JP3145264B2 - Character extraction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character extracting device for extracting a character pattern from a form image pattern.

[0002]

2. Description of the Related Art In conventional character recognition, a character pattern is cut out in units of one character in order to collate the character pattern with a recognition dictionary. In the case of a form in which a character frame defining an entry area for one character is printed, the position of each character frame is known in advance, and this frame position can be used as a cutout position. Does not need to be performed. On the other hand, in the case of a form in which a character column that defines an entry area for a plurality of characters is printed but a character frame is not printed, a cutout position is set for each character described in the character column. Requires a process for detecting.

A conventional apparatus for cutting out a character pattern of a form on which such a character frame is not printed is disclosed, for example, in Japanese Patent Application Laid-Open No. 61-195474.

[0004]

However, in the conventional apparatus disclosed in the above-mentioned publication, a) a basic pattern is detected from a character string pattern including a plurality of character patterns; C) detecting a matching basic pattern; c) for a basic pattern that does not match the cut-out dictionary, obtains a cut-out evaluation value for a single basic pattern and individually obtains a cut-out evaluation value for all possible combinations of basic patterns; C) The extraction position is detected by a process of determining the optimum extraction position with reference to the extraction evaluation value. As described above, the process of matching the basic pattern with the extraction dictionary, the process of obtaining the extraction evaluation value for the basic pattern alone, and the process of obtaining the extraction evaluation value for all possible combinations of the basic patterns are necessary. The cutout position detection process becomes complicated, and the processing speed is reduced. In addition, a large-capacity memory is required to store the extraction dictionary.

An object of the present invention is to provide a character extracting apparatus which can detect an extracting position by a simpler process in order to solve the above-mentioned conventional problems.

[0006]

In order to achieve this object, a character extracting apparatus according to the present invention provides an image pattern of a form having an entry area in which a character to be extracted is entered and an attribute determination area corresponding to the entry area. Then, in a character extraction device that extracts an extraction target character pattern, a peripheral distribution is created by scanning an image pattern of an attribute determination area, and the presence or absence of a check of an attribute determination area is detected using the peripheral distribution of the attribute determination area to detect A cutout allowable number setting means for setting an upper limit value and a lower limit value of the cutout allowable number corresponding to the detection result, and a character element pattern is detected from the image pattern of the entry area, and provisional cutout is performed using the detected position of the character element pattern. Set the position and calculate the total number of temporary cutout characters in the entry area using this temporary cutout position. Is larger than the lower limit of the allowable number of cutout characters, the temporary cutout position is corrected, and the total number of temporary cutout characters is calculated again using the corrected position. When the number is equal to or less than the upper limit of the number and equal to or greater than the lower limit of the allowable number of cutouts, a cutout position determining unit that determines a tentative cutout position that obtains the tentative cutout character total number as a target pattern cutout position, Pattern reading means for extracting a character pattern to be extracted using the target pattern extraction position.

[0007]

According to the invention having such a configuration, the use of the form utilizes the fact that the following situations 1) and 2) exist.

1) A person who fills out a form writes a character to be cut out in a writing area and checks one or a plurality of attribute discrimination areas selected according to the attribute of the character to be cut out. Therefore, the extraction target character has an attribute according to whether or not the attribute determination area is checked.

2) There is a correlation between the total number of characters to be extracted to be written and whether or not the attribute discrimination area is checked, and the total number of characters to be extracted to be written is based on this correlation. it can be the previously accumulated upper limit value G _max and the lower limit value G _min.

A typical example of a form in which such situations 1) and 2) exist is a form in which an address is entered.

[0011] For example, it has an entry area in which the name of the prefecture where the entrant lives is entered and four attribute discrimination areas corresponding to the entry area, and the characters of the capital, the road, the prefecture, and the prefecture are individual. Consider a form printed in the attribute determination area. In this case, only one of the four attribute determination areas is checked.
If the resident is a resident of Tokyo, the user enters Tokyo as a character to be cut out of the entry area, and checks the attribute determination area where the capital is printed. When the attribute discrimination area where the capital is printed is checked, the attribute discrimination area where the road is printed is checked, and the attribute discrimination area where the prefecture is printed is checked. In each of the above, the upper limit value G _max and the lower limit value G _min of the total number of cut-out target characters to be written are G _max = G _min = 2.
Individual. When the attribute discrimination area in which the prefecture is printed is checked, the character to be extracted is Wakayama, Saitama or another prefecture name, so the upper limit G _max and the lower limit G _min are G _max = 3 And G _min = 2. Thus, the presence or absence of attribute discrimination area, there is a correlation between the upper limit value G _max and the lower limit value G _min, on the basis of this correlation, the upper limit value G _max and the lower limit value G _min as data Can be stored.

[0012] As illustrated herein, the total number of cut object character to be entered is the upper limit value G _max or less was it and the lower limit value G _min or more, the total number of cut object character to be entered is, the upper limit It does not become larger than the value _Gmax and does not become smaller than the lower limit _Gmin . Therefore, the total number of temporary cut characters obtained using the temporary cut position is
Whether by determining whether it is within the numerical range of the upper limit value G _max and the lower limit value G _min, the cropping position of the temporary target pattern or extraction position (cut target character extraction position) for use is properly or it can be determined, which the upper limit value G _max and the lower limit value G _min, it is possible to use the upper and lower limits of the cut-out tolerance number.

Further, the peripheral distribution of the attribute determination area has a large value when the area is checked, and has a small value when the area is not checked. By utilizing the fact that the distribution changes, the presence or absence of the attribute determination area can be detected. Therefore, the presence or absence of the check of the attribute determination area is detected,
The upper limit value and the lower limit value of the cutout allowable number corresponding to the detection result are stored in the upper limit value G _max stored as the above-described data.
And the lower limit _Gmin .

[0014]

Embodiments of the present invention will be described below with reference to the drawings. The drawings are only schematically shown to the extent that the invention can be understood, and thus the invention is not limited to the illustrated examples.

FIG. 1 is a functional block diagram schematically showing an entire configuration of an embodiment of the present invention. The character extracting apparatus 10 of this embodiment shown in FIG. 1 includes an image storage unit 12, a format storage unit 14, an allowable number of extraction units setting unit 16, an extraction position determination unit 18, and a pattern reading unit 20.

The image storage means 12 stores an image pattern of a form having an entry area in which a character to be extracted is entered and an attribute determination area corresponding to the entry area. The person who fills out the form indicates the information to be transmitted by the character to be cut out of the entry area, and the attribute of the transmission information by the presence or absence of the check of the attribute determination area.

The format storage means 14 stores at least the address of the area storing the image pattern of the entry area and the address of the area storing the image pattern of the attribute determination area as format information.

The cutout allowable number setting means 16 scans the image pattern of the attribute discrimination area to create a peripheral distribution, detects whether or not the attribute discrimination area is checked using the peripheral distribution of the attribute discrimination area, and performs the detection. The upper limit and the lower limit of the allowable number of cutouts corresponding to the result are set.

The cutout position determining means 18 detects a character element pattern from the image pattern of the entry area, and sets a temporary cutout position using the detected position of the character element pattern. Further, the cut-out position determining means 18 calculates the provisional cut-out character total number in the entry area using the provisional cut-out position, and when the provisional cut-out character total number is larger than the upper limit of the cut-out allowable number and the cut-out allowable number. Is smaller than the lower limit value, the provisional clipping position is corrected, and the total number of provisional clipping characters is calculated again using the correction position. When the number of cut-out characters exceeds the lower limit of the cut-out allowable number, the tentative cut-out position at which the tentative cut-out character total number is obtained is determined as the target pattern cut-out position.

The pattern extracting means 20 extracts a character pattern to be extracted using the target pattern extracting position.

(Form) FIG. 2 shows an example of a form used in this embodiment. In this embodiment, the form 22 is an example of a form in which an address is entered as information transmitted to a form writer, and the form 22 has an entry area for entering the name of a city, a road, a prefecture, or a prefecture. 241 and four attribute determination areas 261 corresponding to the area, an entry area 242 for entering the name of a ward, a city or a county, and three attribute determination areas 262 corresponding to the area, a ward, a town or Entry area 2 for entering the name of the village
43 and three attribute determination areas 263 corresponding to the area
And In the figure, these entry areas 241, 242, 2
The positions of the reference numeral 43 and the attribute determination areas 261, 262, 263 are indicated by dashed lines, but the dashed lines are not printed on the form 22.

On the other hand, the form 22 is printed with a line representing an address entry column, here a solid line 28, a line dividing the address entry column, here a dotted line 30, and an item name of an address column.

The writing area 241, the attribute determining area 261, the writing area 242, the attribute determining area 262, the writing area 243, and the attribute determining area 263 are sequentially arranged along the character string direction X, and these areas 241, 242, 243 are arranged. 261
262 and 263 are surrounded by a solid line 28. Then, a space between adjacent entry areas 241 and 242 is separated by a dotted line 30, and
The adjacent writing areas 242 and 243 are separated by another dotted line 30.

Further, the form 22 has an attribute determining character 321 in each attribute determining area 261 of the entry area 241.
Is printed. Here, different types of attribute determination characters 321 are printed on the respective attribute determination regions 261, and these four types of attribute determination characters 321 are a city, a road, a prefecture, and a prefecture. The attribute determination area 261 is the area 2
61 includes a margin area around the attribute discrimination character 321 printed.

Similarly, an attribute determination character 322 is printed in each attribute determination area 262 of the entry area 242. Here, different types of attribute determination characters 322 are printed in the respective attribute determination regions 262, and these three types of attribute determination characters 322 are ward, city, and county. The attribute determination area 262 includes a margin area around the attribute determination character 322 printed in the area 262.

Each attribute discrimination area 26 of the entry area 243
3, the character 323 for attribute determination is printed.
Here, different types of attribute determination characters 323 are printed in the respective attribute determination regions 263, and these three types of attribute determination characters 323 are ward, town, and village. The attribute determination area 263 includes a margin area around the attribute determination character 323 printed in the area 263.

The person who fills out the form 22 according to the address,
An extraction target character 341 representing the name of a city, a road, a prefecture or a prefecture is entered in the entry area 241, and the attribute determination character 321 corresponding to the target character 341 is printed on the attribute determination area 26.
A check 361 is entered in 1. Similarly, the extraction target character 342 representing the name of a ward, a city or a county is entered in the entry area 2.
42, and a check 362 is entered in the attribute determination area 262 in which the attribute determination character 322 corresponding to the target character 342 is printed. Further, a cut-out target character 343 representing the name of a ward, a town, or a village is entered in the entry area 243, and a check 363 is entered in the attribute determination area 263 on which the attribute determination character 323 corresponding to the target character 343 is printed. .

For example, in the illustrated example, the character 341 to be cut out
2 characters of Tokyo representing the name of the city, 3 characters to be extracted
42, two characters XX representing the name of the city, three characters 切 representing the name of the town as the extraction target character 343, and 町 as the checks 361-363.

(Image Storage Means) In this embodiment, the image storage means 12 inputs the image pattern of the form 22 from the scanner 38 and stores (stores) the input pattern.

The scanner 38 optically scans the form 22 and converts an optical signal from the form 22 into an electric signal quantized in pixel units. Then, the scanner 38 stores the electric signal in the image storage unit 12 as an image pattern of the form 22. Here, the image pattern is a binary electric signal representing a character or a character background.

FIG. 3 is a diagram for explaining the image pattern of a form. In the figure, an image pattern 40 of the form 22 is shown by being surrounded by a two-dot chain line, and among the image patterns 40, an image pattern representing a character is represented in black and an image pattern representing a character background is represented in white.

Here, the solid line 28 and the dotted line 30 of the form 22
The item name of the address column is printed in a color (dropout color) that cannot be read by the scanner 38, for example, red. Therefore, the solid line 28, the dotted line 30, and the light signal of the item name are image patterns representing a character background. Converted to 40.

The attributes determining characters 321 to 323 are printed in a color readable by the scanner 38, for example, black, and the characters 341 to 343 to be extracted and the check 3
61 to 363 are written in a color readable by the scanner 38, for example, black.
The optical signals of 3, 341 to 343 and checks 361 to 363 are converted into image patterns 40 representing characters.

An XY coordinate system is virtually set on the storage area of the image storage means 12, and these coordinate positions X
The image storage means 12 is configured so that the image pattern 44 at the pixel position represented by Y and Y can be read. The image pattern 44 is stored such that the character string direction X of the form 22 is parallel to the X-axis direction.

For example, the form 22 is set so that the main scanning direction of the scanner 38 is substantially parallel to the character string direction X of the form 22.
Is set on the scanner 38, and the form 22 is optically scanned to store the image pattern 44 so that the character string direction X is parallel to the X-axis direction.

(Format storage means) In this embodiment, the format storage means 14 stores the entry areas 241, 2
Addresses for individually reading out the image patterns 40 of 42 and 243 from the image storage means 12 for each area;
Image pattern 40 of attribute determination areas 261, 262, 263
And an address for reading from the image storage unit 12 for each area. Here, these addresses are represented by coordinate positions X and Y set on the storage area of the image storage unit 12 (hereinafter, these addresses are represented as addresses X and Y).

Furthermore format storage unit 14 stores the constants A _n for normalizing the marginal distribution of the attribute discrimination region 261, 262, 263, and a threshold value THL1 for detecting a character element pattern.

(Allowable Extraction Number Setting Means) In this embodiment, the allowable extraction number setting means 16 includes the margin distribution creation means 1.
6a, a check area detecting means 16b and an allowable number storing means 16c.

The margin distribution creating means 16a creates margin distributions individually for each attribute discrimination area, and normalizes each margin distribution.

The check area detecting means 16b detects, for each entry area, the peripheral distribution which is the largest among the normalized peripheral distributions, and determines the attribute discriminating area where the normalized peripheral distribution is the largest. Is detected as an attribute discrimination area with a check for the entry area corresponding to. For each entry area, the attribute discrimination area in which the normalized peripheral distribution is the largest is determined as a checked attribute discrimination area (the checked attribute discrimination area), and the attribute in which the normalized peripheral distribution is not the largest. The determination area is determined as an attribute determination area that is not checked (an attribute determination area without a check). Then, the check area detecting means 16b sets the upper limit values G _max and G _min of the cutout allowable number corresponding to the presence or absence of the check for each entry area.

The permissible number storage means 16c stores the upper limit value G _max and the lower limit value G _min of the permissible number of cutouts corresponding to the detection result of the check for each entry area. , The upper limit G _max and the lower limit G _min corresponding to the presence or absence of the check,
6c.

The detection of the presence / absence of the check will be described more specifically with an example. Here, the description will focus on the attribute determination area 263. 4 and 5 are diagrams for explanation. FIGS. 4A and 4B show the attribute discrimination area 26 when there is no check and when there is a check.
3A and 3B are diagrams showing image patterns, and in these figures, image patterns showing characters and character backgrounds are shown in the same manner as in FIG. 3. FIGS. 4 and 5 show the attribute discrimination area 26 when there is no check and when there is a check.
3 is a diagram showing the peripheral distribution of No. 3 in which the horizontal axis represents the sub-scanning position Y and the vertical axis represents the cumulative number of character pixels f _n (Y)
Is shown.

In the form 22 of this embodiment, three attribute discrimination areas 263 are set for the entry area 243, and the attribute discrimination areas 263 having different wards, towns and villages are used as the attribute discrimination characters 323. Printed inside. Here, the attribute determination area 263 on which the ward, town, and village are printed
Are the first, second, and third attribute determination areas 263, respectively.

The start and end positions of the attribute discrimination area 263 in the character string direction X are X _L and X _R , and the attribute discrimination area 263 in the direction Y intersecting the character string direction X.
Expressed in the position of the start and end with Y _T and Y _B, In the first-th attribute discrimination region _{263, X L = X1, X} R
= X2, and Y _T = Y1 and _{Y B = Y2, X1 ≦ X} ≦ X
The range of 2 and Y1 ≦ Y ≦ Y2 is defined as addresses X and Y for reading the image pattern 40 in the first attribute determination area 263. Also In the first second attribute determination area _{263, X L = X1, X} R = X2, Y T = Y2 and Y _B
= Y3, and the range of X1 ≦ X ≦ X2 and Y2 ≦ Y ≦ Y3 is defined as the image pattern 40 of the second attribute determination area 263.
Are the addresses X and Y for reading out the data. Further, in the third attribute determination area 263, X _L = X1, X _R
= X2, Y _T = Y3 and Y _B = Y4, and X1 ≦ X ≦ X
The range of 2 and Y3 ≦ Y ≦ Y4 is defined as addresses X and Y for reading the image pattern 40 in the third attribute determination area 263.

First, the marginal distribution creating means 16a determines the n-th (n is a natural number satisfying n ≧ 1; here, n = 1,
The addresses X and Y of the attribute determination area 263 of (2) and (3) are read from the format storage unit 14, and the image pattern 4 of the attribute determination area 263 corresponding to the addresses X and Y is read.
0 is read from the image storage unit 12.

Next, the peripheral distribution creating means 16a scans the image pattern 40 in the n-th attribute discrimination area 263 with the main scanning direction as the character string direction X and the sub-scanning direction as the direction Y intersecting with the character string direction X. , The cumulative number of character pixels f _n (Y) on the scanning line is determined for each sub-scanning position Y. Cumulative character pixel number f _n
(Y) is the total number of character pixels existing on the scanning line at the sub-scanning position Y and existing in the n-th attribute determination area 263.

Next, the peripheral distribution creating means 16a obtains a peripheral distribution Δf _n (Y) dY of the n-th attribute discrimination area 263.
The marginal distribution ∫f _n (Y) dY is the n-th attribute determination area 263
Number of character pixels from the start position Y _T to the end position Y _B
It is the sum of f _n (Y).

[0048] Then the peripheral distribution creation unit 16a is the constant A _n for normalizing the marginal distribution ∫f _n (Y) dY of the n-th attribute discrimination region 263, read from the format storage unit 14, the n-th Distribution around the attribute discrimination area 263 of
By normalizing f _n (Y) dY with a constant _An , a normalized marginal distribution 1 / A _n · ∫f _n (Y) dY is obtained.

The normalization constant A _n for normalizing the marginal distribution [integral] F _n (Y) dY is represented by the following equation (Equation 1).

[0050] _{1 / A n · ∫F n (} Y) dY = C ···· ( number 1) However, ∫F _n (Y) dY: n-th attributes previously obtained with no added check Peripheral distribution ∫f _n (Y) dY C of the discrimination area C: a constant that is a positive integer In the form 22 of this embodiment, n = 1, 2, 3 for the attribute discrimination area 263, so that n = 1, 2, Rewriting (Formula 1) as 3 gives the following formula (Formula 2).

1 / A ₁ · ∫F ₁ (Y) dY = 1 / A ₂ · ∫F ₂ (Y) dY = 1 / A ₃ · ∫F ₃ (Y) dY = C (Equation 2) Check Each attribute discriminating area 2 without 363
The marginal distribution ∫F _n (Y) dY is obtained for each 63, and each marginal distribution ∫F _n (Y)
A normalization constant _An that makes dY equal to the constant C is obtained. As described above, the normalization constant _An is individually obtained in advance for each attribute determination area 263, and is stored in the format storage unit 14 in advance.

Next, the check area detecting means 16b calculates the normalized marginal distribution 1 / A _n · ∫f _n obtained for each attribute discrimination area 263.
(Y) From dY, the largest normalized marginal distribution 1 / A _n
n (Y) dY is detected. Then, the check area detecting means 16
b indicates the attribute discrimination area 263 in which the normalized marginal distribution 1 / A _n · ∫f _n (Y) dY is the maximum, and the attribute discrimination area 243 corresponding to the discrimination area 263 is marked with a check 363. It is detected as an area 263.

In the form 22 used in this embodiment,
A plurality of attribute determination areas 26 provided for the entry area 243
A check 363 is attached to any one of the three. Therefore, with respect to the entry area 243, the normalized marginal distribution 1 / A _n
∫f _n (Y) The attribute discrimination area 263 having the maximum dY is detected as the attribute discrimination area 263 with the check 363 (the attribute discrimination area with the check) 263, and the normalized marginal distribution 1 / A _n・ ∫f _n (Y) Attribute discrimination area 26 where dY is not maximum
3 can be detected as an attribute determination area 263 to which no check 263 is attached (an attribute determination area without a check) 263.

The marginal distribution 1 / A _n · ∫f _n thus normalized
(Y) A check 363 is performed for each attribute determination area 263 corresponding to the entry area 243 depending on whether or not dY is the maximum.
Can be detected.

In this embodiment, the marginal distribution ∫f _n (Y) dY
And the normalized marginal distribution 1 / A _n・ ∫f _n (Y)
The attribute determination area 263 where dY is the maximum is checked 263
Is determined to be the attribute determination area 263 marked with.

The marginal distribution 1 / A _n · ∫f _n thus normalized
(Y) Since dY is used, even when the total number of character pixels constituting the attribute determination character 323 is different for each attribute determination area 263, the check 363 protrudes from the attribute determination area 263 to which the check 363 should be attached. Even if it is written in another attribute determination area 263 adjacent to the
The presence or absence of the check 363 in the attribute determination area 263 can be detected with high accuracy.

Similarly, for the entry area 241, a marginal distribution ∫f _n (Y) dY is created for each attribute discrimination area 261 provided for the entry area 241, and the normalized marginal distribution 1 / A _n・ ∫f _n (Y) dY is obtained and normalized marginal distribution
1 / A _n · ∫f _n (Y) dY,
The entry area 241 corresponding to the determination area 261 is detected as an attribute determination area 261 with a check 361 attached. Normalized marginal distribution 1 / A _n · of attribute discrimination area 261
Check 361 depending on whether or not ∫f _n (Y) dY is maximum
Can be detected.

Further, with respect to the entry area 242, a marginal distribution ∫f _n (Y) dY is created individually for each attribute determination area 262 provided for the entry area 242, and the normalized marginal distribution 1 / A _n・ ∫f _n (Y) dY is calculated and normalized marginal distribution
1 / A _n · ∫f _n (Y) dY,
The entry area 242 corresponding to the determination area 262 is detected as an attribute determination area 262 with a check 362 added. Normalized marginal distribution 1 / A _n · of attribute discrimination area 262
Check 362 depending on whether or not ∫f _n (Y) dY is maximum.
Can be detected.

Next, the upper limit values G _max and G of the allowable number of cutouts
The setting of _min will be described more specifically with an example. Here, description will be given focusing on the upper limit values G _max and G _min of the allowable number of cutouts regarding the entry area 241.

When the form 22 of this embodiment is used, the person who fills in the form 22 writes a cut-out target character 341 representing the name of a city, a road, a prefecture, or a prefecture in the entry area 241 and corresponds to the name. Character 321 for attribute discrimination
Check 3 in the attribute discrimination area 261 where prefecture or prefecture is printed
61 will be attached.

Focusing on the upper limit G _max and the lower limit G _min of the total number of characters to be cut out 341 (hereinafter referred to as the total number of characters) to be written in the writing area 241, In the case where the check 361 is added (hereinafter referred to as “checked presence / absence mode 1”), the cut-out target character 341 written in the writing area 241 is Tokyo, so the upper limit G _max and the lower limit G of the total number of characters to be entered are set.
_min is G _max = G _min = 2. When a check 361 is attached to the attribute determination area 261 on which the road is printed (hereinafter, referred to as “check 361”)
In the case 2) with or without the check, the entry area 241
Upper limit value G _max and the lower limit value G _min of the cut-out object character 341 is entered becomes the North Sea thus fill characters total number in the G
_max = _Gmin = 2. In the case where the check mark 361 is added to the attribute discrimination area 261 printed by the government office (hereinafter, whether or not there is a check), the cutout target character 341 to be written in the entry area 241 is Kyoto or Osaka, and thus the input character The upper limit G _max and the lower limit G _min of the total number are G
_max = _Gmin = 2. Further, in the case where a check 361 is attached to the attribute discrimination area 261 on which the prefecture is printed (hereinafter referred to as the presence / absence of check 4), the cutout target character 341 to be entered in the entry area 241 is Wakayama, Saitama, or another character. It is a character representing a prefecture name. Therefore, the upper limit value G _max and the lower limit value G _min of the total number of entered characters are G _max = 3, G _min
= 2.

As described above, the presence or absence of the check 361 in the attribute determination area 261 and the upper limit values G _max and G
There is a correlation that is known in advance between the _min and the _min . Therefore, the upper limit value G _max and the lower limit value G _{min of the} total number of characters to be entered are provided for each aspect of the presence or absence of the check 361 in each attribute determination area 261, here, for each of the aspects 1 to 4 described above.
Can be stored as data.

As will be described later, the character 34 to be cut out
1 In the case of detecting the extraction position, cutout characters total number M of provisional obtained using the extraction position of the provisional, outside the range of values of the upper limit value G _max and the lower limit value G _min of fill characters total number , The provisional extraction position is the extraction target character 3
It can be determined that the cutout position 41 is inappropriate. Also, the total number of temporary cuts M obtained using the temporary cut positions
Is within the range of the upper limit value G _max and the lower limit value G _min of the total number of entered characters corresponding to the presence / absence of the check 361 in the attribute determination area 261, the provisional extraction position is the extraction position of the extraction target character 341. Can be determined to be appropriate.

Accordingly, the upper limit value G _max and the lower limit value G _{min of the} total number of entered characters obtained based on the previously known correlation.
With the upper limit value G _max and the lower limit value G of the cutout allowable number described above.
Can be used as _min .

The upper limit value G _max and the lower limit value G _min of the permissible number of cut-outs that have been determined in advance are classified according to the presence or absence of each check and stored in the permissible number storage means 16c.

When the check area detecting means 16b obtains the detection result of the presence / absence of the check regarding the entry area 241 (that is, the mode of the check), the upper limit value G _max and the lower limit value G _{min of} the cutout allowable number corresponding to the detection result. From the allowable number storage means 16c, and the read upper limit G
The _max and the lower limit value G _min, is set as cut allowed number of upper limit value G _max and the lower limit value G _min relating to the entry region 241 (storage).

Similarly, regarding the entry area 242,
The upper limit value G _max and the lower limit value G _min of the permissible number of cutouts that have been determined in advance are classified according to each check mode, and stored in the permissible number storage unit 16c. When the check area detection unit 16b obtains the detection result of the presence / absence of the check regarding the entry area 242 (that is, the mode of the check),
The upper limit value G _max and the lower limit value G _min of cut allowable number corresponding to the detection result, reading from the allowable number storage unit 16c reads the upper limit value G _max and the lower limit value G _min, cut allowable number relating to the entry region 242 _Are set (stored) as the upper limit value _Gmax and the lower limit value _Gmin .

Regarding the entry area 243, the upper limit value G _max and the lower limit value G _{min of the} number of permissible cutouts that have been determined in advance are known.
Are classified according to each check mode, and stored in the allowable number storage unit 16c. Then check area detecting unit 16b obtains the detection result of the presence or absence of the check relates entry region 243 (i.e., aspects of the check), the upper limit G _max and the lower limit value G _min of cut allowable number corresponding to the detection result, the allowable The upper limit value G _max and the lower limit value G _min read from the number storage unit 16 c are stored in the entry area 2.
Upper limit value G _max and lower limit value G of the permissible number of cuts related to 43
Set as _min .

(Cutout Position Determination Means) In this embodiment,
The cutout position determining means 18 includes a character element detecting means 18a,
Pitch estimation means 18b, cut-out parameter storage means 18
c, end position detecting means 18d and position setting controlling means 18
e.

The character element detecting means 18a calculates the start position X of the character element pattern in the character string direction X for each entry area.
Detecting the _L and end position X _R. The character element pattern is an image pattern of an area in which character pixels are connected, and the image pattern of a character to be extracted, that is, the target character pattern includes one or a plurality of character element patterns. Here, only one line of characters to be cut out is entered in one entry area.

The pitch estimating means 18b uses the start position X _L and the end position X _R of the character element pattern to set the maximum character element width W _B of the character element width W _B in the character string direction X in each entry area. _Bmax and the minimum spacing W _Smin of the character element spacing W _S in the character string direction X are obtained, and the maximum width W _Bmax and the minimum width W are determined for each entry area.
The sum of _Smin is set as the initial value of the estimated character pitch p.
Character element width W _B is the character element pattern width, character element interval W
_S is the space between adjacent character element patterns, that is, the width of a margin pattern sandwiched between adjacent character elements. The margin pattern is an image pattern of an area in which character background pixels are connected.

The extraction parameter storage means 18c reads out and reads the start position X _L and end position X _R of the character element pattern, the estimated character pitch p, the provisional extraction start position X _S and the provisional extraction end position X _E , respectively. Remember freely for rewriting.

The end position detecting means 18d calculates a position separated from the tentative cut-out start position X _S by the estimated character pitch p in the character cut-out direction as a tentative cut-out end position X _E.

The position setting control means 18e sets a position separated by ΔX (ΔX is a positive integer) in the character extraction direction from the temporary extraction end position X _E as the next temporary extraction start position X _S. I do. The minimum starting end position X _L of the starting end position X _L of the character element pattern existing in the same entry area when the character extraction direction is a positive direction, the cut-out starting position X _S of the first formal, character segmentation direction when the negative direction up to the end position X _R of the end position of the character element pattern existing in the same entry area, and clipping start position X _S of the first formal.

[0075] The positioning control unit 18e, when it cut the end position X _E of the provisional end position detecting means 18d has calculated the position of the character element region (excluding the case where the where X _E = X _R) a temporary cut ending position X _E, corrects the end position X _R position or the character element area of the character element between the region adjacent to the character element region. The character element area is an area where a character element pattern exists, and the area between character elements is an area where adjacent character element patterns are interposed, that is, an area where a margin pattern exists.

Further, the position setting control means 18 e
For each area, a temporary cutout start position X_SAnd / or temporary cutout
The total number of detected end positions is the total number of cutout characters in the entry area
M, and for each entry area, the total number of cutout characters M
And the upper limit G of the allowable number of cutouts _max And lower limit G_min With
The following processes 1) to 3) are performed according to the comparison result.
U.

[0077] Process 1); when cut characters total number M is smaller than the lower limit value G _min of cut allowable number adds a positive correction value Δp sets a new estimated character pitch p of the estimated character pitch p the new by using the estimated character pitch p to reset the cut-out start position X _S and cutout end position X _R provisional provisional restarting the end position detection means 18d.

[0078] Process 2); when cut characters total number M is larger than the upper limit value G _max of cut allowable number adds a negative correction value Δp of the estimated character pitch p is set to a new estimated character pitch p the new by using the estimated character pitch p to reset the cut-out start position X _S and cutout end position X _R provisional provisional restarting the end position detection means 18d.

Process 3); When the total number M of cut-out characters is equal to or more than the lower limit value G _{min of the} allowable number of cut-outs and equal to or less than the upper limit G _{max of the} allowable number of cut-out characters, a temporary cut-out start position at which the total number M of cut-out characters is obtained. X _S and temporary cut-out end position X _E
Is determined as the target pattern cutout position.

Next, the flow of the operation of the cut-out position determining means 18 will be described more specifically by way of an example. 6 and 7 are diagrams for explanation. FIG. 6 is a diagram for explaining the detection of the character element pattern and the setting of the estimated character pitch. In FIG. 6A, the image pattern of the entry area 343 is shown in the same manner as in FIG. It is. FIG.
In the diagram (B), the horizontal axis indicates the sub-scanning position X and the vertical axis indicates the cumulative number of character pixels f _n (X) at the sub-scanning position X.
The distribution state of the accumulated character pixel number f _n (X) in the entry area 343 is shown. FIG. 7 is a diagram showing the flow of the operation focusing on the position setting control unit 18e.

First, the character element detecting means 18a reads the addresses X and Y of the writing area 243 and the threshold value THL1 from the format storage means 14, and then reads the writing area 24.
One image pattern 40 is read from the image storage unit 12 using the addresses X and Y. Here, the entry area 24
3, the starting end position X _L and the end position X _R in the character string direction X X _L = X3 and X _R = X4, the starting end position Y _T and end position Y _B in the direction Y perpendicular to the character string direction X Y
X _L ≦ X ≦ X _R and Y, where _T = Y5 and Y _B = Y6
A region in the range of _T ≦ Y ≦ Y _B, the entry region 24
3 addresses X, Y and X _L ≦ X ≦ X _R and Y _T ≦ Y ≦ Y
_B.

Next, the character element detecting means 18a sets the image pattern 40 of the writing area 243 as the direction Y perpendicular to the character string direction X and the sub-scanning direction as the character string direction X.
To obtain the cumulative number of character pixels f _n (X) on the scanning line for each sub-scanning position X. The cumulative number of character pixels f _n (X) is the total number of character pixels existing on the scanning line at the sub-scanning position X and existing in the writing area 243.

Next, the character element detecting means 18a compares the cumulative number of character pixels f _n (X) with the threshold value THL1 for each sub-scanning position X, and determines the area _where f _n (X)> THL1 as a character element area and a THL1 area. the f _n (X) ≦ THL1 become region regarded as the character elements between the regions, and the sub-scanning position X when the changes from between characters element regions in the character element region and start position X _L of the character element regions and character detecting the sub-scanning position X when the change between characters element regions from the element region as a terminal position X _R of the character element region. Then, the character element detecting means 18 a
Each character element each area of the inner, starting end position X _L and the end position X
_R is stored in the extraction parameter storage unit 18c. FIG.
In the is shown a starting position X _L and X _R is detected when the THL1 = 0.

Next, the pitch estimating means 18b sets the entry area 2
Request and each character element width W _B in 43 and the character element spacing W _S in entry region 243. Character element width W _B is equal to the distance of the starting position X _L and the end position X _R of the character elements region, character element spacing W _S is equal to the distance between the character element region adjacent phases.

Next, the pitch estimating means 18b detects the maximum width W _Bmax of the character element widths W _B in the writing area 243, and is adjacent to the character element pattern in which the maximum width W _Bmax is obtained in the writing area 243. The minimum interval W _Smin among the character element intervals W _S is detected, and the sum of the maximum width W _Bmax and the minimum interval W _Smin is obtained as the estimated character pitch p. The pitch estimating means 18b calculates the estimated character pitch p
Is stored in the extraction parameter storage means 18d.

When the pitch estimating means 18b has stored the estimated character pitch p, the position setting control means 18e reads the starting position X _S of the character element area of the writing area 243 from the cut-out parameter storing means 18c. Then, the position setting control unit 18 e determines the minimum start position X _S in the entry area 243.
Detects, and stores the cut parameter storage unit 18c the minimum starting end position X _S as the first cut-out start position X _S of the temporary entry area 243, thereafter, the end position detecting means 1
8d is started (start of FIG. 7).

The activated end position detecting means 18d stores the first temporary cut start position X _S and estimated character pitch p with respect to the entry area 243 from the cut parameter storage means 18c.
Reading the door, X as cut start position X _E of the first formal
To calculate the _{S = X s + p-1} .

Next, the position setting control means 18e checks whether the provisional cut end position X _E calculated by the end position detecting means 18d is a position in the character element area or a position in the inter-character element area. and stores the provisional cut end position X _E in accordance with the test result to the cut parameter storage unit 18c. Provisional cut end position X _E is the case where the position of the inter-character area stores as it is cut parameter storage unit 18c without correcting the end position X _E. In the case the tentative cutout end position X _E is the position of a character region, a temporary cut end position X _E to the position of the inter-character region adjacent to the end position X _R or the character region of the character regions It corrected, cut the corrected temporary cut end position X _E are stored in the parameter storage unit 18c (S1 in FIG. 7).
Next, the position setting control unit 18e determines the total number M of cutout characters.
By adding 1 to (the initial value of M is M = 0), the total number M of cut-out characters is counted (S2 in FIG. 7).

Next, the position setting control unit 18e determines whether the setting of the temporary cutout position has been completed for the entry area 243 (S3 in FIG. 7). Compared with the maximum end position X _Rmax of the end position X _R of the character element regions of the cutout end position X _E in writing area 243, if the X _E <X _Rmax is determined not ended and configuration, X _E ≧ If X _Rmax is reached, it is determined that the setting has been completed.

[0090] When it is determined set not ended and the position setting control section 18e is X as cut start position X _S of the next temporary
Calculate the _S = X _E + _ΔX, and stored in the parameter storage unit 18c cuts out the cut position X _S of the next temporary, thereafter, starts the end position detecting means 18 d (S4 in FIG. 7). For example, ΔX = 1. Activated end position detecting means 18d
Reads out the next provisional extraction start position X _S from the extraction parameter storage unit 18c, and outputs the next provisional extraction end position X S
Calculate _E. Then positioning control unit 18e performs test and storage of the cut end position X _E of the next provisional end position detecting means 18d is calculated (S1 in FIG. 7).

If it is determined that the setting has been completed, the position setting control means 18e sets the total number M of cutout characters in the entry area 2
43 relates to comparing the lower limit value G _min and an upper limit G _max of cut allowable number (S5 in Fig. 7).

[0092] When cut characters total number M is smaller than the lower limit value G _min (M <may become G _min) position setting control unit 1
8e calculates a pitch obtained by adding the negative pitch correction value Δp to the estimated character pitch p as a new estimated character pitch p,
Estimated character pitch p of cut-out parameter storage means 18c
To the new estimated character pitch p. Thereafter, the position setting control unit 18e activates the end position detecting unit 18d to calculate the first temporary cutout end position (S6 in FIG. 7). Activated end position detecting means 18d
Calculates a first cut end position X _E. Thereafter, positioning control unit 18e performs a test and stored in the first tentative cutout end position X _E where the end position detecting means 18 is calculated (S1 in FIG. 7), hereinafter, cutout parameter storage unit 18c tentative rewriting of the cutout start position X _S and the temporary cut-out end position X _E, the provisional cut start position X _S and the temporary cut-out end position X _E obtained by the new estimated character pitch p.

[0093] cutout characters total number M (if made M> G _max) upper limit G if _max greater than, positioning control unit 18e newly pitch obtained by adding a positive pitch correction value Δp of the estimated character pitch p The estimated character pitch p is calculated as a new estimated character pitch p, and the estimated character pitch p in the cut-out parameter storage unit 18c is rewritten to the new estimated character pitch p. Thereafter, the position setting control unit 18e activates the end position detecting unit 18d to calculate the first temporary cutout end position (S7 in FIG. 7). Activated end position detecting means 18d
Calculates a first cut end position X _E. Thereafter, positioning control unit 18e performs a test and stored in the first tentative cutout end position X _E where the end position detecting means 18d is calculated (in FIG. 7 S1), subsequently, the temporary cut-out parameter storage unit 18c the cut start position X _S and the temporary cut-out end position X _E, rewrites the temporary cut start position X _S and the temporary cut-out end position X _E obtained by the new estimated character pitch p.

When the total number M of cut-out characters is not less than the lower limit value G _{min and not} more than the upper limit value G _max (G _min ≦ M ≦
If made G _max), the position setting control unit 18e is the cut characters total number clipping start position of each provisional give the M X _S and the temporary cut end position X _E, the target pattern extraction position of the entry area 243 In order to cut out the pattern to be cut out of the writing area 243.
0 (S8 in FIG. 7), and then the entry area 24
The processing for detecting the target pattern cutout position related to No. 3 is ended (end of FIG. 7).

Similarly, the cut-out position determining means 18
For the other entry areas 241 and 242, the target pattern cutout position is detected for each entry area.

(Pattern reading means) In this embodiment,
When activated by the cut-out position determining means 20, the pattern read-out means 20 reads out from the cut-out parameter storage means 18c the target pattern cut-out position relating to the entry area for which the detection of the target pattern cut-out position has been detected, and starts the start position of the entry area. It reads the Y _T and the end position T _B from the format storage unit 14. Thereafter, starting end position of these target pattern extraction position entry area Y _T and end position Y _B
By using the above, the extraction target pattern is extracted from the image storage unit 12 and output to the post-processing unit 40, for example, the character recognition unit.

In this embodiment, A) a position substantially apart from the provisional cut-out start position X _S by the estimated character pitch p is set as a provisional cut-out end position X _e, and a position adjacent to the provisional cut-out end position X _E is as cut start position X _S provisional of, sequentially tentative extraction position X _S, detects X _E, B) cut start position of the temporary entry region X _S or provisional cutout end position X
Let _{E be the} total number M of cut-out characters, compare the total number M of cut-out characters with the lower limit G _min and the upper limit G _{max of the} allowable number of cut-outs, and C) according to the comparison result, the temporary cut-out position X _S ,
X _E is determined as the target pattern cut-out position or the estimated character pitch is corrected, and the temporary cut-out position X _S , X
_E is detected.

[0098] Since the decision in accordance with the comparison result of the upper limit value G _max and the lower limit value G _min of the thus cut object pattern extraction position characters total number M and cut allowable number, a simple determination of the target pattern extraction position Can be performed with high precision. Because of this, it is possible to speed up the extraction process of the extraction target character pattern,
In addition to this, there is an advantage that the hardware configuration of the device can be simplified and the size of the device can be reduced.

FIG. 8 is a diagram showing another example of a form. In the form 22 used in the above-described embodiment, the kanji of the prefecture or the like is used as the attribute discrimination character. However, as shown in FIG. 8, the total number of characters to be entered in the entry area is used as the attribute discrimination character. May be used. For example, FIG.
In the example of the above, when the writer enters three characters of Tokyo as the extraction target character 341 in the entry area 241,
A check 361 is added to 3 as the character 361 for attribute determination. As described above, even when the attribute discrimination character representing the total number of characters to be entered in the entry area is used, the target pattern cutout position can be detected.

The present invention is not limited to the embodiment described above, and various changes can be made within the scope of the present invention.

For example, as can be understood from FIG. 4B and FIG. 5B, the individual marginal distribution Δf _n (Y) dY in the state where the check 363 is added is checked 363. The individual marginal distribution ∫f _n (Y) dY in the absence state, that is, larger than the marginal distribution ∫F _n (Y) dY. Therefore individual marginal distribution ∫
f _n (Y) dY is compared with a threshold value THL2 which is arbitrarily determined, and a check 36 of the attribute determination area 263 is performed according to the comparison result.
3 can also be detected. That is, the attribute discrimination area 2 in which the individual marginal distribution ∫f _n (Y) dY is _{equal to} or larger than the threshold value THL2.
63 to the attribute determination area 263 with a check 363
And the individual marginal distribution ∫f _n (Y) dY is _equal to the threshold value THL2.
The attribute discrimination area 263 that is less than may be detected as the attribute discrimination area 263 to which the check 363 is not attached.

In this case, the threshold value THL2 (THL
2 _n ) is, for example, THL2 _n = ∫F _n (Y) dY (∫F _n (Y) dY is the individual marginal distribution ∫f _n (Y) dY obtained without the check 363) Or THL2 _n = ∫F _n (Y) d
Y + α (α is a constant).

As shown in FIG. 5, even when the check 363 is out of the corresponding attribute discrimination area 261,
By setting α to a positive integer of any suitable size, the presence or absence of the check 363 can be accurately detected.

In this case, one attribute discrimination area 3
When not only the case where the check 363 is given to the 63 but also the case where the check 363 is given to a plurality of attribute determination areas 363, it is possible to detect which attribute determination area 263 is given the check 363. For example, consider a case in which each number 1, 2, and 3 is used as the attribute determination character 323 in the form 22 instead of the ward, town, and village. In this case,
If the total number of characters entered in the entry area 243 is equal to one selected from the attribute determination characters 323,
One attribute determination area 32 corresponding to the selected character 323
3 is checked 363, and when the total number of characters entered in the entry area 343 is equal to the sum of a plurality of characters selected from the attribute discrimination characters 323, each of the plurality of selected characters 323 is corresponded. Attribute determination areas 3
It is assumed that a check 363 is added to 23. For example, when the total number of cutout target characters 343 entered in the entry area 243 is 5, the person who fills the form 22 has the Arabic numeral 3
A check 363 is attached to each of the two attribute discrimination areas 263 corresponding to and. Even in such a case, the attribute discrimination area 323 where the individual marginal distribution ∫f _n (Y) dY is _{equal to or} larger than the threshold value THL2 is replaced with the attribute discrimination area 32 with the check 363 attached.
3 can be detected. In this case, the presence or absence of the check 363 can be detected by using a positive integer as the constant.

As described above, a marginal distribution is individually created for each attribute discrimination area, and an attribute discrimination area that exceeds a predetermined threshold is detected for each entry area, and a marginal distribution that exceeds a predetermined threshold is obtained. The attribute discrimination area may be detected as an attribute discrimination area in which a check has been made on the entry area corresponding to the discrimination area.

[0106]

As is clear from the above description, according to the character extracting apparatus of the present invention, the total number of characters to be entered in the entry area and the presence / absence of the attribute discrimination area corresponding to the entry area are checked. holds based on the correlation that exists, as previously examined data the lower limit G _min and an upper limit G _max of cut allowable number of the entry region between the. The detected total number of the provisional cutout start position or the provisional cutout end position in the entry area is defined as the total number M of cutout characters in the entry area, and the total number M of cutout characters is the lower limit Gmin of the allowable number of _cutouts.
Less than or cut when the allowable number of larger than the upper limit G _max is cut start position and cut the end position of the provisional provisional obtain the clipped characters total number as cut position of the target pattern extraction position i.e. cut target character pattern It is determined that it is inappropriate, and the temporary cutout position is detected again to correct the temporary cutout start position and the temporary cutout end position.
When the total number M of cut-out characters in the entry area is equal to or more than the lower limit value G _min and equal to or less than the upper limit value G _{max of the} allowable number of cut-out characters, the provisional cut-out start position and the provisional cut-out where the total number of cut-out characters are obtained. The end position is determined to be appropriate as the target pattern cutout position, and the temporary cutout start position and the temporary cutout end position are determined as the target pattern cutout positions.

[0107] Since the decision in accordance with the comparison result of the upper limit value G _max and the lower limit value G _min of the thus cut object pattern extraction position characters total number M and cut allowable number, easy determination of the target pattern extraction position I can do it. Therefore, there is an advantage that the processing for extracting the character pattern to be extracted can be speeded up, and in addition to this, the hardware configuration of the apparatus can simplify the apparatus configuration and reduce the size of the apparatus.

[Brief description of the drawings]

FIG. 1 is a functional block diagram schematically showing a configuration of an embodiment.

FIG. 2 is a diagram illustrating an example of a form that can be used in the character cutout device of the embodiment.

FIG. 3 is a diagram illustrating an example of an image pattern of a form used in the embodiment.

FIGS. 4A and 4B are diagrams showing an image pattern of an attribute determination area and a cumulative number of character pixels relating to the image pattern when no check is made.

FIGS. 5A and 5B are diagrams showing an image pattern of an attribute determination area and a cumulative number of character pixels relating to the image pattern when there is a check;

FIGS. 6A and 6B are diagrams showing an image pattern and a cumulative number of character pixels in an entry area.

FIG. 7 is a diagram showing a flow of operation focusing on the position setting control means of the embodiment.

FIG. 8 is a diagram showing another example of a form that can be used in the character cutout device of the embodiment.

[Explanation of symbols]

 10: Character extraction device 12: Image storage unit 14: Format storage unit 16: Allowable extraction number setting unit 18: Extraction position determination unit 20: Pattern reading unit

Continuation of the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06K 9/00-9/82

Claims (3)

(57) [Claims] 1. A character extracting apparatus for extracting a character pattern to be extracted from an image pattern of a form having an entry area in which a character to be extracted is entered and an attribute determination area corresponding to the entry area. To create a marginal distribution, detect presence / absence of the attribute discrimination area using the marginal distribution of the attribute discrimination area, and set an upper limit value and a lower limit value of the permissible number of pieces corresponding to the detection result. An allowable number setting unit, detecting a character element pattern from the image pattern of the writing area, setting a temporary cutout position using the detected position of the character element pattern, and using the temporary cutout position to temporarily set the temporary setting of the writing area. When the total number of cut-out characters is determined, the provisional cut-out character total number is larger than the upper limit of the cut-out allowable number and smaller than the lower limit of the cut-out allowable number. Correcting the provisional cutout position and recalculating the provisional cutout character total number using the corrected position, and the provisional cutout character total number is equal to or less than the upper limit of the cutout allowable number and the cutout allowable number. And a cutout position determining means for determining, as a target pattern cutout position, a tentative cutout position at which the tentative cutout character total number is obtained, and a cutout target character pattern using the target pattern cutout position. A character readout unit for extracting a character. 2. The character extracting apparatus according to claim 1, wherein the marginal distribution created individually for each attribute discrimination area is normalized, and the marginal distribution which is the largest among the normalized marginal distributions for each entry area. A character segmentation device that detects the attribute discrimination area that has obtained the maximum peripheral distribution as an attribute discrimination area that has been checked for an entry area corresponding to the discrimination area. 3. A character segmentation device according to claim 1, wherein a marginal distribution is created individually for each attribute discrimination area, and an attribute discrimination area exceeding a predetermined threshold is detected for each entry area.
The attribute discrimination area that has obtained the marginal distribution that is equal to or more than the predetermined threshold value,
A character cutout device, which detects an entry area corresponding to the determination area as an attribute determination area with a check.