JPH03207165A - Character area separation system - Google Patents

Abstract

PURPOSE:To securely separate even a thick character by combining a character area separating method which uses a density gradient with a character area separating method which utilizes the presence of black group picture elements on a white background. CONSTITUTION:A white background detection part 2 detects an aimed picture element as a white background picture element and sends it to an AND circuit 4 when there are white group picture elements within a certain constant distance in both the right and left main-scanning directions of the aimed picture element. A black group picture element detection part 3, on the other hand, detects the center picture elements of a matrix as black group picture elements when all picture elements are black picture elements and outputs them to the AND circuit 4, and also outputs the aimed picture element as an active picture element when all the picture elements are black group picture elements, and an expansion processing part 5 expands the aimed picture element according to the active picture element. A character area extraction part 6 which uses ridge picture elements separates a character area and sends the separation result to an OR circuit 7. Consequently, even the thick character is accurately separated as the character area.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a character region separation method for separating character regions from a digital multi-gradation input image containing a mixture of characters and pictures (such as photographs other than text). .

[Conventional technology]

For example, when playing back an image containing a mixture of text and pictures on a binary image playback device such as a digital copying machine or facsimile, in order to improve the quality of the playback image, it is necessary to perform the most appropriate image processing according to each area. It is desirable to apply it adaptively. In order to carry out such image processing, it is necessary to accurately separate character areas and other picture areas in the image as pre-processing.

Conventionally, as a method for separating character areas in such an image, for example, the method of Tetsuya et al. Journal of the Institute of Electronics and Communication Engineers, '84/7 Vol. J67-BNo.
．． 7, pp. 781-788). This method divides the original into blocks of a predetermined size, determines the maximum and minimum density values of pixels in the block, and determines the difference between the maximum and minimum density values when the difference is greater than a predetermined reference value. If it is larger than the reference value, the block is determined to be a binary image area (character area), and if it is smaller than the reference value, it is determined to be a grayscale image area (
Furthermore, in the blocks that are continuous in the main scanning direction or the sub-scanning direction, the outer block is determined to be 2
In the case of a value image area, the blocks inside it are corrected to a binary image area.

On the other hand, the present applicant previously filed an application in Japanese Patent Application No. 63-276000 regarding a line drawing area separation method for separating continuous tone photographic areas from line drawing areas such as characters and line drawings. Line drawing area separation method 0 according to this application detects ridge pixels (maximum density pixels) in an image by applying a two-dimensional ridge pixel detection pattern prepared in advance to an input image, and The number is counted using a predetermined area as a unit, and it is determined whether the pixel of interest or block of interest is a line drawing area based on the magnitude of the count value.

[Problem to be solved by the invention]

All of the above-mentioned methods basically separate regions using the density gradient of the image. For this reason, for example,
In the case of Tetsuya et al.'s method, there is a possibility that the inside of a character is mistakenly determined to be a picture region, and conversely, the edge of a picture may be mistakenly determined to be a character region. To avoid this, Tetsuya et al.'s method uses
In blocks that are continuous in the main scanning direction or sub-scanning direction, if the outer block is a binary image area, the inner block is corrected to the 21 image N area, but up to characters of a certain thickness. There was a problem that could only be addressed.

In addition, patent application No. 63-276000 related to the applicant's earlier application
In the case of the above method as well, since the inside of a thick character is flat in terms of density, ridge pixels within the character cannot be detected, and there is a risk that the thick character may be misjudged as a picture area.

The present invention was made under the above circumstances, and its purpose is to provide a character area separation method that can accurately separate an image including bold characters into a character area and a picture area. It is.

[Means to solve the problem]

In order to achieve the above object, a first region separation method of the present invention is a method for separating character regions of a digital multi-tone input image signal containing a mixture of characters and pictures. The edge part of is separated as a character area, the input image signal is binarized at a low level, and N
When all pixels in a matrix of When doing so, the pixel or block of interest is a white background pixel, the input image signal is binarized at a high level, and when all pixels in the MXM matrix are black pixels, the center pixel of the matrix is set as a black block pixel, When the pixel of interest is a white background pixel and a black block pixel, or when the block of interest is a white background pixel and part of it is a black block pixel, the pixel of interest or the block of interest is subjected to expansion processing. The expanded character area and the character area separated by the density gradient are both output as the character area.

Furthermore, a second character region separation method of the present invention is a method for separating character regions of a digital multi-tone input image signal containing a mixture of characters and pictures, in which edge portions of characters are separated from the input image signal using a density gradient. The input image signal is binarized at a low level, and when all pixels in the NXN matrix are white pixels, the center pixel of the matrix is defined as a white lump pixel, and at least one of the pixel of interest or block of interest is separated as a character region, and the input image signal is binarized at a low level. When a white mass pixel exists within a predetermined distance in two or more predetermined symmetrical directions, the pixel of interest or block of interest is set as a white background pixel, the input image signal is binarized at a high level, and the pixel in the MXM matrix is When all pixels are black pixels, the center pixel of the matrix is set as a black block pixel, the black block pixel is subjected to expansion processing, and the input image signal is binarized at a low level, so that the number of white pixels in the LXL matrix is a predetermined number. When the value is greater than or equal to the value of When the block of interest is a white background pixel, a part of which is a black block pixel that has been subjected to dilation processing, and a part of which is a white background pixel that has been subjected to dilation processing, the pixel of interest Alternatively, the block of interest is determined to be a character area, and both the determined character area and the character area separated by the density gradient are output as the character area.

[For production]

As described above, character areas in an image can generally be separated using a density gradient. However, in the case of this density gradient, as described above, the inside of a character, especially the inside of a thick character, may be erroneously determined to be a picture area, and the edge portion of a picture may be erroneously determined to be a character area. Therefore, the present invention separates character regions in an image by combining the conventional method of character region separation using density gradients and the method of character region separation using the presence of black block pixels on a white background. be. As a result, it is possible to reliably separate thick characters, which were difficult to separate using the conventional concentration gradient method.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an example of a character region separation device to which the first character region separation method of the present invention is applied.

In the figure, 1 is a white block pixel detection section, 2 is a white background detection section, 3 is a black block pixel detection section, 4 is an AND circuit, 5 is an expansion processing section, and 6
is the patent application No. 63-2760 related to the above-mentioned applicant's prior order.
Character area extraction unit using ridge pixel detection No. 00, 7 is OR
It is a circuit.

The white block pixel detection unit 1 binarizes the digital multi-gradation input image signal using a low-level threshold, and then converts it into an NXN matrix, for example, a 5×5 matrix, when all pixels are white pixels. The center pixel of is detected as a white block pixel and sent to the white background detection section 2.

As illustrated in FIG. 3, the white background detection unit 2 detects a pixel of interest as a white background pixel when there are white block pixels within a certain distance in both left and right directions of the main scanning direction of the pixel of interest. It is detected and sent to the AND circuit 4. Through the processing of the white block pixel detection unit 1 and the white background detection unit 2, it is estimated that the area around the pixel of interest is a white background.

On the other hand, the black block pixel detection unit 3 binarizes the digital multi-gradation input image signal using a high-level threshold, and then converts it into an MXM matrix, for example, a 3×3 matrix, when all pixels are black pixels. The center pixel of the matrix is detected as a black block pixel and sent to the AND circuit 4. Through this processing by the black block pixel detection unit 3, it is estimated that the pixel of interest is a pixel forming a character.

The AND circuit 4 ANDs the detection result sent from the white background detection section 4 and the detection result sent from the black block pixel detection section 3, and determines whether the pixel of interest is a white background pixel and a black block pixel. At some point, the pixel of interest is output as an active pixel. This determines that the pixel of interest is a black pixel on a white background, that is, a character.

The expansion processing section 5 performs expansion processing on the pixel of interest based on the active pixels sent from the expansion processing section 5.

For example, if there is even one active pixel in a 5×5 matrix, the pixel of interest is determined to be a character area and sent to the OR circuit 7.

Through this dilation process, internal pixels of a thick character are corrected as a character area.

On the other hand, the character region extracting section 6 using ridge pixels separates character regions as follows, and sends the separation results to the OR circuit 7.

First, a 5×5 matrix, for example, is sequentially applied to an input image, and when the following conditional expression is satisfied within this matrix, a central pixel is detected as a ridge pixel of a character. In other words, the density level of the center pixel of the matrix is L.
c, the density level of a pixel pair located at an arbitrary point symmetrical position with this center pixel Lc in between. , L, then Lc-L. ≧LTN, Ka Le-L. ≧L? )ILTH
: When there is at least one pixel pair that satisfies the threshold, the pixel of interest L
c is detected as a ridge pixel of a character as illustrated in FIG.

Next, the number of ridge pixels detected as described above is counted in units of a matrix of a predetermined size, for example, a 9×9 matrix. If the counted number of ridge pixels is greater than or equal to a predetermined value, the pixel of interest is determined to be a character area and sent to the OR circuit 7.

The OR circuit 7 performs an OR operation on the outputs of the character area extraction section 6 and the expansion processing section 5, and outputs the result as a character area.

Therefore, for characters of normal thickness, the character region extracting section 6 separates the regions by a method using ridge pixels, and for thick characters where ridge pixels cannot be detected, the circuit consisting of symbols 1 to 5 is used. Since character regions are separated based on the presence of black block pixels on a white background, regions can be reliably separated no matter what line thickness the image contains.

FIG. 2 is an example of a character region separation device to which the second character region separation method of the present invention is applied.

The example in FIG. 2 differs from the example in FIG. After ANDing the outputs of the processing section 8 and the expansion processing section 9, O
The point is that the signal is sent to the R circuit 7. Note that circuits that are the same as those in FIG. 1 are given the same reference numerals and their explanations will be omitted.

The expansion processing unit 8 expands the black block pixels obtained by the black block pixel detection unit 3. For example, in a 5x5 matrix, 1
If a black block pixel exists at any time, the pixel of interest is determined to be a black block pixel.

On the other hand, the white background detection unit 9 binarizes the input image signal at a low level, and if the number of white pixels in the LXL matrix, for example, a 5×5 matrix is a predetermined value or more (for example, 3 or more), the The center pixel of the matrix is determined to be a white background pixel.

Then, the expansion processing unit 10 expands the white background pixels obtained by the white background detection unit 9. For example, if even one white background exists in a 5×5 matrix, the pixel of interest is set as a white background.

Through the processing of the white background detection unit 9 and the expansion processing unit 10, in addition to the detection of the white background by the white block pixel detection unit 1 and the white background detection unit 2 described above, a different determination method is used to determine whether the background is white or not. is estimated.

Therefore, in the case of the example in FIG. 2, the white background detection section 9
As a result, the AND circuit 4 generates an output when the pixel of interest is a white background pixel, a black block pixel, and a white pixel, and It is determined that the pixel is a black block pixel on a white background, that is, a character area. Therefore, in the case of the example shown in FIG. 2, region separation can be performed more reliably.

In the above-described embodiments, separation processing is performed pixel by pixel, but processing can also be performed block by block.

In addition, as the character area detection unit 6, the method disclosed in Japanese Patent Application No. 63-276000 related to the applicant's earlier application was used; Of course, the method can be used.

Further, although the white background detection unit 2 performs processing in the main scanning direction as shown in FIG. 3, it can also perform processing in the sub-scanning direction orthogonal thereto, The accuracy can be further improved by performing processing in two directions in the scanning direction simultaneously and ORing the processing results.

〔Effect of the invention〕

As is clear from the above, when using the character area separation method of the present invention, even thick characters can be accurately separated as character areas, and optimal processing can be applied adaptively depending on the image area. This makes it possible to improve the quality of reproduced images. Furthermore, since it is detected whether a pixel of interest belongs to a text area based on the presence of black block pixels on a white background, it is less likely that a picture area will be mistakenly separated as a text area.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of a character region separation device configured by applying the first character region separation method of the present invention, and FIG. 2 is a block diagram of an embodiment of a character region separation device configured by applying the first character region separation method of the present invention. FIG. 3 is a block diagram of an embodiment of the constructed character area separation device; FIG. 3 is an explanatory diagram of detection of white background pixels; FIG. 4 is an explanatory diagram of detection of ridge pixels. ■...White block pixel detection unit, 2...White background detection unit, 3
. . . Black block pixel detection unit, 4 . . . AND circuit, 5, 8.
DESCRIPTION OF SYMBOLS 10... Expansion processing unit, 6... Character area extraction unit based on density gradient, 7... OR circuit, 9... White background detection unit. Renguo Sχti4 Vermilion's Transcription 1 Figure 3 CG) Pig (b) A-A 斯147! l+! Mr. Can L! Figure 4

Claims (2)

[Claims] (1) In a method for separating character regions of a digital multi-gradation input image signal containing a mixture of characters and pictures, the edge portions of characters are separated as character regions from the input image signal using a density gradient, and the input image signal is separated into character regions. When binarized at a low level and all pixels in an N×N matrix are white pixels, the center pixel of the matrix is defined as a white block pixel, and a predetermined value is set in at least one predetermined symmetrical direction of the pixel of interest or block of interest. When a white block pixel exists within a distance of
When all pixels in the ×M matrix are black pixels, the center pixel of the matrix is a black block pixel, and when the pixel of interest is a white background pixel and a black block pixel, or the block of interest is a white background pixel. , and a part of the pixel is a black block pixel, the pixel of interest or the block of interest is expanded to form a character area, and both the expanded character area and the character area separated by the density gradient are formed into a character area. A character area separation method characterized by outputting as . (2) In a method for separating character regions of a digital multi-gradation input image signal containing a mixture of characters and pictures, the edge portions of characters are separated as character regions from the input image signal using a density gradient, and the input image signal is separated into character regions. When binarized at a low level and all pixels in an N×N matrix are white pixels, the center pixel of the matrix is defined as a white block pixel, and a predetermined value is set in at least one predetermined symmetrical direction of the pixel of interest or block of interest. When a white block pixel exists within a distance of
When all pixels in the xM matrix are black pixels, the center pixel of the matrix is set as a black block pixel, the black block pixel is expanded, the input image signal is binarized at a low level, and the L x L matrix is When the number of white pixels in the matrix is greater than or equal to a predetermined value, the center pixel of the matrix is subjected to dilation processing as a white background pixel, and the pixel of interest is a white background pixel, is a black block pixel after dilation processing, and is dilated. When the block of interest is a white background pixel after being processed, or when the block of interest is a white background pixel and a part of it is a black block pixel that has been subjected to dilation processing, and a part of it is a white background pixel that has been subjected to dilation processing. When the pixel is a pixel, the pixel of interest or the block of interest is determined to be a character region, and both the determined character region and the character region separated by the density gradient are output as the character region. Separation method.