JP2662238B2 - Creating dictionary for character recognition - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] The present invention relates to a character recognition dictionary creation method for reading a character printed on paper and creating a character recognition dictionary having characteristics corresponding to the character. The purpose of this paper is to provide a character recognition dictionary creation method that can create a character recognition dictionary that can perform character recognition according to the characteristics of the characters in which the target font is printed. Reads and displays one character on the display unit, reads the same character as the displayed character from the standard format table, displays it on the display unit in comparison with the target character, and simultaneously reads the parameters for creating the standard font. Display on the display unit, compare standard characters, change standard font parameters and set standard font patterns to target characters. The image is converted to a standard font based on the approximation target image at the time when both character patterns are considered to be finally matched with the parameters obtained by the approximation operation. Is configured to be stored.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition dictionary creating method for reading characters (printed characters) printed on paper and creating a character recognition dictionary having features corresponding to the characters.

2. Description of the Related Art In recent years, a character input device that reads and recognizes characters printed on paper, for example, in printed characters, has been developed. However, there are various types of printed characters, such as Gothic and Mincho. Therefore, realization of a character input device capable of recognizing characters in all forms is desired.

[Related Art] When recognizing printed characters printed on paper, a dictionary for character recognition is stored in a character input device in advance, and which character in the dictionary matches the read character is determined. Is used to determine that the matched character is the input character. As the character recognition dictionary, a single template type and a multi-template type are known. In the former, a characteristic amount of a character in a specific form (for example, Mincho) is stored in association with a JIS code. On the other hand, the latter stores a feature amount of a character in a plurality of forms (for example, Mincho, Gothic, etc.) in association with a JIS code.

Since the single plate type is a character recognition dictionary dedicated to a specific character form, the feature amount is stored in detail for the character in that form. On the other hand, the multi-plate type does not specify the character type to be read, and therefore can handle a plurality of character types, but has a small feature amount for each character type.

[Problems to be Solved by the Invention] Therefore, a single-plate type dictionary can obtain a higher recognition rate than the multi-plate type for the same font as the dictionary and can perform recognition processing at high speed.
There is a problem that only one type font can be supported. This problem will be solved if it can be switched each time for the character font to be recognized. However, in order to create a dictionary for one font, a large amount of characters of that font are read by a scanner, and a representative feature of one character type is created using an averaging method or the like. (For example, a code table and a feature amount corresponding to the code table).

Such complicated and time-consuming processing had to be performed.

The present invention has been made in view of such a point, and can reduce the amount of processing by approximating the above processing, that is, character recognition can be performed according to the characteristics of characters printed on paper with a small amount of processing. It aims to provide a dictionary creation method for character recognition.

[Means for Solving the Problems] FIG. 1 is a flowchart showing the principle of the present invention.
According to the present invention, a paper on which a target font is printed is read by a scanner, one of the characters is displayed on a display unit (step), and the same character as the displayed character is read from a standard format table and read on a display unit. In addition to displaying the target characters, the parameters for creating the standard font are also read out and displayed on the display unit (step), and the standard font parameters are changed to change the standard font pattern to the target character image while comparing the compared characters. Is repeated (step), and based on the approximate target image at the time when both character patterns are considered to match finally, the standard font is image-converted with the parameters obtained by the approximate operation and stored as a dictionary The feature is that it is stored in the device (step) To have.

[Operation] The target pattern of the character read on the display unit is displayed in comparison with the standard pattern stored in the apparatus, and the operation of changing the standard pattern parameters to bring the standard pattern close to the target pattern is repeated.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a diagram showing an operation image of a system for implementing the present invention. The operation summary is as follows.

(1) Activate the scanner 2 from the personal computer 1 and display the read image on the CRT of the personal computer 1.

(2) The operator selects one required character (target font) from the displayed image and displays it on the left side of the screen.

(3) The software of the personal computer 1 asks for the font type (for example, Gothic font or Mincho font), and selects an appropriate font type.

(4) A standard format (for example, Gothic type, Mincho type, etc.) of the software of the personal computer 1 is displayed on the right side of the screen. At the same time, selection parameters of the standard format (parameters for creating the standard format, hereinafter simply referred to as parameters) are also displayed.

(5) Change the parameters so that the image on the right side of the screen (standard format side) approaches the image on the left side (target font side).

(6) The approximation operation ends when it is considered that they match. The software of the personal computer 1 sends the approximated target image to the feature generator 3.

(7) The feature quantity generator 3 performs image conversion of the standard font of Gothic font or Mincho font using the parameters obtained by the approximation operation. The software of the personal computer 1 uses the dictionary as a storage device (for example, a floppy disk)
To memorize.

FIG. 3 is a functional block diagram for implementing the present invention. The same components as those in FIG. 2 are denoted by the same reference numerals.
In FIG. 2, reference numeral 2 denotes a scanner, reference numeral 3 denotes a feature generator, and reference numeral 10 denotes font generation software for executing a sequence according to the present invention, which is specifically included in the personal computer 1 shown in FIG. 4 is an operator, 5 is a standard font image parameter table storing parameters of a standard font image, 6 is a floppy disk (FD) storing a character recognition dictionary obtained by the present invention, and 7 is a CRT for displaying font images and the like. It is. The method of the present invention will be described with reference to the block configured as described above.

This will be described with reference to the flowchart shown in FIG. First, the operator 4 places the paper on which the target font is printed on the scanner 2, activates the scanner 2 to perform scanning (step 1), selects one of the characters, and selects the one character from the left side of the screen of the CRT 7. (Step 2).
Next, the operator 4 inputs the same character code as the character (target image) displayed on the screen (step 3). When the font generation software 10 recognizes this code, it extracts the font of the character and its parameters from the standard font image parameter table 5 and executes the CR.
Display on T7 (step 4). At this time, the standard font is displayed on the right side of the screen so that it can be easily compared with the target image (the positional relationship may be reversed).

The operator 4 compares the target image displayed on the screen with the standard font pattern and checks whether or not they are similar (step 5). If not,
The operator changes the parameters so that the standard font pattern on the right side of the screen approaches the target image on the left side (step 6). This change in the parameter is immediately sent to the font generation software 10, which changes the parameter based on the parameter change. Steps 5 and 6 are repeated until they are considered coincident.

When it is deemed that the two images match, the font generation software 10 determines a conversion rule from the parameter, the standard font pattern to the target image (step 7), and applies the conversion rule to the standard format pattern. An image close to the target image is created (step 8). Next, the font generation software 10 sends the approximate target image to the feature amount generator 3 for one character, and the feature amount generator 3 converts the standard font into a parameter obtained by the approximation operation based on the received approximate target image. To convert the image. The image-converted feature amount is sent to the font generation software 10, which stores the character code and the feature amount on the floppy disk 6 to be a dictionary adapted to the target font (step 9). ). When the above operation is performed for all the standard format patterns and stored in the floppy disk 6, the floppy disk 6 can be used as a dictionary for character recognition.

FIG. 5 is a diagram showing a configuration example of a standard font table for a Chinese character "A". In the standard font image parameter table 5 in FIG. 3, such a table is stored for a plurality of characters. As shown in the figure, the standard font table includes image data 5a and a parameter table 5b, and FIG. 6 shows a detailed configuration example in the parameter table 5b. As shown in the figure, parameter table 5b contains kanji codes, Mincho / Gothic flags, character heights,
The data of the character width, the number of character line dots (horizontal line), the number of character line dots (vertical line), and the position of the center of gravity are stored.

Next, a method of generating a conversion rule for standard font image data while comparing with the target image described in step 7 of FIG. 4, that is, a method of determining a conversion rule parameter will be described. The parameters required for the conversion rule are as follows.

Magnification character aspect _{(x; α x, y;} α y) displacement from the center of gravity of the standard parameter tables _{(x; β x, y;} β y) magnification character line (in the case of the Gothic aspect, Akira In the case of the morning body, only the horizontal direction (γ) The coefficient of the affine transformation (δ) The operator changes the parameters above to bring the standard image closer to the target image (see step 7 in FIG. 4).

FIG. 7 is a diagram showing a specific sequence for bringing the standard font image closer to the target image. It is assumed that the target image 20 and the standard font image 21 are displayed on the screen in comparison. Here, in order to bring the standard font image 21 closer to the target image, first, the aspect ratio is adjusted as shown in FIG. Here, in order to lengthen vertically, the magnification ratio α _x = 0.57 and α _y = 0.
Set to 87. As a result, the height of the standard font image 21 is 0.82 times and the width of the standard font image 21 is 0.57 times.
It becomes a vertical character close to 20.

The processing for thickening the character line is performed as shown below. Therefore, the enlargement ratio of the character line is set to γ = 1.33, and the magnification is γ times only in the horizontal direction of the character line. As a result, the character image becomes closer to the target image 20. Further, a process of lowering the position of the center of gravity is performed as shown in FIG. . Here, β _x = 0, β
Set _y = 15 dots. As a result, the center of gravity of the character is lowered by 15 dots in the vertical direction, and does not change in the horizontal direction. Assuming that affine transformation is not performed (that is, δ = 1), the character pattern shown below becomes an approximate target image. The parameter values α _x , α _y , β _x , β _y obtained by the above approximation operation,
The values of γ and δ are used as rules for converting the standard pattern image into the target image.

FIG. 8 is a flowchart showing the operation of the feature generator 3 (see FIG. 3). First, a feature amount (approximate target image) sent from the font generation software 10 (see FIG. 3) is input (step 1), smoothing processing is performed (step 2), and character outlines are extracted (step 3). ). After the contour is extracted, the direction line element of the contour line is determined (step 4), the density of the direction line element in the 8 × 8 mesh is determined (step 5), and output to the font generation software 10 as a feature amount. I do. The font generation software 10 receives this and stores it as a dictionary on the floppy disk 6 (FIG. 3) together with the code.

In the above description, the Gothic type and the Mincho type are used as character types, but it goes without saying that the present invention is not limited to this and can be applied to other character types.
Further, the storage device for storing the created dictionary is not limited to a floppy disk, and other types of storage devices can be used.

[Effects of the Invention] As described above in detail, according to the present invention, the read target image is compared with the standard font pattern, and the parameter is repeatedly changed so that the standard font pattern approaches the target image. This eliminates the need for a process of reading a large number of characters and taking the average as in the conventional example. Therefore, it is possible to provide a character recognition dictionary creating method capable of performing character recognition according to the characteristics of characters printed on paper with a small amount of processing. Moreover, if the present invention is applied to each character type, a dictionary for recognizing all character types can be created, and the practical effect is extremely large.

[Brief description of the drawings]

1 is a flowchart showing the principle of the method of the present invention, FIG. 2 is a diagram showing an operation image of a system for carrying out the present invention, FIG. 3 is a functional block diagram for carrying out the present invention, FIG. Is a flowchart showing the detailed operation of the present invention, FIG. 5 is a diagram showing a configuration example of a standard font table, FIG. 6 is a diagram showing a configuration example of a parameter table, and FIG. FIG. 8 is a flowchart showing a specific sequence, and FIG. 8 is a flowchart showing the operation of the feature amount generator. 2 and 3, 1 is a personal computer, 2 is a scanner, 3 is a feature generator, 4 is an operator, 5 is a standard font image parameter table, 6 is a floppy disk, 7 is a CRT, and 10 is font generation software. It is.

Claims (1)

(57) [Claims] 1. A paper on which a target font is printed is read by a scanner, one of the characters is displayed on a display unit (step), and the same character as the displayed character is read from a standard format table and displayed on a display unit. In addition to displaying the target character in comparison with the target character, the parameters for creating the standard font are simultaneously read out and displayed on the display unit (step). The operation of approaching the image is repeated (step). Based on the approximate target image at the time when both character patterns are considered to match in the end, the reference font is image-converted with the parameters obtained by the approximate operation, and it is converted into a dictionary. It is characterized in that it is stored in a storage device (step). Character recognition dictionary creating method to.