JPH11250183A - Method for discriminating style of printed character, character image processing system and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character image processing system capable of improving a character recognition ratio by accurately and quickly discriminating the style of a character image to be discriminated. SOLUTION: In the system 1, an image conversion part 12 receives a character image and generates a reference style image similar to a Gothic style having almost fixed stroke width. A style discrimination part 14 compares the pixel distribution of the reference style image with that of the inputted character image, and when the mean value of overlapped parts between both the images is >=0.8 and dispersion is <=100, judges the input character image as the Gothic style, but judges it as a Ming style in other cases. A character recognition part 15 selects a dictionary 151 or 152 corresponding to the discrimianated style so as to execute character recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character image processing system, and more particularly, to a method for quickly and accurately determining the typeface of a printed character before processing a character image.

[0002]

2. Description of the Related Art As an example of a character image processing system, an electronic filing system or a system similar thereto has been put to practical use. The electronic filing system captures a document on which characters are printed (hereinafter, a print document) by an image scanner and holds the image data as it is, or holds it as a file after performing a character recognition process. Specifically, the character recognition process compares the characteristics of a character image cut out from image data with each category (character code) of a character recognition dictionary prepared in advance, and determines a character code having the best match in the dictionary. This is the process of extracting from. In performing the character recognition process, if a single print document contains characters of a plurality of types of fonts such as Gothic font and Mincho font, erroneous recognition is likely to occur. Therefore, recently, attempts have been made to perform character recognition processing in consideration of the difference in typeface.

[0003]

Conventionally, when performing character recognition processing in consideration of the difference in typefaces, a first method in which typeface discrimination is performed using a dictionary for typeface discrimination after image feature collation, A plurality of dictionaries corresponding to typefaces are prepared, and one of the second methods for simultaneously performing matching of image features and discrimination of typeface has been adopted.

However, when collation is performed in advance as in the first method, the character recognition rate cannot be improved, and a general dictionary is used so that any typeface can be used. Therefore, there is a problem that the typeface discrimination rate cannot be improved. On the other hand, in the case of the second method, there is a problem that the processing speed is reduced because the collation of the image feature and the typeface determination are performed simultaneously for each character.

SUMMARY OF THE INVENTION An object of the present invention is to provide a method for determining the typeface of printed characters, which can accurately and quickly determine the typeface such as Mincho and Gothic. Another object of the present invention is to provide a character image processing system suitable for implementing the above typeface discrimination method, and a recording medium for implementing the above typeface discrimination method on a computer device.

[0006]

According to the present invention, there is provided a method for determining the typeface of a printed character, comprising the steps of: generating a reference typeface image having a specific shape from a character image whose typeface is unknown; Comparing the pixel distribution with the reference font image thus determined to determine the font of the character image. The reference typeface image is, for example, a character image in which all stroke widths are substantially constant. A specific process different from that of other character images may be performed on a specific one of the character images for which the typeface has been determined.

A character image processing system according to the present invention for solving the above-mentioned other problems includes a plurality of character recognition dictionaries prepared for each typeface and an image conversion for generating a reference typeface image having a specific shape from a character image whose typeface is unknown. Means, a font discriminating means for discriminating the font of the character image by comparing the pixel distribution between the character image and the generated reference font image, and a character recognition dictionary corresponding to the discriminated font. Means for performing character recognition of the character image.

[0008] Another character image processing system of the present invention includes the image conversion means, the typeface discriminating means, and means for performing emphasis processing on a specific one of the character images whose typefaces have been discriminated. It becomes.

[0009] Another character image processing system of the present invention includes the image conversion means, the typeface discriminating means, and a means for extracting and rearranging a specific one of the character images for which the typeface has been discriminated. Become.

In each of the character image processing systems, the image conversion means performs, for example, a skeleton process on the character image to generate a character image having a substantially constant stroke width, and converts the character image to the reference font image. It is constituted so that.

A recording medium according to the present invention for solving the above-mentioned other problems is a computer-readable recording medium in which a program for causing a computer device to execute at least the following processing is recorded. (1) a process of performing a skeleton process on a character image whose font is unknown to generate a reference font image in which all stroke widths are substantially constant; and (2) a pixel distribution between the character image and the generated reference font image. And determining the typeface of the character image by comparing

[0012]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a configuration diagram of a character image processing system according to a first embodiment of the present invention. In this embodiment, an example is shown in which a character image in which a Mincho font and a Gothic font may be mixed is captured by an image scanner, and character recognition is performed based on this character image.

The character image processing system 1 is realized by, for example, a computer device to which an image scanner 11 is connected, and the computer device reads a predetermined program and executes its own operating system (OS).
Image conversion unit 1 formed by executing in cooperation with
2. Reference font image storage unit 13, font discriminating unit 14, character recognition unit 15, dictionary 151 storing categories for each font type
, 52,... Are provided with respective functional blocks of the output control unit 16.

The image conversion unit 12 includes an image scanner 11
The reference font image storage unit 13 functions as a working memory in the process of generating the reference font image and generates the reference font image after the generation. It is stored for processing. The typeface determination unit 14 compares the pixel distributions of the character image and the reference typeface image on a character-by-character basis to determine the typeface for each character image. The character recognition unit 15 performs character recognition using a dictionary of a corresponding typeface. Output control unit 16
Outputs the processing result of the character recognition unit 15 to a display device or a printing device (not shown).

The above-mentioned program is usually recorded in an internal recording device or an external recording device of the computer device, and is read and used as needed in the course of processing described later. It is sufficient that the functional blocks 12 to 15, 151, and 152 can be formed in the above-described manner, and the recording form of the program may be arbitrary. For example, those recorded in a computer-readable form on a portable recording medium distributed in a state separated from the computer device, or taken in through a network connected to the computer device, and installed in the internal recording device or the like at the time of use It may be.

Although there are various methods for generating a reference typeface image in the image conversion unit 12, here, skeleton processing (processing for obtaining a skeleton line) is performed on a character image.
An example will be described in which a character image having a substantially constant stroke width, that is, a character image in a font similar to Gothic is generated.

FIG. 2 is a flowchart showing the process of determining the typeface in the character image processing system 1, and FIGS. 3 and 4 are conceptual diagrams showing the process of generating a reference typeface image. Hereinafter, a processing procedure of the character image processing system 1 will be described with reference to FIGS.
First, a character image is captured using the image scanner 11 (step S101). In the character image, the background is a white pixel (pixel value “0”), and the character portion is a black pixel (pixel value “25”).
5 "). Here, as shown in FIG. 3A, a coordinate system is assumed in which the origin is at the upper left point of the character image, the X axis is rightward, and the Y axis is downward. The coordinates of the lower right point of the image are (X, Y), and in a character image composed of a plurality of pixels, as shown in FIG.
Let i be the coordinate in the X direction and j be the coordinate in the Y direction of each pixel.

The image converter 12 converts the character image shown in FIG. 3B in the first direction, that is, from the (1,1) coordinate point to the (X, 1) coordinate point,
Scan from the (X, 1) coordinate point to the (1,2) coordinate point, from the (1,2) coordinate point to the (X, 2) coordinate point, and then sequentially scan in the direction of the (X, Y) coordinate point to obtain skeleton data Is created (step S102). Specifically, for a pixel (i, j) having a pixel value “255”, g (i,
j) is the pixel value for (i, j), g (i-1, j) and g (i, j-1)
Is smaller than g (i, j), g (i, j) is replaced by g (i-1,
It is assumed that 1 is added to the smaller value of j) and g (i, j-1). As a result, the character image in FIG. 3B is converted into skeleton data as shown in FIG.

Next, data correction is performed by scanning the skeleton data in FIG. 3C in the second direction (step S10).
3). That is, from the (X-1, Y-1) coordinate point to the (1, Y-1) coordinate point, and further from the (X-1, Y-2) coordinate point to the (1, Y-2) coordinate point Are sequentially scanned to the point (0,0). Then, for a pixel (i, j) having a pixel value other than “0”, the smaller of g (i + 1, j) and g (i, j + 1) is smaller than g (i, j). In this case, it is assumed that g (i, j) is obtained by adding 1 to the smaller value of g (i + 1, j) and g (i, j + 1). As a result, the skeleton data of FIG.
Is corrected as follows.

Further, the skeleton data shown in FIG. 3D is scanned in the first direction to perform re-correction (step S104).
That is, from the (1,1) coordinate point to the (X, 1) coordinate point,
2) Scan from the coordinate point to the (X, Y) point in order from the (X, 2) coordinate point. Then, for a pixel (i, j) having a pixel value other than “0”, g (i-1, j), g (i, j−1), g (i + 1, j), g (i , j + 1)) is less than or equal to g (i, j), the new value s (i, j) at that time is “1”, and the other values s (i, j) To “0”
And The point (i, j) where s (i, j) = “1” is the coordinate system forming the skeleton, and g (i, j) is the minimum point from the point (i, j) to the white pixel. Represents a distance value. FIG. 3 (e)
Shows an example of the skeleton data s (i, j) after re-correction.

In this way, a histogram of g (i, j) for all coordinate systems constituting the skeleton is created, and the most frequent value M is obtained (step S105). The probability density of the most frequent value M is assumed to be 60% or more, and if there is no corresponding value, g (i, j) corresponding to a cumulative probability density of 70%
Is M. Then, for all (i, j) where g (i, j) = M, draw a circle with a radius M centered at (i, j), formed from the pixels contained in these circles The image is used as a reference typeface image (step S106).

The typeface discriminating unit 14 compares the pixel distribution between the reference typeface image and the target character image (step S10).
7). Then, the ratio D (i, j) of the overlapping portion is obtained, and it is determined whether the character printed on the document is Gothic or Mincho based on the variance and average value of D (i, j) (step S108). As described above, since the reference typeface image is similar to the Gothic type, if the character image to be discriminated is of the Gothic type, the overlapping portion becomes relatively large.
Therefore, for example, an average value “0.8” or more and a variance “10”
If it is 0 "or less, it will be Gothic, otherwise it will be Mincho.

FIG. 4 shows the above processing more specifically. That is, the image conversion unit 12 creates a skeleton as shown in FIG. 4B for the line segment area shown in FIG. 4A, and creates a histogram of pixels for all coordinate systems constituting the skeleton. In this case, the most frequent value M (the frequent value is represented by the inner diameter of the circle) is n3 as shown in FIG. A circle as shown in FIG. 4E is drawn for the portion of n3, and an image of a region including pixels included in the circle is set as a reference typeface image. FIG. 4 (d)
This is a reference typeface image corresponding to the line segment area in FIG. 4A, and since the stroke width is substantially constant, it is close to a Gothic type.

The character image and the font information (Gothic / Mincho: including coordinates) are sent from the font discriminating unit 14 to the character recognizing unit 15.
Passed to. The character recognition unit 15 performs matching using the dictionaries 151, 152,... Corresponding to the determined typeface,
Perform character recognition. The recognition result is displayed on the display through the output control unit 16 and printed as needed.

Here, examples of experimental results of typeface discrimination for printed documents containing characters of different sizes selected from academic journals, magazines, and catalogs will be described. The printed document contained 3021 characters such as kanji, English characters, hiragana, katakana, and the like. Such a printed document has a resolution of 400
Scan with an image scanner at 300 dpi and dpi
As a result of performing the above-described typeface discrimination processing, of the 2035 characters in the Mincho font, those determined to be the Mincho font are 9 characters at 400 dpi.
6.1% and 95.6% at 300 dpi. Also,
Of the 986 Gothic characters, 94.5% were determined to be Gothic at 400 dpi and 9 at 300 dpi.
It was 3.2%.

As is evident from the experimental results, the method of the present embodiment is a simple method, but can correctly determine a font with sufficient accuracy for practical use. As a result, it is possible to perform typeface discrimination without increasing the load of the conventional character image processing system, and since the typeface discrimination result can be applied to character recognition, recognition accuracy can be significantly improved. Was.

(Second Embodiment) In the present invention, it is also possible to perform processing different from that of other character images on a specific character image by using the typeface discrimination result. For example, when a character image is displayed on a display, the display style of a specific font may be different from that of another font, or only specific fonts may be extracted and rearranged.

FIG. 5 is a block diagram of the character image processing system in this case, and the same elements as those of the first embodiment are denoted by the same reference numerals. The character image processing system 2 includes a setting unit 22 for setting various parameters and a character editing (character typeface, rearrangement, etc.) based on the parameters set in the setting unit 22 at a subsequent stage of the character recognition unit 15. An editing unit 23 for performing the editing is added. These added functional blocks 22 and 23 are also formed by the computer device reading a predetermined program and executing it in cooperation with the OS of the own device.

FIG. 6 is a diagram showing a processing procedure when performing character emphasis processing in the character image processing system 2. In FIG. 6, up to the character recognition processing based on the typeface determination result is the same as in the first embodiment (step S
201). Here, the emphasis parameter set in the setting unit 22, that is, a parameter indicating which typeface is to be emphasized in what manner is acquired by the editing unit 23 (step S202), and after the corresponding typeface character and its position are detected. (Step S203), an emphasis process based on the emphasis parameter is executed (Step S204). As a form of the emphasis processing, for example, the size of a character determined to be Gothic is made relatively large, colored in red or the like, or blinking. The result data of the emphasis processing is output to the output control unit 1.
6 and output to a display device or the like, and displayed in an appropriate form.

FIG. 7 is a diagram showing a processing procedure when rearranging characters. In FIG. 7, up to the character recognition processing based on the typeface determination result is the same as in the first embodiment (step S301). Here, the arrangement parameter set in the setting unit 22, that is, a parameter indicating which typeface should be arranged in what order and how, is acquired by the editing unit 23 (step S302), and the corresponding typeface character and its position are obtained. After the extraction (step S303), a character rearrangement process based on the arrangement parameters is executed (step S304). For example, in the case of a document in which Mincho and Gothic are mixed, the Gothic is often used when the document creator wants to stand out. Therefore, in such a case, if only the Gothic body is extracted and rearranged, the intention and the main point of the document creator can be quickly grasped. The rearranged character group is output to a display device or the like through the output control unit 16 and displayed.

The processing shown in FIGS. 6 and 7 may be performed alternatively by the editing unit 23, or may be performed using both of them. For example, when rearranging characters, the display mode may be changed or the color may be changed.

In the first and second embodiments, the reference typeface is similar to the Gothic typeface. However, the reference typeface may be Mincho typeface or any other typeface.

[0033]

As is apparent from the above description, according to the present invention, it is possible to accurately and quickly determine a typeface, and to enhance the practical value in performing character recognition and character editing. , Has a unique effect.

[Brief description of the drawings]

FIG. 1 is a functional block configuration diagram of a character image processing system according to a first embodiment of the present invention, illustrating an example of performing character recognition.

FIG. 2 illustrates a character image processing system according to a first embodiment.
FIG. 4 is an explanatory diagram of a procedure of a typeface determination method.

3A is an explanatory diagram of a coordinate system set in image data, FIG. 3B is an example of actual image data, FIG. 3C is skeleton data, FIG. 3D is skeleton data after correction, and FIG. () Is a diagram showing an example of skeleton data after re-correction.

4A is a line segment area forming a character, FIG. 4B is a skeleton of a line segment, FIG. 4C is a most frequent pixel value M (frequent value: inner diameter of a circle) in the skeleton, and FIG. FIG. 4A is a diagram illustrating a reference font image, and FIG. 4E is a diagram illustrating a pixel region when a reference font image is obtained.

FIG. 5 is a functional block configuration diagram of a character image processing system according to a second embodiment of the present invention.

FIG. 6 is an explanatory diagram of a procedure of an emphasis process according to the second embodiment.

FIG. 7 is an explanatory diagram of a procedure of a character rearrangement process according to the second embodiment.

[Explanation of symbols]

 1, 2 Character image processing system 11 Image scanner 12 Image conversion unit 13 Reference type image storage unit 14 Type identification unit 15 Character recognition unit 151, 152 Character recognition dictionary for each type 16 Output control unit 22 Setting unit 23 Editing unit

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Tadahiro Kitahashi 3-1-47 Ueno Nishi, Toyonaka-shi, Osaka (72) Inventor Akira Wei 1-130-203 Fujishirodai, Suita-shi, Osaka

Claims (8)

[Claims] 1. A step of generating a reference type face image having a specific shape from a character image whose typeface is unknown, and comparing the pixel distribution between the character image and the generated reference type face image to change the typeface of the character image. Determining the typeface of the printed characters. 2. The typeface discriminating method according to claim 1, wherein the reference typeface image is a character image in which all stroke widths are substantially constant. 3. The typeface discriminating method according to claim 1, further comprising a step of performing a specific process different from other character images on a specific one of the character images whose typefaces have been discriminated. 4. A plurality of character recognition dictionaries prepared for each typeface, image converting means for generating a reference typeface image having a specific shape from a character image whose typeface is unknown, said character image and said generated reference typeface image And a means for selecting a character recognition dictionary corresponding to the determined typeface and performing character recognition on the character image by comparing the pixel distribution of the character image and the typeface of the character image. Character image processing system. 5. An image conversion means for generating a reference typeface image having a specific shape from a character image whose typeface is unknown, and comparing the pixel distribution between the character image and the generated reference typeface image by comparing the pixel distribution of the character image. A character image processing system comprising: a typeface determining unit that determines a typeface; and a unit that performs emphasis processing on a specific one of the character images whose typefaces have been determined. 6. An image conversion means for generating a reference typeface image having a specific shape from a character image whose typeface is unknown, and comparing the pixel distribution between the character image and the generated reference typeface image to obtain a reference type image. A character image processing system comprising: a typeface determining unit that determines a typeface; and a unit that extracts and rearranges a specific one of the character images that have determined the typeface. 7. The image conversion unit according to claim 1, wherein the character image is subjected to a skeleton process to generate a character image having a substantially constant stroke width, and the character image is used as the reference typeface image. The character image processing system according to claim 3. 8. A process of performing a skeleton process on a character image whose font type is unknown to generate a reference font image in which all stroke widths are substantially constant; and a pixel distribution between the character image and the generated reference font image. A computer-readable recording medium storing a program for causing a computer device to execute a process of determining the typeface of the character image by comparing