KR19990056813A - Contact Character Separation Method using Individual / Horizontal Ratio - Google Patents

Abstract

The present invention relates to a method of separating contact letters using the horizontal / vertical ratio of individual letters. In the related art, as the documents become more complicated and more documents are created by mixing various fonts, the width of the conventional individual letters and the average width of the space are increased. When using a method of fusing or cutting physically adjacent characters by using this method, there is a problem in that it is not possible to actively cope with changes in various fonts. Therefore, in the present invention, an individual character extraction step of extracting each individual character from a character string, a contact determination step of checking whether or not the individual character extracted in the individual character extraction step is in contact, and in the contact determination step, An individual character recognition step of recognizing individual characters not in contact, an expected separation region extraction step of extracting an expected separation region using the horizontal / vertical ratio of the individual characters recognized in the individual character recognition step, and the expected separation By using the expected separation area extracted in the area extraction step, it operates as a contact character separation step that separates and recognizes the contact character, enabling active contact separation for various fonts, and significantly reducing the overall processing time. To ensure stable recognition even when the document is in poor condition. It is to ensure.

Description

Contact Character Separation Method using Individual / Horizontal Ratio

The present invention relates to a method of separating contact characters using the horizontal / vertical ratio of an individual character for separating when a plurality of characters are contacted during character recognition. In particular, the present invention estimates a font to be recognized and transposes the font. Expect contact separation area by using / vertical ratio, and then extract the feature values from the expected separation area to separate them. It is about a method.

1 is an operation process diagram for a conventional character separation method, as shown in this, string extraction step of separating the string by obtaining the upper and lower positions of the string by comparing the threshold value and the accumulated black pixels in the horizontal direction for the input image In the step of determining the individual character separation candidate for extracting the individual characters using the cumulative number of black pixels in the vertical direction by obtaining a black pixel histogram in the vertical direction for the string obtained in the character string extraction step, and in the individual character separation candidate determination step Character spacing estimating step of obtaining average width of each character and average spacing between each character by using extracted starting point and end point of individual characters, and combining or truncating between characters using interval information obtained in the character spacing estimation step It consists of the integration and separation between characters.

Referring to the prior art made in this way in detail as follows.

When a binary image of a document input through a scanner is input, the horizontal cumulative black pixels of the read document are obtained, and the upper and lower positions of the string are obtained by comparing the obtained value with the threshold value, and the string is separated from the document.

A vertical black pixel histogram is obtained for the separated character strings, and individual characters are extracted using the cumulative number of black pixels.

The average width of each character and the average interval between each character are obtained by using the start and end points of the individual characters extracted above.

In this way, when the average width and the average interval are obtained, the obtained values are fused using the obtained values, and the contact characters are separated in the case of contact characters.

In this case, if the width of adjacent characters is smaller than the average character width and the width of the combined characters is similar to the already obtained width, the two adjacent characters are combined to make one character, and the width of adjacent individual characters is If a certain amount is larger than the average character interval already obtained, it is cut using the interval information.

Characters are separated and recognized in the same way as above.

However, in the prior art as described above, as documents become more complicated and more documents are created by mixing various fonts, there is a method of physically fusing or cutting adjacent characters using the width of the conventional individual characters and the average width of the space. When used, there is a problem that can not actively cope with changes in various fonts.

For example, in a document written in different fonts as in regions 1, 2, and 3 of FIG. 2, there is a problem in that all characters of each region cannot be satisfactorily separated.

Therefore, an object of the present invention for solving the conventional problems as described above is to extract the individual characters from the character string, to obtain the horizontal / vertical ratio of the non-contact font among the extracted characters and to use the expected separation region of the contact character The present invention provides a method of separating contact characters using the horizontal / vertical ratio of an individual character to extract the.

Another object of the present invention is to extract the expected separation region of the contact character by using the font of the recognition object to be separated contact by using the horizontal / vertical ratio of the individual characters to actively perform contact separation for a variety of fonts It provides a method of character separation.

It is still another object of the present invention to provide a method for separating contact characters using the width / length ratio of an individual character to bring about a drastic reduction in overall processing time.

It is still another object of the present invention to provide a method of separating contact characters using horizontal / vertical ratios of individual characters so that stable recognition results can be ensured even in a poor document state.

1 is an operation process diagram for a conventional character separation method.

Figure 2 is a typeface showing a case in which the conventional horizontal / vertical ratio is different from the general font.

Figure 3 is a narrow font type of characters.

4 is a font type diagram showing a form in which the font of area 1 and the font of area 2 have significantly different characteristics.

5 is an operation process diagram for a contact character separation method using the horizontal / vertical ratio of the individual character of the present invention.

The present invention for achieving the above object is an individual character extraction step of extracting each individual character from the character string, the contact determination step of checking whether the extracted individual character is in contact or not, and the contact is not An individual character recognition step of recognizing individual characters, an expected separation region extraction step of extracting an expected separation region using the obtained horizontal / vertical ratio of the individual characters recognized above, and contacting using the expected separation region extracted above. Characterized in that the contact character separation step to recognize after separating the characters.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

? Is an operation process diagram for a contact character separation method using the horizontal / vertical ratio of the individual characters of the present invention, as shown in the individual character extraction step (S11) for extracting each individual character from the character string, Contact character determination step (S12) for checking whether the individual character extracted in the individual character extraction step (S11) has touched or not contacted, and individual character recognition for recognizing the individual character not touched in the contact status determination step (S12) Step (S13), and the expected separation region extraction step (S14) for extracting the expected separation region using the horizontal / vertical ratio of the individual characters recognized in the individual character recognition step (S13) and using this, and extracting the expected separation region The contact character separation step S15 is performed to recognize the contact character after separating the contact character using the expected separation region extracted in step S14.

When described in detail with respect to the operation and effect of the present invention made of each step as follows.

For example, as shown in FIG. 3, when a character width is narrow and long up and down, a character string different from a general font is input, and then individual characters such as Franz are first extracted from the character string (S11).

After extracting the individual characters, it is checked whether the first letter F is in contact with the next letter or not.

As a result of the check, if the letter F is separated without contact, the letter F is recognized (S13).

After the character is recognized, the width and height ratio of the recognized character is obtained and stored. (S13)

Then, it can be seen that the font is quite narrow from the horizontal / vertical ratio obtained above.

Then read the character after F.

If the read character is a contact character such as "ranz", the height of this contact part is obtained.

The expected separation area is extracted by calculating the heights of the contact parts and the horizontal and vertical ratios of the individual fonts.

That is, the contact portion is predicted (S14).

Contact characters are separated using the expected separation region predicted above, and the separated characters are recognized (S15).

In addition, since "Alt" in the next word is also determined to be the same font, the contact is separated by applying the horizontal / vertical ratio values of the individual fonts already drawn.

When the word is changed or the line is changed, the value of the horizontal / vertical ratio is updated every word, and when the first contact character appears, the similarity with the previous font is determined. If not, update again to separate the contact characters.

As shown in FIG. 4, when the characteristics of the font of the area 1 and the font of the area 2 appear quite differently, the value of the horizontal / vertical ratio obtained in the area 1 is newly updated in the area 2.

Accordingly, the present invention extracts the individual characters from the character string, obtains the horizontal / vertical ratio of the non-contact font among the extracted characters, and extracts the expected separation region of the contact character using the extracted characters to separate the contact characters. It is possible to actively carry out contact separation, to significantly reduce the overall processing time, and to ensure stable recognition even in case of poor document status.

Claims (2)

An individual character extraction step of extracting each individual character from a character string, a contact determination step of checking whether or not the individual character extracted in the individual character extraction step is in contact, and an individual not contacted in the contact determination step; In the individual character recognition step of recognizing the character, the expected separation region extraction step of extracting the expected separation region using the horizontal / vertical ratio of the individual characters recognized in the individual character recognition step, and the expected separation region extraction step Contact character separation method using the horizontal / vertical ratio of the individual character, characterized in that consisting of the contact character separation step to recognize after separating the contact character using the extracted expected separation region. The method of claim 1, wherein the horizontal / vertical ratio of the individual characters is updated every word in the expected separation region extraction step.