JPH096915A - Character segmenting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character segmenting method which accuracy segment a character image wherein adjacent characters touch each other into the characters, one by one. SOLUTION: When connection components are extracted from a character image consisting of characters connected by whisker parts like 'milli' in (A), characters are extracted as one character. For the purpose, a peripheral distribution in a line direction is found in advance, standard character height is estimated, and 1/16 of it is found as line width to be disregarded. In the figure, (B) shows a peripheral distribution perpendicular to the line direction and in an area of >=508 in line width to be regarded, characters at character positions 511-517 in (C) and the intermediate points 521-526 of the characters are found from the peripheral distribution to obtain segmentation candidate positions, thereby estimating standard character width.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character slicing method used in a character recognizing device for locating a character area in a character image and converting it into a code number by inputting a sentence written on a paper as an image. .

[0002]

2. Description of the Related Art In recent years, due to the rapid progress of character recognition devices, character areas can be automatically extracted from various document images, and individual characters can be cut out and recognized to automatically create character files. As a result, various methods have been devised for extracting characters.

For example, a commonly used method is to count the peripheral distribution of the extracted character line in the direction perpendicular to the line direction.

For example, the peripheral distribution in the direction perpendicular to the line direction of extracted character lines as shown in FIG. 4A is counted. In this count, if only the presence or absence of a count value, that is, the presence or absence of a character image is shown in the figure, a marginal distribution as shown in FIG. 4B is obtained. , It becomes possible to know the existing position of the character, and the character can be cut out one by one. However,
As shown in FIG. 4C, in the case where the extracted character lines have overlapping peripheral distributions, the overlapping character portions have a large character width such as “Y” and “o”. Therefore, when it is determined that there are a plurality of characters, the character break is estimated based on the character pitch, and the character is forcibly cut out.

[0005]

However, when characters are cut out by such a method, when the target character image is a constant pitch character image as shown in FIG. 4 (A), the characters are accurately cut out. Is possible, but Fig. 4
In the character image as shown in FIG. 5C, the character image as shown in FIG. 5A, and the character image as shown in FIG. FIG.
(C) character image, FIG. 5 (A) character image, FIG.
Each of the character images in (A) is a proportional character whose character pitch is indefinite. Therefore, unlike the conventional method, it is not possible to estimate the character position and the character pitch simply by using the peripheral distribution in the direction perpendicular to the row direction.

A case will be described in which a character as shown in FIG. 4C is cut out by a conventional method.
FIG. 4C is a character image of the same font as FIG. 4 (A) has a constant character pitch, FIG. 4 (C) shows a proportional character with an indeterminate character pitch. Focusing on FIG. 4D showing the peripheral distribution in the direction perpendicular to the row direction of the character image of FIG. 4C, the peripheral distribution of FIG. 4A is almost the same as that of FIG. 4B. Although they are almost the same, the part corresponding to the peripheral distribution of “Yo” is different. In FIG. 4D, the peripheral distribution of “Y” and the peripheral distribution of “o” overlap each other. Therefore, from such a peripheral distribution, "Y" and "o" are judged as one character, or they are separated at a place different from the actual one. Now, if it is decided that two characters are forcibly divided into two characters and the character pitch is focused, the character images are separated at 1201 shown in FIG. 12, and if the shape of the peripheral distribution is focused, 1202 Will separate character images. No matter which position 1201 or 1202 the character image is separated from, the character "Y" or "o" will be cut off in the middle and will be extracted together with another character, making it impossible to accurately extract the character. is there.

Next, from the character image shown in FIG.
A case where a character is cut out by a conventional method will be described. FIG. 5 (A) is a word composed of seven characters "million". When the peripheral distribution of this word in the direction perpendicular to the line direction is counted and the count value is illustrated, as shown in FIG. 5B, “m”, “i”, “l”,
Peripheral distributions 501 and 50 of five characters "l" and "i"
2, 503, 504, and 505 are linked to form a single peripheral distribution block. Therefore, it is impossible to easily estimate character breaks from this marginal distribution.
If the characters are separated by the character pitch, the characters such as "i" and "l" are half the standard character width, so two characters are judged as one character, and It is likely to be cut out. In addition, if you judge by the shape of the marginal distribution and separate it, you may be able to separate the characters such as "i" and "l" well, but there is a high possibility that "m" will be decomposed, and the reliability of character cutting The sex is extremely small.

Also in the case of the character image shown in FIG. 11A, it is impossible to accurately cut out the character only from the shape of the peripheral distribution, as in the case of FIG.

The present invention solves the above-mentioned problems, and an object of the present invention is to provide a method for accurately cutting out each character from a character image in which adjacent characters are in contact with each other. is there.

[0010]

According to a first aspect of the present invention, there is provided a character recognition device for reading a Western character image written on a paper surface or the like by an optical image input means and recognizing a character in the input image data. In the character cutout method, the standard character height of the character is estimated from the marginal distribution in the line direction, the value of the line width to be ignored is determined based on the standard character height, and the marginal distribution in the direction perpendicular to the line direction The standard character width is estimated by taking statistics of a portion where the value of the marginal distribution is larger than the value of the ignored line width.

According to a second aspect of the present invention, in the character cutting method according to the first aspect, the standard character width is estimated by the statistics in a region where the value of the marginal distribution is larger than approximately half of the standard character height. It is characterized by doing.

The invention according to claim 3 is the character cutting method according to claim 2, wherein the distribution of the clusters in the statistics is such that the distance between clusters is less than or equal to a predetermined value proportional to the standard character height. If so, they are regarded as the same lump.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below based on embodiments.

An example of the outline of the whole process of character extraction is a character recognition in which a character image written on a paper surface or the like is read by an optical image input means, a character in the input image data is recognized and replaced with a code number. It is used for the device, and the standard character height of characters is estimated from the peripheral distribution in the line direction, and the line thickness to be ignored is determined. From the peripheral distribution in the direction perpendicular to the line direction, the standard character interval, the word interval and the standard character width are estimated, the words are extracted, the character cutout candidate positions are estimated from the peripheral distribution, and the characters in the extracted words are extracted. Extraction is performed by extracting the contour of the connected component of the character, extracting the character height and the character width, and if the character width exceeds the allowable size from the standard character width, re-extracts within the range of the character extraction candidate position. When the outline is extracted and the character cutout candidate position does not exist,
The range of the character cutout is determined by the peripheral distribution in the direction perpendicular to the line direction, the contour is extracted again, and the character is cut out by extracting only the inside of the area surrounded by the contour.

Regarding the standard character height of a character, paying attention to the shape of the peripheral distribution in the line direction, the width of the portion where the peripheral distribution changes rapidly and becomes large is defined as the standard character height. The minimum line width of the character line can be estimated.

The standard character interval and the word interval can be estimated by taking the statistics of the size of the portion where no character exists in the peripheral distribution in the direction perpendicular to the line direction.

The standard character width can be estimated by taking statistics of the size of the peripheral distribution in the direction perpendicular to the line direction and the value of the peripheral distribution being larger than the minimum line width.

The standard character width can be determined as the maximum value of the cluster closest to the standard character height in the area larger than 75% of the standard character height in the above statistics.

The classification of lumps in the above statistics can be regarded as the same lump if the distance between lumps is less than a certain value proportional to the standard character height.

By comparing the standard character spacing and the word spacing with the marginal distribution in the direction perpendicular to the line direction, the word position can be extracted.

In the peripheral distribution in the direction perpendicular to the row direction, the central portion of each of the peripheral distribution values smaller than the minimum line width can be used as character cutout candidate positions.

When it is determined that the characters are connected characters based on the extracted character width,
The character cutout candidate position can be cut out preferentially.

When the character cut-out position is estimated from the value of the marginal distribution, it is possible to obtain the portion having the smallest value of the marginal distribution near the 1/2 character width and the l character width.

The extraction of the area surrounded by the outline of the character is performed by providing an image area of the same size as the original image, drawing the outline of the character in the image area, and filling the inside of the outline, and then the common part with the original image. By taking, it is possible to extract only the target character.

As shown in the block diagram of FIG. 1, a character recognition device having a character cutting means as a constituent element is a CPU 10
1, image input device 102, recognized character display device 103, R
It is composed of an OM 104 and a RAM 105.

Thereafter, the RAM 1 is used by the image input device 102.
A character cutting method for cutting out each character from the character image read in 05 will be described in detail based on the flowchart shown in FIG.

A character image read into the RAM 105 by the image input device 102 is shown in FIG. First, the marginal distribution in the row direction is counted. Character image Figure 3 (A)
When the peripheral distribution in the row direction is counted, it is 301 in FIG.
The marginal distribution of the shape is obtained. Western characters are shown in Figure 3.
A character with a large character height corresponding to "d" in (A), "e"
It is composed of three types of characters: a character having a small character height corresponding to, and a character having a large character height corresponding to “y” but having a low character position. Therefore, when the peripheral distribution in the row direction of the character image composed of these three types of characters is counted,
Usually, the peripheral distribution has a shape like 301 shown in FIG. The marginal distribution 301 is made up of three regions 311, 312, and 3l3, and depending on the character image,
In some cases, the area 312 or the area 313 does not exist. However, the area 311 always exists, and the width 304 between the upper limit 302 and the lower limit 303 of this area corresponds to the height of a character having a small character height (hereinafter, referred to as “standard character height”). Therefore, it is possible to know the standard character height by paying attention to the shape of the peripheral distribution in the row direction.

In the case of printed Western characters, the standard character height and the character line thickness usually have a relationship of standard character height: character line thickness ≧ 16: 1. Therefore, when paying attention to the vertical marginal distribution of the character image, the portion where the marginal distribution value is lower than 1/16 of the standard character height is the whiskers of the character,
Alternatively, it can be determined that the characters are in contact with each other. Here, the value of 1/16 of the standard character height, which is the criterion for determination, is obtained as the line width to be ignored.

In the next step, the standard character spacing, the word spacing, and the standard character width are obtained from the vertical distribution of the character lines. First, when the image is projected in the direction perpendicular to the character line, the data as to whether or not the character image exists is obtained. When the character image is shown in FIG. 4A, the data as to whether or not the character image exists, that is, the data projected in the direction perpendicular to the character image is shown in FIG. 4B. An area 401 is a portion where a character image exists, and an area 402 is a portion where no character image exists, that is, a portion corresponding to a character interval. The statistics of the portion 402 in which no character exists are as shown in FIG. 6 (A). Similarly, the projected data when the character image is shown in FIG. 4C. The portion 402 in FIG. 4D where the character image does not exist.
6B is obtained by taking the statistics of. Figure 6
The data shown by the solid lines in (A) and FIG.
The data in (B) and FIG. 4 (D) are shown respectively, but generally, the histogram of the character spacing tends to be indicated by the dotted line shown in FIGS. 6 (A) and 6 (B). Each of these two histograms can be divided into two chunks. One of these clusters can be determined as a cluster of character intervals, and the other can be determined as a cluster of word intervals. Therefore, it is possible to estimate the standard character spacing and the word spacing by the character spacing statistics. Generally, the histogram of the character spacing becomes a histogram as shown in FIG. 6, but sometimes a histogram as shown in FIG. 8 is obtained. The Histogram has many data chunks. If there are many lumps in this way, the lumps are classified by the following method. Printing Western characters are usually printed in regular rows.

Therefore, although the character spacing should be almost the same, sometimes the character spacing varies depending on the character shape. However, the gap is not more than a certain value proportional to the standard character height. Therefore, 1/16 of the standard character height is set as the threshold value 805 for the distance between clusters. Threshold value 805 and lumps 811, 81
2,813,814,815 spacing 801,802,8
By comparing 03, 804, the lumps 811 and 8l2 are the same lumps, and the lumps 813, 814 and 8l5.
Can be judged to be the same block, and the standard character spacing and word spacing can be estimated.

Next, the standard character width of the character image is estimated.
If the standard character width can be estimated, it can be used as a clue for determining whether a character is a single character or a concatenated character when cutting out a character, and for estimating a cutout candidate position. The estimation of the standard character width takes statistics of the character area similarly to the estimation of the standard character spacing and the word spacing. If the character images are as shown in FIGS. 4A and 4C, the character width statistics are as shown by the solid lines in FIGS. 7A and 7B, which are generally indicated by dotted lines. Get a day evening. Since the data shown in FIG. 7A has only one lump, the maximum value of this lump can be set as the standard character width. In the data shown in FIG. 7B, there are two clusters, a cluster 702 and a cluster 703. Normally, the standard character width of printed Western characters is very close to the standard character height.
The maximum value of 02 can be set as the standard character width. In addition, as in the case of estimating the character spacing, when the Histogram sometimes consists of many lumps, in FIG. 9, the distances 901, 902 and 903 between the lumps and the lumps and the threshold 90
By comparing with 4, the mass 9ll, the mass 9
12 and the lump 913 can be determined to be the same lump. Furthermore, even if many lumps are classified by the threshold value between lumps, there may be three lumps as shown in FIG. The reason why the statistical result is as shown in FIG. 10 is that characters such as “i” and “l” are included in the character image.

Then, a mass 100 of "i", "l", etc.
In order not to judge the maximum value of 1 as the standard character width, the maximum value of the cluster is searched for in an area larger than the position 1004 which is 75% of the standard character width. By doing so, it is possible to accurately estimate the standard character width by excluding clusters of "i" and "l" having a width that is about half the standard character width. As described above, the character width of a normal character image can be estimated, but the character image has a character image as shown in FIG. 5 (A). This character image is "m", "i",
It is composed of seven characters, "l", "l", "i", "o", and "n", of which "m", "i",
The five letters "l", "l", and "i" are in contact. Therefore, even if the peripheral distribution in the direction perpendicular to the line direction is counted to estimate the standard character width, the standard character width cannot be estimated.

Therefore, in the present invention, in the printed Western characters, the character line width has a thickness larger than a certain value in proportion to the standard character height, and the whisker part at the bottom shown in FIG.
Paying attention to the line being thinner than a certain value, after counting the peripheral distribution chart 5 (B) in the direction perpendicular to the line direction, the line width to be ignored is determined with a certain value 508 proportional to the standard character height. , The process of rounding down the marginal distribution is established. By this process, only the portion where the value of the peripheral distribution is larger than a certain value 508 is determined as the character region, and the character region is displayed, as shown in FIG. 5C. 511, 512, 513,
514, 515, 5l6 and 5l7 are respectively "m",
The character positions and character widths of "i", "l", "l", "i", "o", and "n" are shown. Although the seven data 511 to 517 are slightly smaller than the actual character width, they always include a portion of information required as a character, and these areas can be determined to be character areas. Therefore, the standard character width can be estimated by taking statistics of the character area of this data FIG. 5 (C).

Next, words are extracted from the character image.
To extract words, the standard character spacing and word spacing are compared with the vertical marginal distribution of character lines. When the character image is as shown in FIG. 4 (A), the marginal distribution is as shown in FIG. 4 (B), which makes it possible to find the word interval from the size of the portion where the character does not exist. It becomes possible to extract words.

Next, the extraction of each character from the extracted word is performed by extracting the connected component, but this also has a problem. Now, suppose that the extracted word is a character image as shown in FIG. In this case, when the connected components are extracted, the five characters "milli" are extracted as one character. This extracted character must be judged as a concatenated character based on its character width, and must be forcibly separated in the middle. The cut-off position is usually performed at a small peripheral distribution value near the standard character width. However, Figure 5
In the case of (A), characters such as "m", "i", and "l" are included. In such characters, since the characters "i" and "l" have a width half the standard character width, two characters in a row have a width of one character and two characters are cut out together. In the case of “m”, since the value of the marginal distribution is once small in the vicinity of the standard character width, it is judged to be a character break, and there is a high possibility that the character will be separated in the middle.

Therefore, in the marginal distribution of FIG. 5B, a region having a certain value 508 or more in proportion to the standard character height, the characters at the character positions 511 to 51 in FIG. 5C and the intermediate points 521 to 52 of the characters.
6 is obtained and set as a cutout candidate position. The cut-out candidate positions 521 to 526 represent boundaries between characters as is clear from FIG. Therefore, when cutting out a character, if the size of the character is larger than the allowable size from the standard character width, it is possible to cut out the character by using such a cutting position, and cutting out the character at the wrong position. Is extremely diminished.

Next, each character is extracted. Extraction of characters is performed by extracting connected components. Most of Western characters have one connected component, and most of the characters can be judged only by the main connected component, even if the character does not have one connected component. Therefore, this method is effective. The extraction of the connected component first extracts the contour of the connected component. By extracting the contour, information on the position and size of the character can be obtained. When the contour is extracted, the already-obtained standard character width is compared with the extracted character width.

If the extracted character width is larger than the standard character width, the extracted character is judged as one character and one character is extracted. When the extracted character width is larger than the standard character width, it is first checked whether or not a cutout candidate position exists in the extraction area. If the cutout candidate position exists, it is most appropriate to cut out the character at that position, and therefore the outline of the character of the connected component is extracted again within the range of that position.

Further, when the character image is a character image as shown in FIG. 11A (in this character image, the middle question point in the area where the peripheral distribution is smaller than the constant size 1110 with respect to the standard character height is There are only 5 positions 1121 to 1125. However, there are 9 characters.),
When the contour of the character image 1152 is extracted, it is determined that the character width is unacceptable than the standard character width. Therefore, the existence of the extraction candidate position is confirmed. However, since the marginal distribution 1142 of the character image 1152 does not have a value smaller than the ignored line width 1110 in the middle of the character, there is no cutout candidate position. In this case, the marginal distribution l14 of the extracted characters
Pay attention to the value near the standard character width of 2. Thereby, it is possible to search for the contact portion 1131 of the characters “a” and “r”, and it is possible to cut out each character. Similarly, the clipping is repeated, and in FIG.
In addition to the five cut-out candidate positions of 1-1125, 11
It is possible to accurately extract nine characters by extracting three cut-out positions 31 to 1133.

Further, after the contour of the connected component of the character is extracted, if the extracted character width is one character width, the character is extracted. Therefore, for example, a character image as shown in FIG. In, you can extract characters without any problems. This character image is called "Y", "o", "u".
It is one contactless character. However, the peripheral distribution map 1
In FIG. 2B, the peripheral distributions of “Y” and “o” are overlapped with each other to form one peripheral distribution 1203. If the character is forcibly separated from the shape of the peripheral distribution 1203 as in the conventional case, the character is separated at the position of 1202. Also, if an attempt is made to separate depending on the size of the peripheral distribution, the separation will occur at the position of l201. Either position will result in the character being cut off at an improper position, but this method does not cause any problems.

After the contour of the connected component of the character is extracted with an appropriate size, the connected component of the character is extracted. Another area of the same size as the character image is provided. FIG.
When extracting the character “P” from the character image 1301 of FIG. 3A, first, the outline of the character “P” of FIG. 13A is drawn in another area 1302. Next, the area surrounded by this contour is painted out to obtain FIG. After that, by extracting a common part between the original image FIG. 13A and the extracted character region image FIG.
The image shown in FIG. 3 (C) is obtained. As can be seen from this image FIG. 13C, it is possible to cleanly extract only the image of “P” from the images of “P” and “e” that are positionally overlapped.

As described above, according to the present invention, the standard character height and the line width to be ignored are obtained from the peripheral distribution in the line direction, and the standard character spacing, the word spacing, and the standard character width are determined with reference to these values. , Accurate standard character spacing and word spacing, standard character width can be obtained. As a result, accurate word segmentation is possible. Further, when cutting out characters, the characters are cut out with reference to the cutting-out candidate position and the peripheral distribution while comparing with the standard character width, so that the characters can be cut out at accurate positions. Further, since character extraction extracts an image of a region surrounded by the contours of the connected components of the character, it is possible to accurately extract only one character from a character image that overlaps in the peripheral distribution and the cutout position is unknown. became.

As described above, according to the present invention, it has become possible to more accurately cut out each character, which has been difficult until now. Therefore, the reliability of the character recognition device using this method as a constituent element can be improved. It has become possible to greatly improve.

Further, since the present invention uses the standard character height and the connected component extraction of characters, it is particularly suitable for printing Western characters.

[0045]

As described above, according to the present invention, the marginal distribution in the row direction is counted, the standard character height is estimated from the shape of the marginal distribution, and the line width to be ignored is determined to determine the line width. In the marginal distribution in the direction perpendicular to the direction, the characters connected by the whiskers are separated and the standard character width is estimated by taking statistics of the portion where the value of the marginal distribution is larger than the value of the line width to be ignored. be able to.

Further, according to the statistics, in the area where the value of the marginal distribution is larger than approximately half of the standard character height,
By estimating the standard character width, it is possible to separate characters at more accurate positions.

Further, regarding the distribution of the lumps in the above statistics, if the distance between the lumps is equal to or less than a predetermined value proportional to the standard character height, it is regarded as the same lump, and more reliable cutting can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram of a character recognition device using a character cutout unit of the present invention.

FIG. 2 shows a flow chart of a character cutting means of the present invention.

FIGS. 3A and 3B are diagrams for explaining a standard character height extracting method according to the present invention.

FIGS. 4A to 4D are diagrams for explaining how to estimate the standard character spacing, the word spacing, the standard character width, and the cutout candidate position according to the present invention.

5 (A) to 5 (C) are diagrams for explaining how to estimate the standard character spacing, the word spacing, the standard character width, and the clipping candidate position according to the present invention.

6 (A) and 6 (B) are diagrams for explaining how the standard character spacing and word spacing, standard character width, and cutout candidate position are estimated according to the present invention.

7 (A) and 7 (B) are diagrams for explaining how to estimate the standard character spacing, the word spacing, the standard character width, and the cutout candidate position according to the present invention.

FIG. 8 is a diagram illustrating a state of estimating a standard character interval and a word interval, a standard character width, and a cutout candidate position according to the present invention.

FIG. 9 is a diagram illustrating a state of estimating a standard character interval and a word interval, a standard character width, and a cutout candidate position according to the present invention.

FIG. 10 is a diagram illustrating a state of estimating a standard character interval and a word interval, a standard character width, and a cutout candidate position according to the present invention.

11 (A) and 11 (B) are diagrams for explaining the state of character extraction according to the present invention.

12 (A) and 12 (B) are diagrams for explaining the state of character extraction according to the present invention.

13 (A) to (C) are diagrams for explaining the state of character extraction according to the present invention.

[Explanation of symbols]

101 ... CPU, 102 ... Image human power device, 103 ... Recognition character display device, 104 ... ROM, 105 ... RAM, 30
1 ... Line direction peripheral distribution, 304 ... Standard character height, 401 ... Character area, 402 ... Character spacing, 50l-507 ... Character peripheral distribution, 51-517 ... Character area, 521-526 ... Character cutout candidate position, 508 ... Lowest Character line width, 70l ...
Standard character height, 80l to 804 ... lump interval, 805 ...
Lump interval threshold, 90l to 903 ... lump interval, 904 ... lump interval threshold, 1004 ... 75% of standard character height, 1l10 ... minimum character line width, 1131-
1133 ... Character cutting position, 120l to 1202 ... Character cutting position.

Claims (3)

[Claims] 1. A character slicing method of a character recognizing device for reading a Western character image written on a paper surface or the like by an optical image inputting means and recognizing a character in input image data. While estimating the standard character height of, the value of the line width to be ignored is determined based on the standard character height, and in the marginal distribution in the direction perpendicular to the line direction, the value of the marginal distribution is A character segmentation method characterized by estimating a standard character width by taking statistics of a portion larger than the value. 2. The character cutting method according to claim 1, wherein the standard character width is estimated in the region where the value of the marginal distribution is larger than approximately half of the standard character height by the statistics. 3. The lump distribution in the statistics is considered to be the same lump if the distance between the lumps is equal to or less than a predetermined value proportional to the standard character height. Character cutting method.