JP2002185782A - Character extracting device, character extracting method and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character extracting device that can efficiently extract characters on a postcard or an original having a background without being affected by the background. SOLUTION: The character extracting device that extracts character on a postcard or an original having a background, reads a character image on the postcard or the original having a background pattern, extracts a major color configuring the image, separates the image into images where the extracted major color is used for a median, selects the character image among the separated images, and extracts characters from the image resulting from applying expansion or contraction processing to the selected character image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character extracting apparatus, a character extracting method, and a recording medium for extracting a lottery number of a postcard with a winning number, such as a New Year's postcard or a mail, and a character of a document having a background. In particular, it reads the text image of a postcard or manuscript with a pattern on the background, extracts the main colors that make up the image, separates them into an image with the extracted main color value as the center value, and from among the separated images A character image is selected, and characters are extracted from an image obtained by expanding or contracting the image.

[0002]

2. Description of the Related Art Conventionally, a process for automatically extracting a lottery number of a postcard with a winning number, such as a New Year's postcard or an e-mail, is performed by using a pattern portion which hinders the extraction of the lottery number. In the case of email, the same color as that of a blue-colored net pattern is emitted, and a drop-out color method is used to invalidate the color of the pattern portion.

[0003]

However, in the prior art, the red pattern in a New Year's postcard has a color variation, and characters cannot be accurately extracted by a single dropout color irradiation. Therefore, there is a problem that efficiency is poor because it is necessary to adjust.

[0004] In addition, if it is made possible to carry out the same kind of mail, if the mail has a lottery number with a blue pattern background, it is determined whether the lottery number is red or blue and the light source of the dropout color is accordingly determined. Has to be switched, and there is a problem that the hardware cost becomes very high.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in consideration of the above-described problems, and is provided with a character extracting apparatus, a character extracting method, and a recording medium for extracting the characters of a postcard or an original having a pattern on the background. In, read the character image of a postcard or manuscript with a pattern in the background, extract the main colors that make up the image, separate the image with the center value of the extracted main color value, from among the separated images By selecting a character image and extracting characters from an image obtained by subjecting the image to expansion or contraction processing, characters on a postcard or an original with a background can be efficiently extracted without being affected by the background.

[0006] In this specification, the term "simple clustering technique" refers to a set of N samples X =
If there is [X1, X2,..., Xn], the threshold T is set, and first, an arbitrary sample, for example, X1, is taken, and this is set as the center of the cluster, Z1 (= X1). Next, Xi
(I = 2,..., N) and the distance D1i between Z1 and Xi
Is calculated, and if D1i = <T, it is determined that Xi belongs to a cluster centered on Z1. X if D1i> T
Let i be the new cluster center Z2. Therefore, the distances D1i and D2i from the remaining samples Xi to Z1 and Z2 are calculated. If any of D1i and D2i is smaller than T, Xi belongs to the cluster. Otherwise, Xi is set to the new cluster center. Let it be Z3. In this way, it refers to a method of examining all samples [X1, X2,..., Xn] and dividing them into several clusters.

[0007] In this specification, the term "K-means clustering technique" refers to a set of N samples X
= [X1, X2,..., Xn], K initial cluster centers Z1 (1), Z2 (1),..., Zk (1) are appropriately determined, and the sample [X] is set to the initial cluster center Z1.
The clusters are classified into K clusters so as to belong to the closest cluster from (1), Z2 (1),..., Zk (1).
Next, the centers Z1 (2) and Z2 of the K clusters classified
(2),..., Zk (2) are obtained, and the sample [X] is similarly set to the cluster centers Z1 (2), Z2 (2),.
Are classified into K clusters so as to belong to the closest cluster. This refers to a method of repeating this until the cluster converges.

In this specification, the term "labeling process" is described by taking a binary image as an example.
This is a process for calculating the number of black aggregates in an image. Points belonging to the same (continuous) black aggregate region are labeled with the same label, points belonging to different (discrete) black aggregate regions are labeled with another label, and the same label is labeled. Refers to the process of enclosing the black aggregate area with a rectangular area.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (1) In a character extracting apparatus for extracting characters on a postcard or a document having a pattern on a background, an image reading section for reading a postcard or a character on a document having a pattern on a background as an image, and an image reading device A representative color extracting unit for extracting a main color (representative color) constituting an image from an image read by the image reading unit, and a central value representing a value of the main color extracted by the representative color extracting unit for the image read by the image reading unit. The image separation unit that separates the images into images to be separated, and the images separated by the image separation unit are each subjected to labeling processing to obtain a rectangular area, and the character image is determined by the number, arrangement, regularity of arrangement, and line segment configuration within the rectangular area. Character image determination unit, and a character image correction unit that corrects the image by performing expansion or contraction processing on the image determined by the character image determination unit, so that a postcard with a background or a character on a document is provided. Without being to be affected by the background it can be extracted efficiently.

(2) In the character extracting apparatus described in (1), a representative color extracting section for calculating main colors constituting an image from an image read by the image reading section by a simple clustering technique, and a calculation by the representative color extracting section. An image separation unit that separates the image read by the image reading unit by the K-means clustering method using the main color values as core values, automatically converts the image into the main image of the background, pattern, and characters. Can be separated.

(3) In a character extracting method for extracting characters on a postcard or an original having a pattern on a background, a character on a postcard or an original having a pattern on a background is read as an image, and a main image forming an image from the read image is provided. Extracts colors, separates the read image into images with the extracted main color values as center values, labels each separated image to obtain a rectangular area, and determines the number, arrangement, and arrangement rules of the rectangular area sex,
A character image is determined based on a line segment configuration in a rectangular area, and the image determined as the character image is expanded or contracted to correct the image, so that a postcard with a background or a character on a document is affected by the background. And can be efficiently extracted.

(4) On a recording medium storing a tool for extracting a character from an image of a postcard or an original having a pattern on the background, a character of the postcard or the original having a pattern on the background is read as an image, and the image is read from the read image. The main colors to be composed are extracted, the read image is separated into images having the extracted main color values as center values, and the separated images are subjected to labeling processing to obtain rectangular regions, and the number and arrangement of the rectangular regions. A computer executes a recording medium storing a program for determining a character image based on the regularity of arrangement and the line segment configuration in a rectangular area, and expanding or contracting the image determined as the character image to correct the image. As a result, it is possible to efficiently extract a postcard with a background or characters on a document without being affected by the background.

[0013]

FIG. 1 is a block diagram showing the configuration of the present invention. In the figure, reference numeral 1 denotes a character extracting device. The character extracting device 1 forms an image from an image reading unit 11 that reads a postcard having a pattern in the background or characters on a document as an image, and an image read by the image reading unit 11. A representative color extraction unit 12 for extracting a main color (representative color); and an image separation unit for separating an image read by the image reading unit 11 into an image having a center value of a main color value extracted by the representative color extraction unit 12. A character image that determines a character image based on the number, arrangement, regularity of arrangement, and line segment configuration within the rectangular area by obtaining a rectangular area by performing labeling processing on each of the images separated by the image separating unit 13 and the image separating unit 13 Judgment unit 14
Character image correction unit 1 that corrects an image by subjecting the image determined by character image determination unit 14 to expansion or contraction processing
5 is comprised.

The character extracting method of the present invention will be described with reference to FIGS. FIG. 2 is a processing flowchart of one embodiment of a method of extracting a character of a lottery number of a postcard with a winning number, and FIG. 3 is an explanatory diagram of a method of extracting a character of a lottery number. This example is an example of extracting a lottery number with a pattern on the background of a postcard with a winning number.

The operation will be described below with reference to the flow chart of FIG.

Step S201: The image of the lottery number portion of the postcard is read. The read image has a 6-digit lottery number on the pattern as shown in FIG. 3A, and is composed of three main colors of the pattern (red), the lottery number (black), and the background (white). Is what it is.

Step S202: The main colors constituting the image are determined from the read image by a simple clustering method. As a result, three colors close to the main red, black, and white that constitute the image are obtained. The details of this process will be described later.

Step S203: Using the obtained main color (representative color) as a core value, the image of the read lottery number portion is separated by the K-means clustering method. The separated image is separated into background, pattern, and character images as shown in FIG. The details of this processing will be described later.

Step S204: Labeling processing is performed on each of the separated images to obtain a rectangular area, and a character image is determined based on the number and arrangement of the rectangular areas. Note that the separated image is as shown in FIG. 3C by the labeling process, a large number of very small rectangular areas arranged separately are obtained for the background image, and one rectangular area is obtained for the pattern image. , The character images are arranged in a line in the horizontal direction.
One rectangular area is required. Therefore, since the lottery number to be extracted is known to be 6 digits, the character image is 6
It can be determined that two rectangular areas are obtained images.

When the lottery number of the postcard is to be extracted as in this example, since the number of characters of the lottery number is known in advance, a character image is determined based on the number and arrangement of the rectangular areas. In the case of extracting the character image, the character image may be determined by analyzing the number, arrangement, regularity of arrangement, line segment configuration in the rectangular area, and the like.

Step S205: The character image is expanded / contracted to correct the image so that the character can be easily recognized. The character image is expanded as shown in FIG. 3D, and the recognition rate of character recognition performed in the next process can be increased. Then, the process ends.

By performing such processing, it becomes possible to efficiently extract postcards with backgrounds and characters on originals with a low-cost hardware configuration.

Next, the simple clustering process in step S202 and the K-means clustering process in step S203 in FIG. 2 will be described in detail.

FIG. 4 shows a processing flowchart of one embodiment of the simple clustering processing. Hereinafter, the operation will be described according to this flow.

Step S401: A distance (permissible value) is set as a reference as to whether or not the object belongs to the same class. Normally, the distance is a Lab that matches the value of each pixel described in the RGB coordinate system, which is a color system of a computer system, to the human color sense.
The coordinate system is converted to a coordinate system, and a color difference distance = 8 to 12 at which a human judges a different color in the Lab space is set.

Step S402: Any one of the target images
One pixel is selected, one class Zi (i = 1) is set, and the center value is set.

Step S403: A pointer P pointing to the first pixel of the target image is set.

Step S404: It is checked whether the pixel indicated by the pointer P belongs to any of the set classes Zm (m = 1 to n). That is, the distance between the pixel pointed by the pointer P and the center value of the set class Zi is obtained, and it is determined whether the distance is within the allowable value.

Step S405: If the processing result of step S405 is within the allowable value, the process proceeds to step S407, and if not, the process proceeds to step S406.

Step S406: The pixel pointed by the pointer P is set to a new class Zi (i = i + 1), and the pixel pointed to by the pointer P is set to a class Zi (i = i +
The central value of 1) is used. Then, the process returns to step S404.

Step S407: The pixel pointed to by the pointer P is assigned to a class within an allowable value.

Step S408: It is determined whether or not all pixels of the target image have been processed. If all pixels have been processed, the process proceeds to step S410. If all pixels have not been processed, the process proceeds to step S409.

Step S409: The pointer P is updated so as to point to the next pixel.

Step S410: The upper three classes are selected in descending order of the number of pixels belonging to the class, the center value is set as a representative value, and the result is passed to the subsequent processing. Then, the process ends.

By performing such processing, three representative colors constituting the target image can be obtained.

FIG. 5 shows a flowchart of an embodiment of the K-means clustering process. Hereinafter, the operation will be described according to this flow.

Step S501: The representative values of the three colors obtained by the simple clustering process are set as the first three core values (center values).

Step S502: classify all pixels of the target image into a class having the smallest color difference among the first three kernel values;
In other words, they are assigned to belong to the class with the most similar color.

Step S503: With respect to the three classes, the representative values (average values) of the respective classes are obtained, and are set as second three core values.

Step S504: It is determined whether the color difference between the first kernel value and the second kernel value is equal to or less than a predetermined value. If the value is equal to or smaller than the predetermined value, the process proceeds to step S506.

Step S505: The obtained second three kernel values are set to the first three kernel values. And step S
Return to 502.

Step S506: The three classes are transferred to the subsequent processing as an image obtained by separating the target image into three representative colors. Then, the process ends.

By performing the above-described processing, the target image can be separated into images of three optimum representative colors.

[0044]

The present invention is embodied in the form described above, and has the following effects.

It is possible to efficiently extract the lottery number without being affected even if the color of the red pattern has a variation like the New Year's postcard or the blue pattern like the mail. Become. In addition, characters on a document having a background can be efficiently extracted without being affected by the background.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of the present invention.

FIG. 2 is a processing flowchart of one embodiment of a method for extracting a character of a lottery number of a postcard with a winning number;

FIG. 3 is an explanatory diagram of a character extraction method of a lottery number.

FIG. 4 is a processing flowchart of one embodiment of a simple clustering process.

FIG. 5 is a processing flowchart of one embodiment of a K-means clustering process.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 character extracting device 11 image reading unit 12 representative color extracting unit 13 image separating unit 14 character image determining unit 15 character image correcting unit

Claims (4)

[Claims] An image reading unit (11) for reading a postcard or a character on a document having a pattern on the background as an image, and a representative for extracting main colors constituting the image from the image read by the image reading unit (11). A color extracting unit (12); an image separating unit (13) for separating an image read by the image reading unit (11) into an image having a center value of a main color value extracted by the representative color extracting unit (12); Character image determination for determining a rectangular image by labeling each of the images separated by the image separating unit (13) to determine a rectangular region, and determining the character image based on the number, arrangement, arrangement regularity of the rectangular regions, and the line segment configuration in the rectangular region A character extraction device comprising: a unit (14); and a character image correction unit (15) that corrects the image by performing expansion or contraction processing on the image determined by the character image determination unit (14). 2. The character extracting device according to claim 1, wherein a representative color extracting unit (12) for calculating a main color forming the image from the image read by the image reading unit by a simple clustering method, and a representative color extracting unit. A character extraction device comprising: an image separation unit (13) that separates an image read by an image reading unit by a K-means clustering method using the main color values calculated in (1) as kernel values. 3. A postcard or a character on a document having a pattern on a background is read as an image, a main color constituting the image is extracted from the read image, and a value of the main color obtained by extracting the read image is a central value. Is separated into images, and the separated images are each labeled to obtain a rectangular area, and the number of rectangular areas, arrangement, regularity of arrangement, and line segment configuration within the rectangular area are used to determine a character image. A character extracting method, wherein the determined image is expanded or reduced to correct the image. 4. A recording medium storing a tool for extracting character information from an image including character information, wherein a postcard having a pattern on a background or a character on a document is read as an image, and a main image constituting the image from the read image is read. The color is extracted, the read image is separated into images having the extracted main color values as center values, and the separated images are respectively subjected to labeling processing to obtain rectangular regions.
A computer-readable recording medium storing a program for determining a character image based on the regularity of arrangement and the line segment configuration in a rectangular area, and expanding or contracting the image determined as the character image to correct the image.