JP3833867B2 - Concatenated character separation device, concatenated character separation method, and computer-readable recording medium recording a program for causing a computer to execute the method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a connected character separation technique for determining a separation position of a connected character in which a plurality of characters are connected in a document image, and separating the connected character based on the determined separation position, and in particular, without setting a character separation range. The present invention relates to a connected character separating apparatus, a connected character separating method, and a computer-readable recording medium on which a program for causing a computer to execute the method is recorded.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, in a character recognition device, in order to improve a character recognition rate, a connected character (a black pixel forming one character and a black pixel forming another character are connected in a document image read by a scanner or the like). A connected character separation technique for separating each character) is known.
[0003]
For example, in Japanese Patent Laid-Open No. 4-248688, the characteristic value of each scanning position is calculated based on the peripheral distribution and the line density obtained from the connected characters, and the characteristic value is minimized within a preset scanning position range. A technique is disclosed in which a scanning position is determined as a separation position, and concatenated characters are separated based on the separation position.
[0004]
Specifically, for a horizontally written document image, first, a connected character is scanned in the vertical direction, and the peripheral distribution is obtained by counting the number of black pixels for each scanning position, and from the white pixels for each scanning position. The line density is obtained by counting the number of inversions to black pixels. Next, the characteristic value of each scanning position is calculated by integrating the peripheral distribution and the line density for each scanning position. Then, a scanning position range of 0.75 to 1.25 times the character width (average character width obtained separately) from the beginning of the connected character image is set as the character separation range, and the characteristic value is minimum within the range. Is determined as the separation position.
[0005]
FIG. 11 is an explanatory diagram showing an example of determining a separation position of connected characters in this prior art. As shown in the figure, a characteristic value for each scanning position is calculated for the connected character “Tatsuno”. If the range B is set as the character separation range from the character width, the connected character “Tatsuno” can be separated into “dragon” and “field”.
[0006]
[Problems to be solved by the invention]
However, the character separation range is obtained on the basis of the separately obtained character width, the character width is not uniform for handwritten characters, etc., and when two or more characters are connected, the character The width cannot be obtained properly. As a result, it is difficult to set the character separation range accurately. In FIG. 11, the separation position is not set to the range B, and may be set to the range A, for example, so that the characteristic value is minimized. The scanning position is Y, and there is a problem that the connected characters cannot be separated accurately.
In this prior art, it is conceivable that the minimum value of the characteristic value in the entire range of the connected character is set as the separation position. In this case, however, separation is performed at the position indicated by Y as shown in FIG. As a result, there is a problem that the connected characters cannot be accurately separated. In this prior art, since the characteristic value is calculated using the line density obtained by counting only the complexity of the character in one direction (scanning position direction) of the connected character, this characteristic value indicates the complexity of the character. This is because it is insufficient to represent.
[0007]
Therefore, in order to solve the above-described problems caused by the prior art, the present invention provides a connected character separating apparatus, a connected character separating method, and a connected character separating method that can separate connected characters directly and accurately without setting a character separation range. An object of the present invention is to provide a computer-readable recording medium on which a program for causing a computer to execute the method is recorded.
[0008]
[Means for Solving the Problems]
In order to solve the above-described problems and achieve the object, a connected character separation apparatus according to claim 1 detects a connected character in which a plurality of characters are connected in a document image, and scans the detected connected character. In the connected character separation device that determines the separation position of the connected character and separates the connected character based on the determined separation position, the peripheral distribution calculation that calculates the peripheral distribution by counting the pixel of the character for each scanning position of the connected character A circumscribing width calculating unit that calculates a circumscribing width by calculating an interval between character pixels at both ends for each scanning position of the connected character; and a character adjacent to each pixel in the XY direction for each scanning position of the connecting character. A local line density calculating unit that calculates a local line density based on a line edge interval; and a characteristic value calculating unit that calculates the characteristic value by adding the peripheral distribution, circumscribed width, and local line density for each scanning position. And the special Characterized by comprising a separation position determining means for determining the separating position the scanning position to take the lowest minimum value among the characteristic values calculated by the value calculating means.
[0009]
According to a second aspect of the present invention, there is provided the connected character separation device according to the first aspect of the invention, wherein the local line density calculating means is adjacent to the interval between the character line edges and the Y direction adjacent to each other in the X direction. The character line edge interval is calculated as a line interval, the larger of the calculated line intervals is calculated as a pixel line density value, and the calculated pixel line density value is added for each scanning position. A local linear density is calculated.
[0010]
According to a third aspect of the present invention, there is provided a connected character separation method that detects a connected character in which a plurality of characters are connected in a document image, scans the detected connected character, determines a separation position of the connected character, and In a connected character separation method for separating connected characters based on a determined separation position, a peripheral distribution calculating step for calculating a peripheral distribution by counting character pixels for each connected character scanning position, and for each connected character scanning position A circumscribing width calculation step of calculating a circumscribing width by calculating a spacing between character pixels at both ends, and a local line density based on a spacing between character line edges adjacent to each pixel in the XY direction for each scanning position of the connected character Calculated by a local line density calculating step, a characteristic value calculating step of calculating the characteristic value by adding the peripheral distribution, circumscribed width and local linear density for each scanning position, and the characteristic value calculating step Characteristic value Minimum separation position determining step of determining the scanning position as a separated position which takes a minimum value in out, characterized in that it contains.
[0011]
According to a fourth aspect of the invention, there is provided the connected character separation method according to the third aspect of the invention, wherein the local line density calculation step is adjacent to each pixel in the X direction, adjacent to the character line edge interval and in the Y direction. The character line edge interval is calculated as a line interval, the larger of the calculated line intervals is calculated as a pixel line density value, and the calculated pixel line density value is added for each scanning position. A local linear density is calculated.
[0012]
Further, the recording medium according to the invention of claim 5 records the program for causing the computer to execute the method described in claim 3 or 4 so that the program can be read by a machine. The operation of 4 can be executed by a computer.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Exemplary embodiments of a concatenated character separation device, a concatenated character separation method, and a computer-readable recording medium storing a program for causing a computer to execute the method will be described in detail below with reference to the accompanying drawings. explain. In the present embodiment, the case where the present invention is applied to a character recognition device is shown.
[0014]
First, the configuration of the character recognition device used in this embodiment will be described. FIG. 1 is a block diagram showing a configuration of a character recognition device used in the present embodiment. As shown in the figure, the character recognition device 1 used in the present embodiment includes an image input unit 2, a character string cutout unit 3, a character block cutout unit 4, a character block storage unit 5, a connected character. The block separation unit 6, the character candidate segment creation unit 7, the character recognition unit 8, and the knowledge processing unit 9 are configured. The connected character block separation unit 6 corresponds to the connected character separation device according to claim 1 or 2.
[0015]
In FIG. 1, an image input unit 2 is a scanner that reads a document image composed of handwritten characters written on paper or the like. In the present embodiment, a case where a horizontally written document is input as a binary document image (an image composed of white pixels and black pixels) in the image input unit 2 will be described.
[0016]
The character string cutout unit 3 is a processing unit that cuts out a character string from a horizontally written document image input by the image input unit 2. For example, for a horizontally written document image, a method of measuring a horizontal projection (projection with respect to a vertical axis) and cutting out a range in which the projection value exceeds a specific numerical value can be applied.
[0017]
The character block cutout unit 4 is a processing unit that cuts out a character block from the character string cut out by the character string cutout unit 3. For example, a method of measuring a vertical projection (projection with respect to a horizontal axis) and cutting out a range where the projection value exceeds zero as a character block can be applied to a horizontally written character string.
[0018]
The character block storage unit 5 is a memory that stores the character blocks cut out by the character block cutout unit 4 and the character blocks separated by the connected character block separation unit 6.
[0019]
The concatenated character block separation unit 6 detects a concatenated character block (a character block in which two or more characters are concatenated) from the character blocks stored in the character block storage unit 5, determines a separation position of the concatenated character block, A processing unit that separates the connected character blocks at the determined separation position.
[0020]
Here, as a method of detecting a concatenated character block from character blocks stored in the character block storage unit 5, for example, an aspect ratio of the character block (width of the character block / width of the character block) is measured, and the aspect ratio is measured. For example, a method of detecting a character block having a specific value (for example, 2) as a concatenated character block can be applied.
[0021]
In addition, the determination of the separation position of the connected character block is performed by calculating, for each scanning position, a characteristic value relating to black pixels in the connected character block (a value reflecting the complexity of characters in the vertical and horizontal directions in the connected character block). Then, the scanning position having the minimum minimum value among the calculated characteristic values is determined as the separation position.
[0022]
As a method of calculating the characteristic value, for example, there is a method of adding a peripheral distribution, a circumscribed width, and a local line density to be described later for each scanning position. Note that the characteristic value is the local linear density, the characteristic value is obtained by adding or integrating the local linear density and the peripheral distribution, and the characteristic value is obtained by adding or integrating the local linear density and the circumscribed width. It is also possible.
[0023]
Note that the character block separated by the connected character block separation unit 6 is re-registered in the character block storage unit 5, and if the re-registered character block is still a connected character block, the connected character block separation unit 6 further To be separated.
[0024]
The character candidate segment creation unit 7 is a processing unit that creates a character candidate segment by synthesizing each character block stored in the character block storage unit 5. For example, if Kanji characters that consist of Hen and Tsukuri are stored as separate character blocks, a character candidate segment is created by synthesizing this character block. To do.
[0025]
The character recognition unit 8 is a processing unit that performs character recognition on the character candidate segment created by the character candidate segment creation unit 7, and the knowledge processing unit 9 uses the recognition result of the character recognition unit 8 to perform character candidates. It is a processing unit that collates a combination of segments with a word registered in advance in the word dictionary.
[0026]
Next, an example of processing of each unit in the character recognition device 1 used in the present embodiment will be described. FIG. 2 is an explanatory diagram showing an example of character block extraction in the character block extraction unit 4 shown in FIG. 1, and FIG. 3 shows an example of character block storage in the character block storage unit 5 shown in FIG. FIG. 4 is an explanatory diagram showing an example of creating character candidate segments in the character candidate segment creating unit 7 shown in FIG.
[0027]
When the character string cut-out unit 3 cuts out the character string of “Katsubuki Hondacho, Tatsuno-shi, Hyogo” shown in FIG. 2A, the character block cut-out unit 4 sets the vertical projection value to zero. As a result of cutting out the exceeding range as a character block, as shown in the figure (b), "Hyo", "Kou-ken", "Tatsuno", "City", "Honor", "Ta", "Town", "Fragment" And each character block of “Blow” is cut out.
[0028]
In the character block storage unit 5, the concatenated character block separation unit 6 separates the concatenated character blocks “Ku Prefecture” and “Tatsuno” into character blocks “Ku”, “Prefecture”, “Ryu”, and “Field”. As a result, as shown in FIG. 3, “Soldier”, “Warehouse”, “Prefecture”, “Ryu”, “Field”, “City”, “Honor”, “Ta”, “Town”, “Fragment” And each character block of “Blow” is stored.
[0029]
In the character candidate segment creation unit 7, as a result of synthesizing each character block stored in the character block storage unit 5 to create a character candidate segment, as shown in FIG. ”,“ Dragon ”,“ field ”,“ city ”,“ honor ”,“ field ”,“ town ”,“ strip ”,“ buki ”,“ Hyogo ”,“ Kyo prefecture ”,“ Tatsuno ”,“ field city ” And each character candidate segment of “Kabuki” is created.
[0030]
The reason why two characters are also used as character candidate segments is a result of considering the case in which bias and 旁 are stored as separate character blocks for kanji characters consisting of bias and 旁. In other words, when a certain condition is satisfied, as a result of combining two blocks, things such as “Hyogo” and “Hyogo Prefecture” are also character candidate segments.
[0031]
Next, the configuration of the connected character block separation unit 6 shown in FIG. 1 will be described. FIG. 5 is a block diagram showing the configuration of the connected character block separation unit 6 shown in FIG. As shown in the figure, the connected character block separation unit 6 includes a separation control unit 10, a peripheral distribution calculation unit 11, a circumscribed width calculation unit 12, a local line density calculation unit 13, and a characteristic value calculation unit 14. And is configured.
[0032]
The separation control unit 10 corresponds to the separation position determination unit described in claim 1, the peripheral distribution calculation unit 11 corresponds to the peripheral distribution calculation unit described in claim 1, and the circumscribed width calculation unit 12 includes The local line density calculating unit 13 corresponds to the local line density calculating unit according to claim 1, and the characteristic value calculating unit 14 corresponds to the circumscribed width calculating unit according to claim 1. 1 corresponds to the characteristic value calculating means described in 1.
[0033]
The peripheral distribution calculation unit 11 is a calculation unit that calculates the peripheral distribution of a concatenated character block (a character block in which two or more characters are concatenated), and specifically scans a horizontally-written concatenated character block in the vertical direction. The number of black pixels is counted for each scanning position, and a value obtained by dividing the count value at each scanning position by the maximum count value in the connected character block is calculated as the peripheral distribution.
[0034]
Here, a calculation example of the peripheral distribution in the peripheral distribution calculation unit 11 will be described. FIG. 6 is an explanatory diagram illustrating a calculation example of the peripheral distribution in the peripheral distribution calculation unit 11 illustrated in FIG. When calculating the peripheral distribution for the concatenated character block shown in the figure, first, the count value at each scanning position is obtained as follows.
Count value of scanning position A = T1 + T2 = 3
Count value of scanning position B = T1 = 1
Although only two positions A and B are shown as scanning positions, this is for explanation, and scanning is performed pixel by pixel from the left end to the right end of the connected character block.
[0035]
Then, the peripheral distribution at each scanning position is calculated as follows.
Peripheral distribution of scanning position A = 3 ÷ 3 = 1
Peripheral distribution of scanning position B = 1 ÷ 3 = 0.34
[0036]
The circumscribing width calculation unit 12 shown in FIG. 5 is a calculation unit that calculates the circumscribing width of the concatenated character block. Specifically, the horizontal concatenated character block is scanned in the vertical direction and positioned at the lowest position at each scanning position. The width of the black pixel to be measured and the black pixel located at the top is measured, and a value obtained by dividing the measurement width at each scanning position by the maximum measurement width in the connected character block is calculated as the circumscribed width.
[0037]
Here, an example of calculating the circumscribed width in the circumscribed width calculating unit 12 will be described. FIG. 7 is an explanatory diagram illustrating an example of calculating the circumscribed width in the circumscribed width calculating unit 12 illustrated in FIG. 5. When calculating the circumscribing width for the concatenated character block shown in the figure, first, the measurement width at each scanning position is obtained as follows.
Measurement width at scanning position A = T3 = 4
Measurement width at scanning position B = T4 = 1
Although only two positions A and B are shown as scanning positions, this is for explanation, and scanning is performed pixel by pixel from the left end to the right end of the connected character block.
[0038]
Then, the circumscribed width of each scanning position is calculated as follows.
Circumscribed width of scanning position A = 4 ÷ 4 = 1
Circumscribed width of scanning position B = 1 ÷ 4 = 0.25
[0039]
The local line density calculation unit 13 illustrated in FIG. 5 is a calculation unit that calculates the local line density of a connected character block. Specifically, in each pixel, the interval between the character line edges adjacent in the horizontal direction and the interval between the character line edges adjacent in the vertical direction are calculated, and the larger reciprocal of the reciprocal of these line intervals is calculated for the pixel. Calculated as a linear density value. Then, a value obtained by adding the line density of each pixel to each scanning position is obtained, and a value obtained by dividing the added value of each scanning position by the maximum added value in the connected character block is calculated as the local line density.
[0040]
In addition, as a calculation formula of the line interval in the local line density calculation unit 13, when there is no black pixel on one side in the horizontal direction, when there is no black pixel on both sides in the horizontal direction, or when there is no black pixel on one side in the vertical direction For example, a predetermined calculation formula is used according to each situation.
[0041]
Here, a calculation example of the linear density value in the local linear density calculation unit 13 will be described. FIG. 8 is an explanatory diagram illustrating a calculation example of the line density value in the local line density calculation unit 13 illustrated in FIG. 5. When calculating the line density value at the pixel of interest on the scanning position A for the connected character block shown in the figure, first, the horizontal line interval (Lx) and the vertical line interval (Ly) are calculated as follows. Is done.
Lx = (L2-L1 + L4-L3) / 2 = 9
Ly = (L2-L1 + L4-L3) / 2 = 3
Note that L1 to L4 are the coordinates of the character line edge located closest in the vertical and horizontal directions of the target pixel.
[0042]
That is, for the X direction,
L1: The nearest coordinates where the pixel changes from white to black when viewed from the target pixel to the left
L3: nearest coordinates where the pixel changes from black to white when viewed from the target pixel to the left
L2: nearest coordinates where the pixel changes from black to white when viewed from the target pixel in the right direction
L4: nearest coordinates where the pixel changes from white to black when viewed from the target pixel in the right direction
Similarly, for the Y direction,
L1: The nearest coordinates where the pixel changes from white to black when viewed downward from the target pixel
L3: the nearest coordinates where the pixel changes from black to white when viewed downward from the target pixel
L2: nearest coordinates where the pixel changes from black to white when viewed upward from the target pixel
L4: nearest coordinates where the pixel changes from white to black when viewed upward from the target pixel
It is. In FIG. 8, it is 5 mm per pixel.
[0043]
Then, the linear density value of the target pixel is obtained as follows.
Linear density value = max (Wx / Lx, Wy / Ly)
= Max (19 / 9,6 / 3) = 2.1
Wx is the horizontal width of the connected character block, and Wy is the vertical width of the connected character block.
[0044]
In this way, the line density value of each pixel is obtained, and this line density is added for each scanning position. Then, a value obtained by dividing the added value at each scanning position by the maximum added value in the connected character block is calculated as the local linear density.
[0045]
The characteristic value calculation unit 14 illustrated in FIG. 5 includes the peripheral distribution calculated by the peripheral distribution calculation unit 11, the circumscribed width calculated by the circumscribing width calculation unit 12, and the local line density calculated by the local line density calculation unit 13. It is a calculation unit that calculates the characteristic value of each scanning position by adding the line density for each scanning position.
[0046]
In the present embodiment, the characteristic value is calculated by simply adding the marginal distribution, circumscribing width, and local line density in the characteristic value calculation unit 14. By doing so, the characteristic value can also be calculated.
[0047]
The separation control unit 10 is a processing unit that controls each part of the connected character block separation unit 6. Specifically, it is determined whether or not the character block input from the character block storage unit 5 is a connected character block, the determination of the separation position of the connected character block based on the characteristic value calculated by the characteristic value calculation unit 14, Processing such as re-registration of each character block separated at the separation position in the character block storage unit 5 is performed.
[0048]
The separation position of the concatenated character block is determined as the separation position of the concatenated character block, among the characteristic values calculated by the characteristic value calculation unit 14, the scanning position of the characteristic value that takes the minimum minimum value.
[0049]
Here, an example of determination of the separation position of the connected character block in the separation control unit 10 will be described. FIG. 9 is an explanatory diagram showing an example of determining the separation position of the connected character block in the separation control unit 10 shown in FIG. As shown in FIG. 9, for the connected character block “Tatsuno”, the characteristic value calculation unit 14 adds the peripheral distribution (a), the circumscribed width (b), and the local line density (c) to obtain the characteristic value ( d) is calculated.
[0050]
Then, in the separation control unit 10, as shown in FIG. 4D, the scanning positions A, B, and C at which the characteristic values are minimum values are set as separation position candidates, and among these A, B, and C candidates, The scanning position B having the minimum characteristic value is determined as the separation position of the connected character block “Tatsuno”. As a result, the connected character block “Tatsuno” is separated into “dragon” and “field”.
[0051]
Next, the connected character separation procedure of the connected character block separation unit 6 having the above configuration will be described. FIG. 10 is a flowchart showing a concatenated character separation procedure in the concatenated character block separation unit 6 shown in FIGS. 1 and 5. As shown in the figure, the concatenated character block separation unit 6 first inputs a character block stored in the character block storage unit 5 in the separation control unit 10 (step S1001), and the input character block is a concatenated character. It is determined whether it is a block (step S1002).
[0052]
If it is determined in step S1002 that the character block is not a concatenated character block (No in step S1002), it is determined whether all character blocks have been input (step S1009). If it is determined that all character blocks have been input (Yes at step S1009), the connected character separation process is terminated. On the other hand, if it is determined that all character blocks have not been input (step S1009). No), the next character block is input (step S1001).
[0053]
If it is determined in step S1002 that the character block is a connected character block (Yes in step S1002), the peripheral distribution calculation unit 11 calculates the peripheral distribution (step S1003), and the circumscribing width calculation unit 12 The width is calculated (step S1004), and the local linear density calculation unit 13 calculates the local linear density (step S1005). The calculation of the peripheral distribution (step S1003), the circumscribed width (step S1004), and the local linear density (step S1005) may be processed in any order, or may be processed in parallel. Also good.
[0054]
The characteristic value calculation unit 14 calculates the characteristic value by adding the peripheral distribution, the circumscribed width, and the local line density (step S1006). In the separation control unit 10, the scanning position at which the characteristic value takes the minimum value is calculated. Is determined as the separation position (step S1007), the connected character blocks are separated at this separation position, and each separated character block is re-registered in the character block storage unit 5 (step S1008).
[0055]
Then, the separation control unit 10 determines whether or not all character blocks have been input (step S1009). If it is determined that all character blocks have been input (Yes in step S1009), the process of separating connected characters is performed. On the other hand, if it is determined that not all character blocks have been input (No at step S1009), the next character block is input (step S1001).
[0056]
As described above, in the present embodiment, in the connected character block separating unit 6, the peripheral distribution calculating unit 11 calculates the peripheral distribution by counting the black pixels for each scanning position, and the circumscribing width calculating unit 12 The circumscribing width is calculated by obtaining the interval between the black pixels at both ends for each scanning position, and the local line density calculation unit 13 performs local detection based on the spacing between the character line edges adjacent to each other in the vertical and horizontal directions for each scanning position. The line density is calculated, the characteristic value calculation unit 14 adds the peripheral distribution, the circumscribed width, and the local line density for each scanning position to calculate the characteristic value, and the separation control unit 10 calculates the smallest of the characteristic values. Since the scanning position having the minimum value is determined as the separation position, it is possible to separate the connected characters directly and accurately without setting the character separation range.
[0057]
In the present embodiment, the local line density calculation unit 13 calculates the interval between the character line edges adjacent to each pixel in the vertical direction and the interval between the character line edges adjacent in the horizontal direction as a line interval. The larger of the reciprocal of the interval is calculated as the pixel line density value, and this pixel line density value is added for each scanning position to calculate the local line density, so that the complexity of the characters is properly expressed. By calculating the characteristic value, it is possible to determine the separation position of the connected character directly and accurately.
[0058]
The concatenated character separation method described in this embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, floppy disk, CD-ROM, MO, and DVD, and is executed by being read from the recording medium by the computer. The program can be distributed via the recording medium and a network such as the Internet.
[0059]
【The invention's effect】
As described above, according to the first aspect of the present invention, the peripheral distribution calculating unit calculates the peripheral distribution by counting the pixel of the character for each scan position of the connected character, and the circumscribing width calculating unit is configured to connect the connected character. A character line edge that calculates a circumscribing width by obtaining an interval between character pixels at both ends for each character scanning position, and the local line density calculating means is adjacent to each pixel in the XY direction for each connected character scanning position. The local line density is calculated based on the interval, and the characteristic value calculating means calculates the characteristic value by adding the peripheral distribution, circumscribed width and local linear density for each scanning position, and the separation position determining means, Since the scanning position that takes the minimum minimum value among the characteristic values calculated by the characteristic value calculating means is determined as the separation position, the connected characters can be directly and accurately set without setting the character separation range. Can be separated An effect that shaped separation device is obtained.
[0060]
According to a second aspect of the present invention, the local line density calculation means uses the interval between the character line edges adjacent to each other in the X direction and the interval between the character line edges adjacent to the Y direction as a line interval. Since the larger of the reciprocal of the calculated line spacing is calculated as the pixel line density value, and the calculated pixel line density value is added for each scanning position, the local line density is calculated. By calculating the characteristic value that appropriately represents the complexity of the characters, it is possible to obtain a connected character separation device that can determine the separation position of the connected characters directly and accurately.
[0061]
According to a third aspect of the present invention, the peripheral distribution calculating step calculates the peripheral distribution by counting character pixels for each scan position of the connected characters, and the circumscribed width calculating step includes the scan position of the connected characters. The circumscribing width is calculated by obtaining the spacing between the pixels of the characters at each end, and the local line density calculation step is based on the spacing between the character line edges adjacent to each other in the XY direction of each pixel at each scanning position of the connected characters. The local line density is calculated, and the characteristic value calculating step calculates the characteristic value by adding the peripheral distribution, circumscribed width and local line density for each scanning position, and the separation position determining step calculates the characteristic value. Since the scanning position having the smallest minimum value among the characteristic values calculated in the process is determined as the separation position, it is possible to separate the connected characters directly and accurately without setting the character separation range. Possible concatenated character separation methods An effect that is.
[0062]
According to a fourth aspect of the present invention, the local line density calculation step uses the interval between the character line edges adjacent to each pixel in the X direction and the interval between the character line edges adjacent to the Y direction as a line interval. Since the larger of the reciprocal of the calculated line spacing is calculated as the pixel line density value, and the calculated pixel line density value is added for each scanning position, the local line density is calculated. By calculating the characteristic value that appropriately represents the complexity of the characters, it is possible to obtain a connected character separation method capable of directly and accurately determining the separation position of the connected characters.
[0063]
According to the invention of claim 5, by recording a program for causing a computer to execute the method described in claim 3 or 4, the program can be machine-readable. There is an effect that a recording medium whose operation can be realized by a computer can be obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a character recognition device used in the present embodiment.
FIG. 2 is an explanatory diagram illustrating an example of character block extraction in the character block extraction unit illustrated in FIG. 1;
FIG. 3 is an explanatory diagram showing an example of storing character blocks in the character block storage unit shown in FIG. 1;
FIG. 4 is an explanatory diagram showing an example of creating character candidate segments in the character candidate segment creating unit shown in FIG. 1;
FIG. 5 is a block diagram illustrating a configuration of a connected character block separation unit illustrated in FIG. 1;
6 is an explanatory diagram illustrating an example of calculating a peripheral distribution in the peripheral distribution calculating unit illustrated in FIG. 5;
7 is an explanatory diagram illustrating an example of calculating a circumscribed width in the circumscribed width calculating unit illustrated in FIG. 5;
8 is an explanatory diagram illustrating an example of calculating a line density value in the local line density calculating unit illustrated in FIG. 5;
9 is an explanatory diagram illustrating an example of determining a separation position of a connected character block in the separation control unit illustrated in FIG. 5;
10 is a flowchart showing a concatenated character separation procedure in the concatenated character block separation unit 6 shown in FIGS. 1 and 5. FIG.
FIG. 11 is an explanatory diagram illustrating an example of determining a separation position of connected characters in the prior art.
[Explanation of symbols]
1 Character recognition device
2 Image input section
3 Character string extraction part
4 Character block cutout
5 Character block storage
6 Concatenated character block separator
7 Character candidate segment creation part
8 Character recognition part
9 Knowledge Processing Department
10 Separation control unit
11 Peripheral distribution calculator
12 circumscribed width calculation part
13 Local linear density calculator
14 Characteristic value calculator

Claims (5)

In a connected character separation device that detects a connected character in which a plurality of characters are connected in a document image, scans the detected connected character, determines a connected character separation position, and separates the connected character according to the determined separated position ,
Peripheral distribution calculation means for calculating the peripheral distribution by counting character pixels for each scan position of the connected character;
Circumscribing width calculating means for calculating a circumscribing width by obtaining an interval between character pixels at both ends for each scanning position of the connected character;
Local line density calculating means for calculating a local line density based on an interval between character line edges adjacent in the XY direction of each pixel for each scanning position of the connected character;
Characteristic value calculation means for calculating the characteristic value by adding the peripheral distribution, circumscribed width and local line density for each scanning position;
A separation position determining means for determining, as a separation position, a scanning position that takes a minimum minimum value among the characteristic values calculated by the characteristic value calculating means;
A connected character separating apparatus comprising: The local line density calculation means calculates the interval between the character line edges adjacent to each pixel in the X direction and the interval between the character line edges adjacent to the Y direction as a line interval, and the reciprocal of the calculated line interval. 2. The connected character separation device according to claim 1, wherein a larger one of the calculated values is calculated as a pixel line density value, and the calculated pixel line density value is added for each scanning position to calculate a local line density. . In a connected character separation method for detecting a connected character in which a plurality of characters are connected in a document image, scanning the detected connected character to determine a connected character separation position, and separating the connected character by the determined separated position ,
A peripheral distribution calculating step of calculating a peripheral distribution by counting the pixels of the character for each scanning position of the connected character;
A circumscribing width calculating step of calculating a circumscribing width by calculating an interval between character pixels at both ends for each scanning position of the connected character;
A local line density calculating step for calculating a local line density based on an interval between character line edges adjacent to each pixel in the XY direction for each scanning position of the connected character;
A characteristic value calculating step of calculating the characteristic value by adding the peripheral distribution, circumscribed width and local line density for each scanning position;
A separation position determination step for determining a scanning position that takes a minimum minimum value among the characteristic values calculated by the characteristic value calculation step as a separation position;
Concatenated character separation method characterized by including. In the local line density calculation step, the interval between the character line edges adjacent to the X direction of each pixel and the interval between the character line edges adjacent to the Y direction are calculated as line intervals, and the reciprocal of the calculated line interval is calculated. 4. The method of separating connected characters according to claim 3, wherein a larger one of them is calculated as a pixel line density value, and the calculated pixel line density value is added for each scanning position to calculate a local line density. . 5. A computer-readable recording medium on which a program for causing a computer to execute the method according to claim 3 or 4 is recorded.

Images