JPH11312218A - Character optimally binarizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character optimally binarizing device, capable of correctly segmenting a character by calculating an optimal binarized threshold value for each character. SOLUTION: This device consists of a histogram acquiring means which acquires the histogram by each character, a histogram smoothing means 2 taking the moving average of histogram to smooth the histogram, a histogram valley position detecting means 3 detecting the valley position of the histogram, a histogram evaluation value calculating means 4 for calculating an evaluation value expressing the form of the histogram, a binarized threshold setting means 5 for setting the binarized threshold from a histogram evaluation value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optimal character binarizing apparatus used for character recognition, and more particularly to an optimal character binarizing apparatus for each character.
A threshold value is determined and a character can be cut out one character at a time based on the threshold value.

[0002]

2. Description of the Related Art A binarizing apparatus for extracting characters and binarizing character information is commonly provided in an OCR (optical character reading apparatus). 2. Description of the Related Art A binarizing device of an OCR (optical character reading device) usually binarizes all characters described in a form with a fixed threshold.

Conventionally, characters written on a form are of various character types such as printing, stamping, and handwriting. Further, the density of characters such as thin, dark, blurred, and stained with noise is also reduced. Various. If all the characters described in the form are binarized with a fixed threshold, the binarized character image will be blurred or crushed, which will adversely affect later character recognition.

Accordingly, for example, Japanese Patent Application Laid-Open No. 61-14607
According to the technology disclosed in Japanese Patent Publication No. 0-200, an optimal binarization process according to image characteristics can be performed by using a binarization circuit having a different threshold value in addition to a character binarization circuit for area determination. However, there is a problem that the circuit configuration becomes complicated because a binarization circuit having a different threshold value must be prepared separately from the character binarization circuit.

[0005]

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a character optimum binarizing apparatus which determines an optimum binarization threshold for each character based on a histogram and can cut out an accurate character. With the goal.

[0006]

In order to achieve the above object, the present invention provides a histogram acquiring means for acquiring a histogram for each character, a histogram smoothing means for taking a moving average of the histogram and smoothing the histogram, Histogram valley position detection means for detecting the valley position of the histogram, histogram evaluation value calculation means for calculating the evaluation value representing the shape of the histogram,
It is characterized by comprising a binarization threshold value determining means for determining a binarization threshold value.

With the above configuration, according to the present invention, an optimum binarization threshold can be determined for each character, so that a correct character can be cut out.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention according to claim 1 of the present invention comprises a histogram acquiring means for acquiring a histogram for each character, a histogram smoothing means for taking a moving average of the histogram and smoothing the histogram, and a histogram smoothing means. The apparatus includes a histogram valley position detecting means for detecting a valley position, a histogram evaluation value calculating means for calculating an evaluation value representing the shape of the histogram, and a binarization threshold value determining means for determining a binarization threshold value from the histogram evaluation value. This is a character-optimized binarizing apparatus characterized in that it has an effect that an optimal binarization threshold can be determined for each character.

According to a second aspect of the present invention, there is provided a histogram obtaining means for obtaining a histogram for each character,
Histogram smoothing means for taking a moving average of the histogram and smoothing the histogram; histogram valley position detecting means for detecting a valley position of the histogram; histogram evaluation value calculating means for calculating an evaluation value representing the shape of the histogram; A binarization threshold value deciding unit for deciding a binarization threshold value from a value, a black pixel ratio calculation unit for calculating a black pixel ratio with respect to the entire area after binarization, This is a character optimum binarization device characterized by comprising a binarization threshold value correcting means for correcting, and it is possible to calculate an optimum binarization threshold value for each character property from a histogram distribution. Having.

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

(First Embodiment) FIG. 1 is a block diagram showing a configuration of a character optimum binarizing apparatus according to a first embodiment of the present invention. In FIG. 1, the character optimum binarization device includes a histogram acquisition unit 1 for acquiring a histogram for each character, a histogram smoothing unit 2 for taking a moving average of the histogram and smoothing the histogram, and detecting a valley position of the histogram. Histogram valley position detection means 3
A histogram evaluation value calculation means 4 for calculating an evaluation value representing the shape of the histogram;
And a binarization threshold value determining means 5 for setting a binarization threshold value.

Next, the operation of the character optimum binarizing device according to the first embodiment of the present invention will be described. The histogram acquisition unit 1 acquires a histogram for each character as is well known. The histogram smoothing means 2 smoothes the histogram by taking a moving average of the acquired histogram as is well known.

The histogram valley position detecting means 3 detects a valley position of the smoothed histogram. The detection of the histogram valley position will be described in more detail with reference to FIGS.

FIG. 2 shows the general tendency of a normal character histogram. In FIG. 2, generally, a rectangular histogram circumscribing a character includes a background portion in a mountain having the lowest density. Since the binarization of the character is performed to separate the character from the background, it is necessary to find the end of the background, that is, the valley position th_l of the peak with the lowest density. Also, since the peak on the side with the highest concentration is the peak in the character part,
The valley position th_r of the peak having the highest concentration is obtained. The optimal binarization threshold is set between th_l and th_r.

FIGS. 3 and 4 show a histogram valley position th_.
14 explains a calculation procedure for calculating l and th_r. Multiply the character histogram by a moving average with a moving average width of 17. In the terminal, when the gradation sampling number is less than 17, the calculation is performed assuming that the data of the sample value 0 exists. The histogram is expressed as h for the gradation value x (0 ≦ x ≦ 255).
h_s (x) is the value obtained by multiplying (x) and h (x) by the moving average. 5
H_sub (x) = h_s (x + 5) −h_s
Define (x). Starting from x = 0, the first gradation value that satisfies h_sub (x) <0
If x0, h_sub (x) ≤ 0 and h_sub
A gradation value th_l1 = x + 1 that satisfies (x)> 0 is obtained. The number of continuous rises after th_l1 (the number of samples that continuously satisfy h_sub (x) ≧ 0) is determined as up_continue_count. h_ continuously for th_l1 ≦ x ≦ th_l1 + up_continue_count
The number up_rate_count that satisfies sub (x)> 2 is obtained (see the right end of FIG. 3). When up_rate_count ≧ 5, it is determined that there is a peak after that, and th_l1 is set to the valley position th_l, but h_sub before th_l1
If (x) = 0 is continuously present, th_l1 is returned to the start point th_l2 of h_sub (x) = 0, and the process is terminated. When up_rate_count <5 (see FIG. 4), a number flat_count that becomes h_sub (x) = 0 or −1 continuously before th_l1 and its start point th_l3 are obtained. When flat_count ≧ 5, th_l3 is set to th_l. flat_cou
When nt <5, the process returns to the start point th_l1 without repeating determination as a valley, and repeats in the same manner. The th_r calculation method is the same as the above th_l calculation method, except that the search is performed from the side with the higher gradation.

The detection of the histogram valley position is performed according to the above-described procedure. After that, the histogram evaluation value calculation means 4 calculates an evaluation value representing the shape of the histogram.

Here, the effectiveness of calculating the evaluation value of the histogram will be described. The density of the characters written on the form such as faint characters and characters with noise varies. As shown in FIG. 5, generally, a histogram of a blurred character has a background mountain, and the tip of the histogram changes smoothly, and the area of the mountain having the highest density tends to be small.
Also, as shown in FIG. 6, the histogram of a character with noise has a background mountain, several peaks ahead of it, and undulations are severe, and the area of the peak with the highest density is large.

In order to determine these character types, an evaluation value th_para described below is effective. When this evaluation value is used, a faint character has a low evaluation value, and a noisy character has a high evaluation value, which is effective for character type discrimination.

As the evaluation value for determining the threshold value,
A weight normalized by multiplying the histogram change amount after th_l by a gradation value is used. That is, if the evaluation value is th_para,

(Equation 1) Is calculated based on

Since the evaluation value for determining the threshold value has been calculated, the binarization threshold value setting means 5 determines the binarization threshold value from the histogram evaluation value.

That is, the threshold value is determined as follows from the value of th_para described above. When th_para <TH_PARA_L th = th_l When TH_PARA_L ≦ th_para ≦ TH_PARA_H A value that sets the area ratio between th = th_l and th_r to 0.5 th_para> When TH_PARA_H th = th_r By extracting characters based on the character, accurate characters can be extracted.

(Second Embodiment) FIG. 7 is a block diagram showing a configuration of a character optimum binarizing apparatus according to a second embodiment of the present invention. In FIG. 7, the character optimum binarization device includes a histogram acquisition unit 1 for acquiring a histogram for each character, a histogram smoothing unit 2 for taking a moving average of the histogram and smoothing the histogram, and detecting a valley position of the histogram. Histogram valley position detection means 3
A histogram evaluation value calculation means 4 for calculating an evaluation value representing the shape of the histogram;
A binarization threshold determining means 5 for setting a binarization threshold;
It comprises a black pixel ratio calculating means 6 for calculating a black pixel ratio with respect to the entire area after binarization, and a binary threshold value correcting means 7 for correcting a binary threshold value from the black pixel ratio.

Next, the operation of the character optimum binarizing device according to the second embodiment of the present invention will be described. The operation up to the point where the binarization threshold value is determined by the binarization threshold value determination means 5 is the same as that of the first embodiment.

However, this method may not be sufficient for faint characters or characters with noise. Therefore, the black pixel ratio calculating means 6 calculates
The ratio of black pixels to the total area after binarization is calculated. Then, the binarization threshold value determined by the binarization threshold value determination unit 5 is corrected by the binarization threshold value correction unit 7 from the ratio of black pixels. This correction operation is repeated until an optimum binarization threshold value is determined, and the character is cut out based on the finally determined correction binarization threshold value. Can also accurately extract characters.

[0025]

As described above, the character optimum binarization device of the present invention calculates the optimum binarization threshold value for each character, so that characters can be cut out more accurately than the fixed threshold value. Has an effect.

Further, since an optimum binarization threshold is calculated for each character property from the histogram distribution, it is possible to binarize a blurred character or a noisy character to cut out an accurate character.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a character optimum binarization device according to a first embodiment of the present invention;

FIG. 2 is a diagram showing a general tendency of a histogram of a normal character;

FIG. 3 is a diagram for explaining a calculation procedure for calculating a histogram valley position;

FIG. 4 is a diagram for explaining a calculation procedure for calculating a histogram valley position;

FIG. 5 is a diagram showing a general tendency of a histogram of a blurred character;

FIG. 6 is a diagram showing a general tendency of a histogram of a noisy character;

FIG. 7 is a block diagram showing a configuration of a character optimum binarization device according to a second embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Histogram acquisition means 2 Histogram smoothing means 3 Histogram valley position detection means 4 Histogram evaluation value calculation means 5 Binary threshold value setting means 6 Black pixel ratio calculation means 7 Binary threshold correction means

Claims (2)

[Claims] A histogram acquisition unit for acquiring a histogram for each character; a histogram smoothing unit for taking a moving average of the histogram to smooth the histogram; a histogram valley position detection unit for detecting a valley position of the histogram; A character optimum binarizing apparatus comprising: a histogram evaluation value calculating unit that calculates an evaluation value representing a shape; and a binarization threshold value determining unit that determines a binarization threshold value from the histogram evaluation value. 2. A histogram acquisition unit for acquiring a histogram for each character, a histogram smoothing unit for taking a moving average of the histogram and smoothing the histogram, a histogram valley position detection unit for detecting a valley position of the histogram, and a histogram valley position detection unit. A histogram evaluation value calculating means for calculating an evaluation value representing a shape, a binarization threshold value setting means for setting a binarization threshold value from the histogram evaluation value,
A character optimum 2 comprising: a black pixel ratio calculating unit that calculates a black pixel ratio with respect to the entire area after binarization; and a binarization threshold value correcting unit that corrects a binarization threshold value from the black pixel ratio. Pricer.