown in FIG.
Distribution value B _x '(x) (x _s ' ≤x≤
x _e '), B _Y ' (y) (y _e '≤ y ≤ y _s ') and the distribution widths D _x 'and D _y ' are recorded and the values are recorded in the characterless pattern distribution value memory 10. Keep it. FIGS. 7A to 7C
FIG. 6 is a diagram for explaining the processing contents of the reverse character pattern cutout unit of the first embodiment. For example, a perfect circle character-less pattern as shown in FIG. 7A, a square character pattern as shown in FIG. 7B, and a rectangular pattern as shown in FIG. Suppose you have recorded in. The width per bit on the X-axis and Y-axis is Δx in the character pattern memory and the character-less pattern distribution value memory, respectively.
And Δy. Process 2-1 and Process 2-described below
2, the character-free white character pattern corresponding to the white character pattern output from the white character pattern determining means is selected from the patterns recorded in the characterless pattern distribution value memory 10.

Process 2-1: From the patterns recorded in the characterless pattern distribution value memory 10, a pattern satisfying the following condition 2-1 for appropriately set values D _xn and D _yn is selected as a candidate. Conditions _{2-1: | D x -D x '} | <D xn (| D x -D x' | may be _{≦ D xn) | D y -D} y '| <D yn (| D y -D y' | ≦ D _{yn is} also acceptable) The condition 2-1 is that the size of the contour of the characterless pattern is
It indicates that the size is almost the same as the outline of the input character pattern. For example, in FIG. 7, (a), (b),
In the patterns of (c), if (a) and (b) satisfy the above conditions and (c) does not, the patterns of (a) and (b) are selected as candidates. As a result of the process 2-1, the process 2-2 is performed on the pattern selected as the candidate. Process 2-2: Appropriately set values N _x2 , N _y2 , δ _x2 , δ
_{For y2} , a character pattern that satisfies the following conditions 2-2 and 2-3 is selected.

Condition 2-2: │B _x (x _s + n _x Δx) -B _x '(x _s ' + n _x Δx) │ <δ x ₂ (n _x = 0,1,2, ... N _x2 ) and | _{B x (x e -n x Δx} ) -B x '(x e' -n x Δx) | <δx 2 (n x = 0,1,2, ... n x2) and | B _{y (y s} -n _{y Δy) -B y '(y} s' -n y Δy) | <δy 2 (n y = 0,1,2, ... n y2) and _{| B y (y e + n} y Δy) -B y' ( _{y e '+ n y Δy)} | <δy 2 (n y = 0,1,2, ... n y2) n values of _x2 and n _y2 are below the _{x ds, x de, y ds} , relative y _de, A value as large as possible that satisfies the following condition 2-3 is set.

Condition 2-3: x _s + N _x2 Δx <x _ds and x _de <x _e −N _x2 Δx and y _ds <y _s −N _y2 Δy and y _e + N _y2 Δy <y _de where δx ₂ And δy ₂ are minute values determined in consideration of noise at the time of image input. Condition 2-2 is based on the fact that a reversed character pattern appears in the outline character pattern or that the outlines of the characterless patterns corresponding to the left, right, upper and lower contours after the end of the reversed character pattern are almost the same. There is. For example, in FIGS. 7A and 7B, if the pattern of (a) satisfies the condition 2-2, the pattern of (a) is the outline character pattern output from the outline character pattern determining means. It is selected as a white pattern without characters corresponding to.

As described above, the processing 2-1 and the processing 2-
By sequentially performing 2 and 2, it is possible to select a blank character pattern without characters corresponding to the blank character pattern output from the blank character pattern determining means 7. In addition, processing 2-1
The result is the same even if the process 2-2 is first performed and then the process 2-1 is performed by reversing the order of the process 2-2. Next, the reverse character cutout unit 9 cuts out a reverse character pattern by the following processing. 8 (a) and (b),
It is a diagram showing the processing contents of the reverse character pattern cutout means,
In FIG. 8A, X _d represents the x coordinate when | B _X (x) −B _X '(x) | is shown in the vertical direction, and Y _d is | B.
_{Y (y) -B Y '(} y) | its y when showing laterally
Indicates coordinates. The distribution values B _X (x) and B _Y (y) of each white character pattern output from the white character pattern discriminating means 7 and stored in the white character pattern memory 8 and the process 2
-1, The difference between the distribution values B _X ′ (x) and B _Y ′ (y) of the white character pattern without characters selected as a result of processing 2-2 is obtained.

In order to eliminate an error in the distribution value due to noise or the like at the time of image input, the threshold values THxd and Tx are previously set.
Set Hyd to an appropriate value. Then, as shown in FIG. 8A, | B _X (x) −B _X '(x) |> THxd (≧
_However , the minimum value of x that satisfies
Define as _de . │B _Y (y) -B _Y '(y) ｜ ＞ THyd
Of y satisfying (≧), the maximum value is defined as y _ds and the minimum value is defined as y _de . This is because the outline character pattern has a different density in the area of the outline character pattern from the corresponding non-character pattern, so the distribution value | B _X (x) −B _X '(x) |,
| Because reversed character pattern in the x-coordinate or y-coordinate that exceeds the threshold is present _{| B Y (y) -B Y} '(y). As shown in FIG. 8B, the starting point is (x _s ,
y _s ), the end point is (x _e , y _e ), the start point is (x _ds , y _ds ), and the end point is (x _de , y _de ). The area between the outer rectangular area and the inner rectangular area is ANDed with white (bit value is 0) and the inner rectangular area is black (bit value is 1).

FIGS. 9A and 9B are views showing the processing result of the reverse character pattern cutout means 9. As a result, FIG.
As shown in (a) and (b), an outline character pattern of an area surrounded by a rectangular area circumscribing the outline character pattern is obtained. Then, it is recorded in the reverse character pattern memory 11. The character pattern white / black reversing means 12 sets a rectangular area having a start point (x _ds , y _ds ) and an end point (x _de , y _de ) on the reverse character pattern memory 11 for each white character pattern. , 0 and 1 in this area are inverted. FIG. 10 is a diagram showing a processing result of the character pattern white / black inverting means 12. As shown in FIG. 10, the character pattern white / black inverting means 12 provides a pattern which can be recognized by the character pattern recognizing means 13. The character pattern recognition unit 13 reads the reversed character pattern from the reversed character pattern memory 10 if it is a blank character pattern, performs character pattern recognition by referring to the recognition dictionary, and if it is not a blank character pattern, the character pattern is recognized. The character pattern is read from the memory 6 and the character is recognized.

As described above, according to the first embodiment, there is no need to prepare a white character recognition dictionary.
The white character pattern is distinguished from the character pattern in which the white pattern and the character patterns other than the white character pattern are mixed,
It is possible to determine a white character pattern from a character pattern in which character patterns other than the white character are mixed and convert it to a character pattern that can be recognized by the white character pattern recognition means. Therefore, there is an advantage that the dictionary capacity can be reduced without creating a white character recognition dictionary.

Second Embodiment FIG. 11 is a functional block diagram of an optical character reader according to a second embodiment of the present invention. Elements common to those in FIG. 1 are designated by common reference numerals. I am doing it. The outline character recognition device according to the second embodiment of the present invention is different from the outline character recognition device according to the first embodiment in that it does not use the characterless pattern distribution value memory 11 in FIG. Means 20
In other words, the inversion pattern is cut out from the gradient of the white pattern distribution values B _X (x) and B _Y (y). The operation of FIG. 11 will be described below. Similar to the first embodiment, the image input unit 1 extracts a document image including white characters as shown in FIG.
The black is binarized to "1" and input to the image memory 2, and the line cutout unit 3 cuts out the document image line by line and inputs it to the line memory 4. Then, the character pattern cutting means 5 is
From the line-by-line document image stored in the line memory 4, 1
A character pattern for each character is input to the cut-out character pattern memory 6. The outline character pattern determining means 7 determines whether the character pattern is an outline character pattern and inputs the outline character pattern to the outline character pattern memory 8.

FIGS. 12A and 12B are views showing the processing contents of the reverse character pattern cutout means 20 of the second embodiment. The reverse character pattern cutout unit 20 projects a white character pattern on the X axis and the Y axis to obtain a distribution value, as shown in FIG. If there are no black pixels in the projected area, the distribution value is 0.
Becomes The distribution value at the point x is defined as B _X (x), and the distribution value at the point y is defined as B _Y (y). Of x that satisfies B _X (x)> 0, the minimum value is defined as x _s and the maximum value is defined as x _e .
Among y satisfying B _Y (y)> 0, the maximum value is defined as y _s and the minimum value is defined as y _e . Next, FIG. 12 (a), the (b), the midpoint of x _s and _{x e (x s + x e} ) / 2 and y _s and y _e midpoint (y _s + y _e) / 2 Ask for. On the outline character pattern memory 8, the 1-bit width on the X-axis is defined as Δx, and the 1-bit width on the Y-axis is defined as Δy. In addition, the optimum values B _Xn (> 0) and B as the gradient for obtaining the reverse character pattern
Set _Yn (> 0).

As shown in FIG. 12 (a), x _s ≤x <
In the range of (x _s + x _e ) / 2, B _X (x−Δx) −B _X
The minimum x that satisfies (x)> B _Xn is defined as x _ds . (X
_s + x _e ) / 2 ≦ x ≦ x _e , B _X (x + Δx)
The maximum x that satisfies −B _X (x)> B _Xn is defined as x _de . Further, as shown in FIG. _{12 (b), (y s} + y e)
/ 2 ≦ y <in the range of _{y s, B Y (y +} Δy) -B
_The maximum y satisfying _Y (y)> B _Yn is defined as y _ds . y
in the range of _{e ≦ y ≦ (y s +} y e) / 2, B Y (y-Δ
The minimum y that satisfies y) −B _X (y)> B _Yn is defined as y _de . This is the reverse character pattern X of the blank character pattern.
It is based on the fact that black is reduced due to the reverse character pattern at the start point and the end point on the axis and the Y-axis, and there is a certain or greater gradient between adjacent contours. The starting point is (x _s , y _s ),
A rectangular area whose end point is (x _e , y _e ) and whose start point is (x _ds ,
y _ds ), a rectangular area whose end point is (x _de , y _de ) is set on the outline character pattern, and the area between the outer rectangular area and the inner rectangular area is white (bit value is 0). , The inner rectangular area is ANDed with black (the bit value is 1).

The character pattern white / black reversing means 12 _creates a rectangular area in the reversed character pattern memory 11 with respect to each white character pattern as a rectangular area having a start point (x _ds , y _ds ) and an end point (x _de , y _de ). , And 0 and 1 in this area are inverted. If the character pattern recognition means 13 is a white character pattern,
The reversed character pattern is read from the reversed character pattern memory 10, the character pattern is recognized by referring to the recognition dictionary, and if it is not a blank character pattern, the character pattern is read from the character pattern memory 6 and the character is recognized. As described above, according to the second embodiment, there are the same advantages as the first embodiment. Note that the present invention is not limited to the above embodiment, and various modifications are possible. For example, there are the following modifications.

(1) The outline character may be white or dropout color as long as it can be binarized into white and black. (2) FIG. 13 is a diagram showing an example of the outline of an outline character pattern. In the first and second embodiments, the contour shape of the background of the cut-out white character is limited (a perfect circle), but if the contour shape and size are (a) and (b) below, As shown in FIG. 13, it can be applied to outline characters of various shapes (outline). (A) The shape and size are such that the white character pattern determining means 7 can distinguish between a white character pattern and a pattern other than the white character. (B) Reversed character pattern cutout means 1 of the first embodiment
In the processing of 0, the characterless white character pattern corresponding to the white character pattern output from the white character pattern discriminating means is defined as a characterless white character pattern distribution value. Even if it is assumed that the shapes and sizes differ from each other so that only one can be selected from the width of the distribution and the width of the distribution, or that the outline of the outline character is all black, the second embodiment X _ds , x _de , y _ds , y _de must have a shape and size that cannot be determined even if the reverse character pattern cutout means 20 is executed. (3) In the characterless pattern distribution value memory 9 of the first embodiment, only one characterless blank pattern having the same shape is recorded, and the starting point (x _s , relative y _s), the starting point of the character without pattern _{(x s ', y s'} ) is a character without patterns to match its starting point after translation, to compare the inputted white pattern You may

[0024]

As described in detail above, the first to fifth embodiments
According to the invention, since the white character pattern discriminating means, the reverse character pattern cutting means, and the character pattern white / black reversing means are provided, it is not necessary to create a white character recognition dictionary, and the dictionary capacity is reduced. You can

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an optical character reading device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an example of a document pattern.

FIG. 3 is a diagram showing an example of a cut-out character pattern.

FIG. 4 is a projection view of white character patterns on the X axis and the Y axis.

FIG. 5 is a diagram showing the processing contents of an outline character pattern determination means.

FIG. 6 is a diagram showing the contents of a characterless pattern distribution value memory.

FIG. 7 is a diagram illustrating processing contents of a reverse character pattern cutout unit according to the first embodiment.

FIG. 8 is a diagram showing processing contents of a reverse character pattern cutout unit.

FIG. 9 is a diagram showing a processing result of a reverse character pattern cutout unit.

FIG. 10 is a diagram showing a processing result of a character pattern white / black inverting means.

FIG. 11 is a functional block diagram of an optical character reading device according to a second embodiment of the present invention.

FIG. 12 is a diagram illustrating processing contents of a reverse character pattern cutout unit according to the second embodiment.

FIG. 13 is a diagram showing an example of the outline of an outline character pattern.

[Explanation of symbols]

1 image input means 2 image memory 3 line cutout means 4 line memory 5 character pattern cutout means 6 character pattern memory 7 white character pattern judgment means 8 white character pattern memory 9 no character pattern distribution value memory 10, 20 reverse character pattern cutout Means 11 Reversed character pattern memory 12 Character pattern White / black Reversed means 13 Character pattern recognition means

Claims (5)

[Claims] 1. An image input means for inputting an image of a document written on a paper whose ground color is white or dropout color to an image memory, and one character based on the image of the document input to the image memory. In an optical character reading device including a character pattern cutout unit for cutting out a character pattern and a character pattern identification unit for recognizing the cutout character pattern of one character by referring to a recognition dictionary, the character pattern is A white character pattern determination means for determining whether or not a white character pattern is drawn in a black contour of a predetermined shape, and prerecorded data of a white character pattern without characters in which a plurality of contours are all black Characterless pattern distribution value memory to be stored, comparing the characterless white character pattern and the white character pattern recorded in the characterless pattern distribution value memory, A characterless white pattern corresponding to the blank character pattern is selected, and a reverse character pattern of an area including the blank character pattern is selected based on the blank character pattern and the blank blank pattern corresponding to the blank character pattern. And a character pattern white / black reversing means for reversing an image of the reversed character pattern from white to black and from black to white. In the case of a pattern, the optical character reading device is characterized in that character recognition is performed from the character pattern reversed by the character pattern white / black reversing means. 2. An image input means for inputting an image of a document written on a paper whose ground color is white or dropout color to an image memory, and one character based on the image of the document input to the image memory. In an optical character reading device including a character pattern cutout unit for cutting out a character pattern and a character pattern identification unit for recognizing the cutout character pattern of one character by referring to a recognition dictionary, the character pattern is White character pattern determining means for determining whether or not a white character pattern is drawn in a black contour of a predetermined shape, and a reverse character pattern cutout for cutting out a reverse character pattern of an area including the white character pattern. An optical character reading device comprising: a means and a character pattern white / black reversing means for reversing an image of the reversed character pattern from white to black and from black to white. 3. The white character determining means is a distribution value B _X (x), B _Y (y) representing the number of black or white pixels obtained by projecting a character pattern on the X axis and the Y axis, and X. 3. The optical character reading device according to claim 1, wherein it is determined whether the character pattern is a white character based on the widths D _x and D _y of the distribution in the direction and the Y direction. 4. The characterless pattern distribution value memory stores a characterless white pattern in which a plurality of contours are all black on the X-axis,
And distribution values B _X '(x) and B _Y ' (y) representing the number of black or white pixels projected on the Y-axis, and distribution widths D _x 'and D _Y ' in the X and Y directions. The reverse character pattern cutout means stores the distribution values B _X (x) and _BY (y) of the white character patterns output from the white character pattern determination means, and the distribution widths D _X and D. _Y and the distribution value B _X '(x), B _Y ' of the white characterless pattern recorded in the characterless pattern memory
(y) is compared with the distribution widths D _x '(x) and D _Y ' (y), and the distribution value B _X '(x) of the character-free white pattern corresponding to the read white characters. , B _Y ′ (y) and a width D _x ′ (x) and D _Y ′ (y), and a difference between the B _X (x) and the selected B _X ′ (x), 2. The reverse character pattern is cut out by obtaining the area of the reverse character pattern from the result of obtaining the difference between the B _Y (y) and the selected B _Y '(y). Optical character reader. 5. The inversion character pattern cutout means extracts an area of the inversion character pattern from the gradient of the distribution values B _X (x) and B _Y (y) of the white character pattern output from the outline character pattern determination means. 3. The optical character reading device according to claim 2, wherein the reversed character pattern is cut out in order to obtain it.