JPH04264993A - One character segmenting method - Google Patents

Abstract

PURPOSE:To obtain a character separate position deciding method to deal with connected characters with different character width and character intervals in a document and operated at high processing speed by using simultaneously the peripheral distribution and line density of a part of the connected characters. CONSTITUTION:The peripheral distribution and the line density are found from a character image, and are stored in the storage area of a RAM 13, respectively. Thence, the product of the peripheral distribution stored in a peripheral distribution storage area 13e and the line density stored in a line density storage area 13f is found, and a result is stored in a product storage area, however, the product can be found by multiplying the value of the peripheral distribution by that of the line density at the same scanning position. The separate position of the connected characters can be decided by using a numeric value stored in the product storage area 13g. A deciding method is to set the scanning position where the minimum product value can be obtained as the separate position of the character.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a single character extraction method for separating connected characters in image information of a document in a character recognition device.

0002

2. Description of the Related Art In character recognition devices, black pixels may be connected between characters in the document itself, or because the resolution of the scanner is insufficient. Characters concatenated in this way must be forcibly separated.

Therefore, when characters are connected, conventionally, in order to separate them into single characters, for example, in the case of a horizontally written document, the vertical projection (projection on the horizontal axis) of the connected characters is calculated, and the projected value is specified. A method of forcibly separating characters that is less than or equal to the value of , or finding the vertical projection of the entire string containing connected characters, and determining the separation position of the connected characters from the periodicity of the position where the projected value is zero. did.

0004

[Problems to be Solved by the Invention] However, according to the method for separating connected characters described above, there are multiple parts where the projected value is less than a specific value, and the separation position cannot be determined, or the character width or Due to unequal letter spacing,
There was no periodicity in the position where the projected value was zero, and the separation position of connected characters was sometimes incorrect.

[0005]The present invention is intended to solve these problems, and its purpose is to solve the problems of documents with unequal character widths and character spacing by using the peripheral distribution and line density of connected character parts together. The object of the present invention is to provide a method that can deal with connected characters in the text, has a high processing speed, and can determine character separation positions because it processes only the range in which character separation positions are to be searched.

[0006]

[Means for Solving the Problems] In the character separation method according to the present invention, when there are connected characters in each line to be processed, a range is set to search for the character separation position, and the connected characters are searched within the set range. Find the marginal distribution and line density from the image,
The method is characterized in that the product of the obtained marginal distribution and line density at each scanning position is obtained, and the separation position of connected characters is determined from the obtained product.

[0007]

EXAMPLES (Example 1) Examples of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing the equipment configuration necessary for implementing the present invention. 10 is a scanner for inputting document images; 11 is a CPU for executing processing; 12
13 is a ROM that stores programs for each process, and RA is for storing data related to images and processing, and processing results.
It is M. Although the program for each process is stored in the ROM 12 in FIG. 1, the program for each process may be loaded in the RAM instead of the ROM 12 before starting the process.

[0009] The processing contents of the above-mentioned apparatus configuration example will be explained.

The image of the document inputted by the scanner 10 is stored in an image storage area 13a in the RAM 13. Note that the image of the separate book handled here is a binary image, that is, one composed of 0 (white pixel) and 1 (black pixel). The CUP 11 cuts out lines from the document image stored in the image storage area 13a according to the line cutting program 12a, and converts the image of the cut lines into RA.
It is stored in the row image storage area 13b in M13. This line cutting is performed by, for example, measuring the horizontal projection (projection on the vertical axis) of a document (in this case, a horizontally written document).
This is done by cutting out the range where the projection value exceeds a specific numerical value as a row range. Note that the line extraction method itself may be arbitrary, such as a method of dividing a line into a plurality of blocks and performing line extraction by horizontal projection for each block.

From the cut out line image, the CPU 11 cuts out characters according to the character cutting program 12b, and stores the cut out character images in the character image storage area 13c in the RAM 13. At the same time, the character width is determined and stored in the character width storage area 13d in the RAM 13. This method of cutting out characters is, for example, by measuring the vertical projection (projection on the horizontal axis) of a character example (in this case, a horizontally written document), and then cutting out the part where the projection value is greater than zero as the range of characters. etc. The character width is determined by using the average value of the range of cut out characters. Further, the character width may be given in advance and stored in the character width storage area 13d in the RAM 13. Note that the method of cutting out characters and calculating the character width is arbitrary.

When cutting out a single character, there may be a connection of black pixels between characters in the document itself, or there may be a connection of black pixels between characters due to insufficient resolution of the scanner 10. Therefore, the CPU 11 checks whether any characters are registered in the character image storage area 13c while being connected to each other according to the character separation program 12c, and determines that the characters are registered in the character image storage area 13c while being connected to each other. For character images that have been registered, the characters are separated and re-registered. The flow of processing of the character separation program 12c will be explained with reference to FIG.

[0013] The CPU 11 executes the character cutting program 12.
According to b, the width of the registered character image is compared with the obtained character width multiplied by α (step 202). Here α
is a numerical value indicating the width when characters are connected to each other; for example, α=2.

As a result of the comparison, if the width of the character image is equal to or larger than the character width multiplied by α, it is determined that two or more characters are connected, and a range to search for character separation positions is set ( Step 203). The range in which the character separation position is searched is centered on the position leading by the character width from the beginning of the character image, and is set to the range β times the character width to the left and right. Here, β is a numerical value used when setting the range to search for character separation positions, and is set to β=0.25, for example. This is because when printed characters have the same number of points, the virtual bodies of the characters (Fig. 3, 301) are the same, but the face (Fig. 3, 302) shown by the rectangle circumscribing the characters differs depending on the character, and the virtual body of the characters (Fig. 3, 301) is the same. The width of the font is on average 5% to 2% of the width of the body.
This value was set based on the fact that it is 5% smaller.

The peripheral distribution is determined from the character image within the range set in step 203, and is stored in the peripheral distribution storage area 13e in the RAM 13 (step 204). In the case of a horizontally written document, the peripheral distribution is obtained by scanning in the vertical direction and counting pixels with a pixel value of 1 (black pixels) at each scanning position.

Next, within the range set in step 203,
The line density is determined from the character image and stored in each line density area 13f in the RAM 13 (step 205). In the case of a horizontally written document, the line density is determined by scanning in the vertical direction and counting the locations where the pixel value is reversed from 0 to 1 at each scanning position.

The product of the marginal distribution stored in the marginal distribution storage area 13e and the linear density stored in the linear density storage area 13f is calculated. It is stored in the product storage area 13g in the RAM (step 206). The product is obtained by multiplying the value of the marginal distribution and the value of the linear density at the same scanning position.

Using the numerical value stored in the product storage area 13g, the separation position of the connected characters is determined (step 2
07). The position is determined by determining the scanning position where the product value is the minimum as the separation position of the connected characters. If there are multiple minimum values within the range, from the beginning of the character image,
The scanning position of the minimum value closest to the position advanced by the character width is set as the character separation position.

The processed character image is deleted from the character image storage area 13c, separated at the character separation position determined in step 205, and re-registered in the character image storage area 13c (step 208).
.

Next, it is determined whether a character image satisfying the conditions used in step 202 remains in the character image storage area 13c, and if there is no character image, the CPU 1
1 ends the processing of the character separation program 12c. If it exists, similar processing is continued (step 201).

FIG. 4 shows that the CPU 11 runs the character separation program 1.
An example of determining a separation position from connected character images according to 2c is shown below. FIG. 4(a) shows the character image storage area 13.
This is an example of a character image registered while being connected to c. You can see that the characters ``smoke'' and ``road'' are connected. Moreover, the range r shown in FIG. 4(a) is
This is the range to search for the character separation position set in step 03. FIG. 4(b) shows the marginal distribution obtained from FIG. 4(a) within the set range. The minimum value of the marginal distribution is point p11
2 points can be seen from to p12. FIG. 4(c) shows the linear density determined from FIG. 4(a) within the set range. The minimum value of the linear density is found at two points from point p21 to p22. Figure 4(d) is similar to Figure 4(b) and Figure 4(c).
This is the product obtained from . The minimum value is only at point p31,
The character separation position is determined to be point p31.

As shown in FIG. 4, there are a plurality of scanning positions where the minimum value is obtained based only on the marginal distribution and linear density, and it is difficult to determine the separation position of connected characters. However, when characters are connected, there is a high probability that there is only one connected part, so by multiplying the marginal distribution by the linear density and weighting it, it becomes easier to determine the position where the characters are connected.

[0023]

Effects of the Invention As explained above, according to the present invention, character separation positions are determined by using both marginal distribution and linear density in separating connected characters, thereby improving character strings with irregular character intervals. It is possible to separate connected characters inside, narrow down candidates for character separation positions, and set the search range for character separation positions, so processing speed is fast and separation positions can be determined. have an effect.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of the configuration of a device necessary for implementing the present invention.

FIG. 2 is a flowchart showing the flow of a single character extraction process according to the present invention.

FIG. 3 is a diagram for explaining the size of characters.

FIG. 4 is a diagram showing an example of determining a separation position from connected characters by the process shown in FIG. 2;

[Explanation of symbols]

10 Scanner 11 CPU 12 ROM 13 RAM

Claims (1)

[Claims] 1. In each line of a document to be processed, if there are connected characters in the line, a range is set to search for character separation positions, and within the set range, peripheral distribution is determined from the connected character images. A method for cutting out one character, characterized in that the linear density is determined, the product of the peripheral distribution and the linear density at each corresponding scanning position is determined, and the scanning position where the product has a minimum value is set as the separation position of the connected characters. .