JPH10171924A - Character recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character recognizing device which recognizes image data with high accuracy by effectively eliminating linear noise that is included in the image data. SOLUTION: A character segmenting part 30 segments a character part from image data, a histogram generating part 50 counts one value in binary data of the image data in each line or each string and generates a histogram, and a peak value detecting part 60 detects a peak value of the histogram. A noise deciding part 70 decides whether the detected peak value is noise or not and when it decides that it is noise, a noise eliminating part 40 eliminates the noise.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for removing noise from image data of characters and symbols (hereinafter referred to as images) and converting the image data from which noise has been removed into character codes such as characters. About.

[0002]

2. Description of the Related Art Conventionally, a character recognition device that cuts out a character or the like from image data including a printed character or a handwritten character, performs character recognition and converts the character into a character code,
In order to improve the accuracy of character recognition, various methods have been proposed for removing noise other than images included in image data. For example, there is a method in which an isolated point that is separated from the image and is isolated from the image data is searched, and if the size of the isolated point is equal to or smaller than a predetermined value, the isolated point is regarded as noise and removed. Also, assuming that the image data is represented by, for example, black and white binary data, and then the image data is subdivided into dots,
In this method, for example, the number of black points in the area near the subdivided dot is counted, and if the number of black points is equal to or less than a predetermined value, the dot itself is not a black point.

[0003]

However, the above-mentioned method is, for example, a scattered read noise which is often found in image data read by a fixed document type optical reading device (such as a fixed document type scanner). For example, although noise is effectively removed, for example, noise removal may not work for a specific linear noise that is often seen in image data sent by a document feed type reading device (FAX or the like). is there. That is, when an image written on a sheet is transmitted by facsimile, the image is often read while moving the sheet and transmitted as image data. Then, as shown in FIG. 1, a linear noise may enter in the direction (vertical direction) in which the paper is moved. However, the conventional noise elimination method cannot cope with such noise.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and effectively eliminates linear noise contained in image data sent by facsimile or the like to thereby improve the quality of image data. It is an object of the present invention to provide a character recognition device with improved recognition accuracy.

[0005]

According to a first aspect of the present invention, there is provided a character recognition apparatus for converting an image of a character or a symbol, etc., into binary data, the image data other than the image. A character recognition device comprising: a noise removing unit that removes noise; and a character recognition unit that recognizes image data from which noise has been removed by the noise removing unit and converts the image data into a character code. Of the binary data, one of the values is counted for each row or column, and a histogram creating means for creating a histogram; a peak value detecting means for detecting a peak value of the histogram; and a peak value detecting means. A noise determination unit configured to determine whether the detected peak value is caused by noise.

In the character recognition device of the present invention having the above-described structure, when image data representing an image of a character or a symbol by binary data is sent, the noise removing means extracts noise other than the image from the image data. Is removed. That is, the histogram creating means creates a histogram by counting one of the binary data of the image data for each row or each column,
The peak value detecting means detects a peak value of the histogram. Then, the noise determination means determines whether or not the peak value detected by the peak value detection means is caused by noise, and when it is determined that the noise is noise, the noise is removed. Then, the character recognition unit recognizes the image data from which noise has been removed and converts the image data into a character code.

[0007] The character recognition device according to claim 2 is
2. The character recognition device according to claim 1, wherein the noise determination unit is configured such that when a width of a peak value detected by the peak value detection unit is smaller than a width of an outline of the image,
The peak value is determined to be due to noise.

[0008] Furthermore, a character recognition device according to claim 3 is the character recognition device according to claim 1 or 2,
The noise removing unit removes the noise as noise only when binary data adjacent to the peak value detected by the peak value detecting unit is different from the data of the peak value.

[0009]

Embodiments of the present invention will be described below with reference to the drawings.

In this embodiment, the character recognition apparatus of the present invention is mounted on a facsimile apparatus.

FIG. 2 is a block diagram showing a schematic configuration of the character recognition device according to the present embodiment. As shown in FIG.
The AX data receiving unit 10 is connected to the external line 100. On the other hand, the FAX data receiving unit 10 is also connected to a character extracting unit 30 and a preprocessing unit 20 including a noise removing unit 40 as a noise removing unit. Further, the noise removing unit 40 includes a histogram creating unit 50 as a histogram creating unit, a peak value detecting unit 60 as a peak value detecting unit, and a noise determining unit 70 as a noise determining unit. Further, the preprocessing unit 20 is connected to a character recognition unit 80 as a character recognition unit, and the character recognition unit 80 is connected to a character recognition dictionary 90.

In the character recognition device configured as described above, when image data representing an image as binary data by another facsimile is transmitted through the external line 100, F
The AX data receiving unit 10 receives the image data. The received image data is processed by the preprocessing unit 20. That is, the preprocessing unit 20 includes a character cutout unit 30 and a noise removal unit 40, and the character cutout unit 30 cuts out the image data sent from the FAX data receiving unit 10 for each image.

The noise elimination unit 40 eliminates noise other than the image extracted by the character extraction unit 30. Further, the noise removing unit 40 includes a histogram creating unit 50, a peak value detecting unit 60, and a noise judging unit 70. The histogram creating unit 50 converts the binary data of the image data extracted by the character extracting unit 30 into, for example, “ One value of “black” and “white”, for example, “black” is counted for each row or each column to create a histogram. Further, the peak value detecting section 60 detects a peak value of the histogram created by the histogram creating section 50. Further, the noise determination unit 70
Determines whether the peak value detected by the peak value detection unit 60 is noise, and when the peak value is determined to be noise, the noise removal unit 40 removes the noise. The character recognition unit 80 recognizes the image data processed by the pre-processing unit 20 and refers to a character recognition dictionary 90 described later,
Convert to character code. The character recognition dictionary 90 stores data of characters to be referred to when converting image data into character codes.

Next, an example of the case of recognizing a number entered in a frame as shown in FIG. 1 will be described with reference to a flowchart of FIG.

First, in step 1 (hereinafter abbreviated as S1; the same applies to other steps), the FAX data receiving unit receives image data from the external line 100. FIG. 1 shows an example of the image data received at this time, and the numerals (1234, 5, 6789) in the frame of FIG. In S2, the character cutout unit 30 cuts out one character from the number in the frame. First cut out is school number 1
It is assumed that the first digit of the school number is “2”, but the second digit of the school number is “2” in order to explain the noise removal. FIG. 3A shows an image image cut out at this time.

Next, in S3, the histogram creation unit 50
Creates a histogram by dividing image data of one character into dots and counting one value, for example, "black", of binary data of the divided dots, for example, "white" or "black" in the column direction. I do. A histogram is created from the image of FIG. FIG. 3B shows a histogram created from the image data of FIG. 3A. Where there is noise, the number of black spots is large and the histogram protrudes. In S4, the peak value detection section 60 detects a peak value. When a position where the value of the histogram takes a local maximum value is detected as a peak value, a position containing noise is detected as a peak value.

Next, in S5, the noise determination section 70 determines whether or not the peak value is noise. In the present embodiment,
The peak value is determined to be noise if the width over which the value of the peak value is 0.9 times or more continues is within two dots and the value of the peak value is more than twice the left or right value. I do.
In FIG. 3B, the peak value is 50, the width following the peak value is one dot width, and the left or right value of the peak value is 15 dots.
Therefore, the above condition is satisfied, and this peak value is determined to be noise. When determining whether or not noise is present, the width of the peak value is used as a condition.
In order to prevent the part from being judged as noise and not to be deleted, when the peak value appears with a width smaller than the average thickness of the character outline, it is determined to be noise. It is.

If it is determined in S5 that the noise is present, S
In 6, the noise removing unit 40 removes noise. In this case, if all of the “black” dots at the peak value are removed, a part of the character may also be removed. Only when the dot located at a position other than the peak value is not a “black” dot, the “black” dot is removed as noise. In the example of FIG. 3, since the width of the peak value is one dot width, if both the left and right dots among the “black” dots on the peak value are “white” dots, the peak value "Black" dots on the value are removed as noise. Then, the black point at the portion where the noise intersects with the outline of “2” is not removed as noise, so that a part of the outline of “2” is not lost. Then, in S7, the character recognition unit 80 performs character recognition on the image data from which noise has been removed.

Thereafter, in S8, it is further checked whether or not there is a character to be recognized. If it is determined that there is a character, the process returns to S2, and the character cutout unit 30 again cuts out the next character in the frame by one character. In the example of FIG. 1, the third digit "3" of the school number is cut out, and the above processing is repeated.

However, "3" in the third digit of the school number is S5
Since the peak value is not determined to be noise, the noise removal processing in S6 is not performed, and the character recognition processing in S7 is performed.

In this embodiment, an example has been shown in which the numbers in the frame are cut out one character at a time to remove noise. However, noise may be removed after cutting out a character area of a normal text without a frame. Characters may be cut out after removing noise.

The characters may be printed characters or handwritten characters.

Further, in the present embodiment, an example in which the character recognition apparatus of the present invention is mounted on a facsimile apparatus has been described as an example, but it is apparent that the apparatus can be mounted on another apparatus.

[0024]

As is apparent from the above description,
According to the character recognition device of the first aspect of the present invention, one of the binary data of the image data is counted for each row or each column, and the histogram is created. The linear noise can be effectively removed, and the recognition accuracy of the image data can be increased.

According to the character recognition device of the present invention, when the width of the peak value of the histogram is smaller than the width of the outline of the image, the peak value is determined to be noise. It is possible to prevent a part from being erroneously determined to be noise.

Furthermore, according to the character recognition device of the present invention, only when the binary data adjacent to the peak value is different from the peak value data, the binary data is removed as noise. It is possible to further prevent a part from being erroneously removed as noise.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of image data sent by FAX.

FIG. 2 is a block diagram illustrating a schematic configuration of a character recognition device according to the present embodiment.

3A and 3B are diagrams illustrating data of one character cut out from image data, in which FIG. 3A is a diagram illustrating image data of one character cut out, and FIG. 3B is a diagram calculated from image data of one character. FIG. 6 is a diagram showing a histogram.

FIG. 4 is a flowchart showing the operation of the present embodiment.

[Explanation of symbols]

 Reference Signs List 40 noise removal unit 50 histogram creation unit 60 peak value detection unit 70 noise judgment unit 80 character recognition unit 90 character recognition dictionary

Claims (3)

[Claims] 1. A noise removing means for removing noise other than an image such as an image such as a character or a symbol from binary data represented by binary data, and recognizing the image data from which noise has been removed by the noise removing means. In the character recognition device having a character recognizing means for converting into a code, the noise removing means counts one value of the binary data of the image data for each row or column and creates a histogram. A histogram creation unit; a peak value detection unit that detects a peak value of the histogram; and a noise determination unit that determines whether a peak value detected by the peak value detection unit is caused by noise. Character recognition device. 2. The method according to claim 1, wherein the noise determining means determines that the peak value is caused by noise when the width of the peak value detected by the peak value detecting means is smaller than the width of the outline of the image. The character recognition device according to claim 1, wherein 3. The method according to claim 2, wherein the noise removing unit removes the noise as noise only when binary data adjacent to the peak value detected by the peak value detecting unit is different from the data of the peak value. The character recognition device according to claim 1.