JPH03269690A - Character segmenting device - Google Patents

Abstract

PURPOSE:To obtain this character segmenting device with high accuracy without generating positional deviation and segmenting failure by detecting the position of a character frame from the density distribution of the multi-valued density data of an input picture and segmenting a character pattern at every detected character frame. CONSTITUTION:Multi-valued density data inputted from an input terminal 80 are stored in a memory 10. The multi-valued density data of the memory 10 are referred to and the respective density ranges of a character frame and a character are derived from the distribution of respective density in a character frame/character density detecting circuit 20. A binary picture to make the pic ture elements of the density range of the character frame detecting the circuit 20 black and the other picture elements white is generated and stored in a memory 40 among input pictures in a binary picture generating circuit 60. The position of the character frame is detected from the binary picture stored in the memory 30 and the character pattern of the binary picture involved in the detected character frame position among the binary pictures stored in the memory 40 in a pattern segmenting circuit 70.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to a character cutting device, and more particularly, to an optical character reading device for reading characters written on paper on which character frames are printed. The present invention relates to a character cutting device for cutting out characters written on a paper.

[Conventional technology]

Conventionally, this type of character cutting device estimates the position of the character frame based on character frame position information given in advance, and extracts the character pattern written in each character frame within a certain range from the density of the printed color of the character frame. It is cut out from a binary image in which pixels with a density exceeding 1 are black and other pixels are white.

[Problem to be solved by the invention]

The conventional character cutting device described above estimates the character frame position based on the character frame position information given in advance.
If there is an error in the character frame position information, there is a drawback that a shift occurs between the estimated character frame position and the actual character frame position, resulting in digit shift or cropping defects.

FIGS. 5(1) and 5(2) are diagrams illustrating problems with conventional character cutting devices.

Character frame 1a on paper 3 shown in FIG. 5(1).

"1", "blank", "3" from the left in lb and lc
Frames 2a and 2b are indicated by broken lines as character frame position information for an input image in which characters are written in this order.

Suppose that 2C is given in advance. At this time, in the conventional character cutting device, the frame 2a shown by the broken line.

Since the characters are cut out according to 2b and 2c, the cut out characters are equivalent to the input characters shown in FIG. 5(2).

Therefore, the character rlJ originally cut out at the - digit is 2
Misalignment occurs when the digit is cut out as the digit, and defective cutting occurs where the characters ``blank'' and ``3'', which are originally cut out as the 2nd and 3rd digit, disappear or are missing.

An object of the present invention is to provide a highly accurate character cutting device that does not cause digit shifts or cutting defects.

[Means to solve the problem]

The character cutting device of the present invention includes a first memory in which multivalued density data of an input image is stored, a second memory, and a second memory. Referring to the multilevel density data stored in the third memory and the first memory,
A character frame/character density detection circuit that calculates the density range of the character frame and the character density range from the distribution number of each density, and pixels in the density range of the character frame detected by the character frame/character density detection circuit in the input image. A first binary image generation circuit that generates a binary image in which pixels are black and other pixels are white and stored in a second memory, and pixels detected in the input image by the character frame/character density detection circuit. a second binary image generation circuit that generates a binary image in which pixels in a character density range are black and other pixels are white, and stores the generated binary image in a third memory; a pattern cutting circuit that detects the position of a character frame from a value image and cuts out a character pattern of the binary image included in the detected character frame position from among the binary images stored in a third memory; has.

[Function J] The character cutting device of the present invention refers to the multilevel density data of the input image, calculates the density range of the character frame and the density range of the character from the distribution number of each density, and extracts the density range of the character frame from the input image. The position of the character frame is automatically detected by generating binary pixels in which pixels in the density range are black and other pixels are white, and the density range of characters in the input image that is included within the position of this character frame is calculated. By cutting out a character pattern from a binary image in which pixels are black and other pixels are white, digit shifts and cutting defects are prevented.

[Example] Next, an example of the present invention will be described with reference to the drawings.

Fig. 1 is a block diagram showing an embodiment of the character cutting device of the present invention, Figs. 2 (1), (2), (3), Figs. FIG. 3 is a diagram illustrating the operation of the device.

This character cutting device 100 has an input terminal 80 to which multi-value density data obtained by converting the density of each input pixel into multi-values using an image sensor, etc., and an input terminal 80 to which multi-value density data of an input image is stored. By referring to the multilevel density data stored in the first memory 10, second memory 30, third memory 40, and first memory 10, the density range of the character frame and A character frame/character density detection circuit 20 for determining the character density range, and pixels in the character frame density range detected by the character frame/character density detection circuit 20 in the input image are set to black, and other pixels are set to white. Generate a binary image and
The first binary image generation circuit 50 stored in the memory 30 of the input image and the pixels in the character density range detected by the character frame/character density detection circuit 20 are set as black, and the other pixels are set as white. A second binary image generation circuit 60 generates a binary image and stores it in the third memory 40, detects the position of a character frame from the binary image stored in the second memory 30, and generates a third
a pattern cutting circuit 70 for cutting out a character pattern of a binary image included in a detected character frame position from among the binary images stored in the memory 40;
and an output terminal 90 to which the cut out character pattern is output.

The operation of this character cutting device 100 will be described below with reference to FIGS. 2(1), (2), and (3), FIGS. 3, and 4.

Character frame 1a on paper 3 shown in FIG. 2(1).

An example of cutting out characters "2" and "3" written in lb will be explained. Here, each character is written on the white paper 3 at a higher density than the printing color density of the character frames 1a and 1b.

Multi-value density data indicating the density of each pixel on the paper 3, which has been multi-valued by an image sensor or the like, is input from the input terminal 80. Multilevel density data input from the input terminal 80 is stored in the first memory 10. A character frame/character density detection circuit 20 calculates a density distribution from the multi-value density data of all pixels stored in the first memory 10. Since the density of the characters is higher than the density of the printing color of the character frames 1a and lb, the density distribution is as shown in FIG.

The distribution is concentrated at density A, which indicates the density of 1b, and B, which indicates the density of characters. Therefore, from this density distribution, the density range of the character frame (the range between density C and dark in Figure 2 (2)) and the density range of the character (the range above density in Figure 2 (2))
is required. The first binary image monitoring circuit 50 compares the multi-value density data of each pixel stored in the first memory 10 with the density C and the density, so that the density is within the density range of the character frame. Set pixels to black (0), other pixels to white (
1) A binary image is generated. In this binary image, the character frames 1a and lb are black, and the white background of the paper 3 and the characters are white, so that the data is only for the character frames 1a and lb, as shown in FIG. 2(3). This binary image data is stored in the second memory 30. The second binary image generation circuit 60 compares the density with the multi-value density data of each pixel stored in the first memory 10, and selects pixels with a density within the character density range as black ( 0), other pixels white (1)
A binary image is generated and stored in the third memory 40. In this binary image, only the characters are black, and the paper 3 and the character frames 1a and 1b are white, so that the data consists of only the characters, as shown in FIG. 3.

The pattern cutting circuit 70 calculates the distribution number of black pixels in the horizontal and vertical directions of the paper 3 for the binary image stored in the second memory 30, and obtains the distribution shown in FIG. 2 (3). A diagram is obtained. As a result, by finding the peak position of the distribution number in this distribution map,
Axis coordinates (horizontal direction of paper 3) Xi, Xi, X
s, X4oJ: Y-axis coordinate (vertical direction of paper 3) Y
l, Yz are detected. Also, character frame 1a and character frame 1b
Since there is always an X coordinate where the distribution number is O between
The coordinates of the four corners are (Xi, YI), (Xl, YI).

(Xs, Yz), (Xl, YI), the coordinates of the character frame 1a and the four corners are (X3, Yr) (
Xs, Yz), (X4.Yz), (x4゜y+)
It is possible to separate the character frame 1b represented by . Therefore, as shown in FIG. 4, from the binary image data stored in the memory 40 of FIG. 3 in which only the characters are represented in black, the data of each pixel within the coordinates of the four corners described above can be calculated. Character patterns can be cut out by sequentially extracting them. The coordinates of the four corners are (Xi, YI), (
Xi.

By extracting the data of each pixel within Yz)-(Xi, Y2), (Xl, YI) (inside the diagonal line at the bottom left in Figure 4), the character pattern "2" is cut out, and the coordinates of the four corners are ( Xs,
YI ), (X3, Y2) (X4.YI), (
X4. The character pattern "3" is cut out by extracting the data of each pixel within Yl) (inside the diagonal line at the bottom right in FIG. 4). The character pattern thus cut out is output from the output terminal 70 to a character reading device (not shown).

Therefore, in this character cutting device 100, the character frame 1a
, lb position is automatically detected and the character pattern is extracted by extracting a binary image of only the characters within the detected character frame, so characters can be extracted with high precision. .

In the above explanation, cutting out two characters shown in FIG. 2 (1) was taken as an example, but even if the number of characters per line and the number of lines are large, the character cutting device 100 performs the same operation. Needless to say, this is not affected by the size of the character frame, as can be easily understood from the automatic detection of the coordinates of the four corners of the character frame.

〔Effect of the invention〕

As explained above, the present invention automatically detects the position of a character frame from the density distribution of multilevel density data of an input image, and cuts out a character pattern for each detected character frame, thereby eliminating digit shift and cutting out. This has the effect of not causing defects and improving the precision of character extraction.Also, since the coordinates of the four corners of the character frame can be determined, the relative positional relationship between the characters and the character frame can also be determined accurately. This has the effect of improving the accuracy of character discrimination (for example, comma and apostrophe), which is distinguished based on the relative positional relationship between the character and the character frame.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing an embodiment of the character cutting device of the present invention, Figs. 2 (1), (2), (3), Figs. A diagram explaining the operation of the device, Fig. 5 (
1) and (2) are diagrams illustrating problems with conventional character cutting devices. la, lb...Character frame, 3...Paper, 10...First memory, 20...Character frame/character density detection circuit, 30...Second
memory, 40... third memory, 50... first binary image generation circuit, 60... second binary image generation circuit, 70... pattern cutting circuit, 80...・Input terminal, 90...Output terminal, 100...
・Character cutting device, A, B, C, D... Density, XI, Xi, Xs, X4. Yl, Y*', coordinates.

Claims (1)

[Scope of Claims] 1. A character cutting device that is installed in an optical character reading device that reads characters written on paper on which character frames are printed, and that cuts out the characters written in each character frame, The multi-value density data stored in the first memory, the second and third memories, and the first memory are stored, and characters are generated from the distribution number of each density. A character frame/character density detection circuit that calculates the density range of the frame and the character density range, and pixels in the density range of the character frame detected by the character frame/character density detection circuit in the input image are black, and other pixels are A first two-dimensional image that generates a binary image to be white and stores it in a second memory.
A value image generation circuit generates a binary image in which pixels in the character density range detected by the character frame/character density detection circuit are black and other pixels are white in the input image, and stores the image in a third memory. A second binary image generating circuit detects the position of the character frame from the binary image stored in the second memory, and detects the position of the character frame from the binary image stored in the third memory. A character cutting device comprising a pattern cutting circuit for cutting out a character pattern of a binary image included in a character frame position.