JPH0365584B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character separation device, and more particularly to a device for identifying whether a character string image on a printed paper including touching characters is a variable pitch character string or a constant pitch character string.

BACKGROUND ART Conventionally, devices for optically reading printed or handwritten characters (hereinafter referred to as OCR) for alphanumeric characters and katakana have already been put into practical use. However, when OCR scans items such as mail and documents with various printed characters and qualities, and the limitations of paper and printing machines are relaxed, it is difficult to read the prints that occur when printing with a typewriter, etc. Increasingly, contact between adjacent characters occurs due to blurring, misalignment, etc. In order to separate character string images on paper that include such touching characters, it is first necessary to identify character strings that are evenly spaced to be read. To distinguish between variable character pitch strings and regular pitch strings, if you use a statistical method to check based on the frequency distribution of all character widths, including characters that touch, as the number of characters that touch increases, equal pitch characters Identification of column or variable pitch string becomes inaccurate. If the identification of a constant pitch character string or a variable pitch character string becomes inaccurate, it may become difficult to detect a block of characters that touch each other in a constant pitch character string, or it may become difficult to detect a block of characters that touch a character string with a relatively small number of characters. This results in a precise separation.

An object of the present invention is to set the range of possible widths of character blocks that can become one character from the average height of a plurality of character blocks, and to obtain the most frequent character block within the range of existence determined by a predetermined allowable width. By updating the upper limit value of the width based on frequency information and obtaining a single character width class, the width and commas of contacting character blocks of evenly pitched character strings can be removed, making it easier to identify character pitches. An object of the present invention is to provide a character pitch identification device that allows for accurate character pitch identification.

Another object of the present invention is to provide a character pitch identification device that identifies character pitches based on the degree of width of a character block in one character width class.

Still another object of the present invention is to provide a character pitch identification device which can stably identify character pitch from various printed characters whose character pitch values cannot be known in advance.

According to the present invention, in a character separation device that scans a character string image written on a paper surface and separates it into individual characters, the height and width of a plurality of character blocks are calculated from a binarized character string image. a frequency table memory for obtaining frequency information regarding the width of a plurality of character blocks; a means for obtaining an average height of a character block from the heights of a plurality of character blocks; and a frequency table memory for obtaining frequency information regarding the width of a plurality of character blocks; Set the existence range of the width of a possible character block that can be one character, and set the upper and lower limit values of the width of the most frequent character block within the existing range from the frequency information in the frequency table memory with a predetermined allowable width. means to update the upper limit value based on the frequency information in the frequency table memory and obtain the upper limit value and lower limit value of a single character class; There is obtained a character pitch identification device characterized in that it has means for identifying whether it is a pitch character string or an equal character pitch character string.

Hereinafter, the present invention will be described with reference to a specific embodiment. FIG. 1 is an example showing a part of an evenly spaced character string image including contact between characters.
In the figure, the white background indicated by diagonal lines shows a separable character part image, that is, a character block, as a rectangular area.
In the figure, V _i and H _i (i=1, . . . , 8) indicate the width and height of each character block. Here, the character block width V ₁ ,
The character block with V ₈ becomes a character image containing two characters due to the contact of the character blocks.

FIG. 2 shows an example of the frequency distribution of the widths of a plurality of character blocks on a paper surface as shown in FIG. 1, and the principle of the present invention will be explained using this. In Figure 2, the horizontal axis of the frequency distribution regarding multiple character blocks indicates the value of the character block width V, and the vertical axis
NUM indicates the number of character blocks at a given character block width value, that is, the frequency value. In the present invention, while referring to the frequency distribution _described above, _{the character width section A 1 that} can be considered as one character shown in _FIG . Identification of constant pitch/variable pitch character strings is automatically performed by comparing the average height H _n of character blocks with a threshold value set based on the average height H n of character blocks. Therefore, extraction of the character width section _A1 that can be considered as one character will be explained. First, since there is a close relationship between the width and height of a character, the average height of a character block H _n and the constants C ₁ and C ₂ (however, C ₁ < C ₂ , e.g. C ₁ = 0.5, C ₂ =1.2) to set the one-character possible range A. Next, referring to the frequency distribution, a character block having the most frequent value within the range A is extracted as the character width that is most likely to be one character. Here, when obtaining the above-mentioned mode, an allowable width ΔT (ΔT = 3 in the figure) is usually set, in order to take into account cases where the number of observable character blocks is small, such as in addresses on mail, etc.
Upper limit value V′ _{U and lower limit value of the character block width interval A S where} the frequency of the number of character blocks included in the allowable width ΔT is maximum
Calculate V′ _L. In addition, here, one character possible range A
In order to calculate the mode within the allowable width ΔT,
V′ _U −V′ _L =ΔT holds true.

Next, the section of character width that is most likely to be one character mentioned above.
By setting A _S as the initial value and updating its upper limit value V′ _U in the direction of increasing it while referring to the frequency distribution, we can create an interval A ₁ of character width that can be considered as one character.
The lower limit value V _U and lower limit value V _L can be extracted.
Regarding the lower limit value V _L , for example, period,
Remove characters with small width such as commas,
In order to improve the reliability of discrimination between a constant pitch character string and a variable pitch character string, the lower limit value _V'L of the character width section A _S that is most likely to be one character is not updated in a direction in which its value decreases. That is, the lower limit value V _L is equal to the lower limit value V′ _L. Also, update the above lower limit value V′ _U and 1
The update operation for determining the upper limit value V _U of the character width section A ₁ that can be considered as a character is completed based on the constant C ₃ and the average height of the character block, as explained later using Figure 3.
This can be carried out using H _n (where C ₃ <C ₂ , for example C ₃ =0.4). Here, it will be briefly described that a constant pitch character string or a variable pitch character string can be identified based on the value of the character width section _A1 . In other words, excluding characters with small character widths such as periods and commas, in the case of evenly spaced character strings, the width of the character block most likely to be one character as described above (i.e., A _S
On the other hand, in variable pitch character strings, even if characters with small widths such as periods and commas are excluded, the character strings have a tendency to vary without being concentrated around the width of one character block. ing. Therefore, as shown in FIG. 2, by detecting the section _A1 that can be regarded as one character wide, it is possible to distinguish between a constant pitch character string and a variable pitch character string.

FIG. 3 is a logic block diagram showing a specific embodiment of the present invention. In the figure, the signal is represented by adding S to the end of the signal line. A scanning device 1 optically scans a printed or handwritten character string on a paper surface, converts it into an electrical signal, and sequentially writes a binary quantized character string image into a character string image memory 2. 3 is a character block extraction device that sequentially detects character images surrounded by white bits (hereinafter referred to as character blocks) from the character string images stored in the character string image memory 2, and determines the starting position and size of each character block. The value is stored in the character block register 4. Note that the size of a character block represents the width and height of the character block. Further, such a character block extraction device can be obtained using, for example, the technique disclosed in Japanese Patent Application No. 1983-27512 filed by the same applicant. Character block register 4
The widths of the plurality of character chunks stored in are sequentially transferred to the control device 6. The control device 6 converts the width values of the character blocks that are sequentially transferred into addresses in the frequency table memory 5, reads out the contents of the corresponding frequency table memory 5, increments them by +1 addition circuit 11, and converts them into addresses in the frequency table memory 5. Write to the same storage location in memory 5. In this way, the frequency value of the width Vi of the character block extracted from the character string image memory 2 is stored in the frequency table memory 5 at the address Vi of the frequency table memory 5.

It is assumed that the frequency table memory 5 is initially cleared to 0. Next, the heights of the plurality of character blocks stored in the character block register 4 are sequentially transferred to the average height detection section 7. The average height detection unit 7 is
Calculate the average height of multiple character blocks. still,
In the average height detection section 7, a circuit that calculates the height having the most frequent value may be used.

15 is a constant register, which stores constants C ₁ , C ₂ , C ₃ (C ₁ <C ₂ , C ₃ <C ₂ ) explained in FIG. The constants C ₁ and C ₂ are used to determine the possible range of the width of one character block from the average height Hm of the character block detected by the average height detection unit 7, and are constants.
_C3 is for generating a threshold value to be used in the one character width class detection unit 8 surrounded by a dotted line in the figure. The multiplication unit 10 multiplies the average height Hm of the character blocks by constants C ₁ , C ₂ , and C ₃ . Multiplication section 10
Of the values obtained by C ₁・Hm and C ₂・
Hm (C ₁ ·Hm<C ₂ ·Hm) is stored in the one character range setting register 12 as the lower limit and upper limit of the possible width of one character, respectively, and C ₃ ·Hm is stored in the threshold value register 17. 16 is a constant register, in which a constant ΔT indicating a preset allowable width is set. 13 is a most frequent width detection section. The lower limit value C ₁ · Hm and the upper limit value C ₂ · of the possible width of one character stored in the one character range setting register 12
When the frequency value of the width of the character block included in Hm is inputted sequentially from the frequency table memory 5 to the most frequent width detecting section 13 by the control device 6, the most frequently occurring width detecting section 13 inputs the frequency value of the width of the character block contained in the constant register 16. Constant △ which is the content
Count the frequency values at intervals of T, and calculate the lower and upper limits V _L , V _U (however, C ₁・Hm≦V _L ≦C ₂・Hm−△ T, C ₁ : Hm+
△T≦V _U ≦C ₂・Hm, V _U −V _L = △T is satisfied. )
Detect. The lower limit value V _L and upper limit value V _U of the width of the character block having the most frequent value are stored in the single character initial class register 81 .

The one-character width class detection unit 8 updates the upper limit value V U of the width of the character block having the above-mentioned most frequent value based on the frequency information in the frequency table memory 5, and updates the upper limit value V _U of the width of the character block having the above-mentioned most frequent value, and calculates the interval of the width of one character, that is, as shown in FIG. This is a circuit for finding the upper and lower limits of the interval _A1 that can be regarded as one character width, as shown in FIG. 85 comparison part 8
6. It is composed of an OR circuit 87. The upper limit value of the width of a character block stored in the one-character initial class register 81 is set in the counter 82, and then incremented by 1, and the frequency value corresponding to the width of the character block, which is the content of the counter 82, is transferred to the control device 6. Accordingly, the data is read from the frequency table memory 5 and stored in the frequency value register 83. The zero detection unit 84 detects whether the content of the frequency value register 83 is 0 or not. If it is 0, the output signal 841S is set to "1" to open the OR circuit 87 and the counter 82 becomes 1. It will be counted up.

On the other hand, in the zero detection unit 84, if the content of the frequency value register is not 0, the output signal 841S
“0” is output from the controller 6, and the subtracter 8 subtracts the upper limit value of the width of a character block, which is the content of the one-character initial class register 81, and the content of the counter 82.
Transfer to 5. The subtraction circuit 85 includes a counter 82
The upper limit value of the width of a character block, which is the content of the initial character class register 81, is subtracted from the content of the initial character class register 81, and the subtracted value is transferred to the comparison unit 86. The comparison unit 86 compares the contents of the threshold register 17 and the output value from the subtraction unit 85, and if the contents of the threshold register 17 is larger than the output value of the subtraction unit 85, outputs “1” from the output signal 861S. . When “1” is output from the output signal 861S, the control device 6 uses the contents of the counter 82 to update the upper limit value of the width of a character block stored in the single character initial class register 82, and also performs OR. The circuit 87 is opened and the counter 82 is counted up by one. On the other hand, if the contents of the threshold register 17 are smaller than or equal to the output value of the subtraction section 85, the output signal 861 of the comparison section 86
"0" is output from S, and the control device 6 transfers the lower limit value of the character block width, which is the content of the one-character initial class register 81, and the upper limit value updated as described above to the one-character width class register 14. be done. Here, the single character width class detection unit 8 described above
Summarize the behavior of That is, the operation of updating the upper limit value of the character block width set as an initial value in the 1-character initial class register 81 by the contents of the counter 82 indicating the character block width under inspection is based on the contents of the counter 82 and the 1-character initial class. The process ends when the difference from the upper limit value stored in the register 81 becomes greater than the value C ₃ Hm that is the content of the threshold register 17, and the upper limit value stored in the one-character initial class register 81 is transferred to the one-character width register 14. will be done.

Regarding the conditions for updating the upper limit value of the character block width, when the frequency value of the frequency distribution corresponding to the character block width under inspection, which is the content of the counter 82, is 0 (detected by the zero detection unit 84), Unconditionally counts up the contents of the counter 82 (i.e., the width of the next character block)
However, when the frequency value is not 0 and the operation of updating the upper limit value of the character block width described above is not yet completed, the contents of the counter 82 indicating the character block width under inspection are set as the upper limit value of the single character initial class register 81. By transferring, the upper limit value of the character block width will be updated.

A determining unit 9 surrounded by a dotted line in the figure determines whether the character string image stored in the character string image memory 2 is a constant pitch character string or a variable pitch character string. 91
is a subtraction unit which subtracts the lower limit value from the upper limit value of the width of a character block, which is the content of the one-character width class register 14, and transfers the output value to the comparison unit 93. Reference numeral 92 is a threshold value register in which a preset threshold value is stored. Note that the contents of the threshold value register 92 may be set based on the average height of the character blocks obtained by the average height detection section 7, like the contents of the threshold value register 17. A comparison section 93 compares the contents of the threshold register 92 and the output value of the subtraction section 91 . If the content of the threshold register 92 is larger than the output value of the subtraction unit 91, "1" is output from the output signal 931S, thereby making it possible to detect an equally pitched character string. If the contents of the threshold register 92 are smaller than or equal to the output value of the subtraction unit 91, the variable pitch character string can be detected by outputting "0" from the output signal 931S.

As described above, by applying the present invention, it is possible to stably identify whether a character string is a variable pitch character string or a constant pitch character string even when adjacent characters are likely to come into contact with each other.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a part of an evenly pitched character string image as an example. FIG. 2 is a diagram showing an example of the frequency distribution of character block widths shown to illustrate the principle of the present invention. FIG. 3 is a logic block diagram showing one specific implementation of the present invention. In the figure, 1 is a scanning device, 2 is a character string image memory,
3 is a character block extraction device, 4 is a character block register, 5 is a frequency table memory, 6 is a control device, 7 is an average height detection unit, 10 is a multiplication unit, 11 is a +1 addition circuit,
12 is a single character range setting register, 13 is the most frequent width detection unit, 14 is a single character width class register, 15,
16 is a constant register, 17 is a threshold register, 8 is a one-character width class detection section, 81 is a one-character initial class register, 82 is a counter, 83 is a frequency value register, 84 is a zero detection section, 85 is a subtraction section, and 86 is a comparison section. , 87 is an OR circuit, 9 is a determination section, 92 is a threshold register, 91 is a subtraction section, and 93 is a comparison section.

Claims (1)

[Claims] 1 A character separation device that scans a character string image written on a paper and separates it into individual characters extracts multiple character blocks surrounded by white bit strings from the binary quantized character string image. , means for obtaining the height and width of the character blocks; a frequency table memory for obtaining frequency information regarding the widths of the plurality of character blocks; and a means for obtaining the average height of the character blocks from the heights of the plurality of character blocks. A range of widths of possible character blocks that can become one character is set using the average height of the character blocks, and the frequency information in the frequency table memory is used to obtain the highest frequency within a predetermined allowable width. means for obtaining upper and lower limit values of the width of a character block within the existing range; and means for updating the upper limit value based on frequency information in the frequency table memory to obtain the upper and lower limit values for a single character width class. and means for checking the spread of the upper limit value and lower limit value of the one character width class and identifying whether it is a variable character pitch character string or an equal character pitch character string.