JPH09219791A - Binarization processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To properly execute the separation of a character picture area from a gradation picture area. SOLUTION: A picture judging part 7 calculates the gradation difference between an area adding a picture element to be considered and the area adding the picture element which is separated from the picture element to be considered by planned distance, that is, a density pitch. The picture elements of an n-th line and the n-1-th line are objects concerning density data of the picture element for calculating the density pitch. The area is judged to be the density picture area when the calculated density pitch is within a planned range and is judged to be the character picture area when the pitch is out of the planned range and the judgement result is inputted to a difference diffusing circuit 3. The difference diffusing circuit 3 gives a bias for diffusing difference to input data at the time of the density picture area and supplies input data to a binarizing circuit by '0' bias at the time of the character picture area.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a binarization processing device, and more particularly to a binarization processing device for processing image information obtained by reading an original image in which a character image and a grayscale image are mixed.

[0002]

2. Description of the Related Art When an original including a character image and a grayscale image is binarized using a single threshold, the original image can be beautifully reproduced in the character image area, but the original image is reproduced in the grayscale image area. The sex is not good. On the other hand, for such manuscripts,
When a halftone processing method such as dithering or an error diffusion method is used, the grayscale image area is reproduced beautifully, but in the character image area, the character edge is blurred and a clear character cannot be obtained. As described above, it is known that there are binarization methods suitable for the character image and the grayscale image.

Therefore, it is often the case that the character image area and the grayscale image area are well separated and binarization processing suitable for each is performed. The division between the character image area and the grayscale image area is called image area separation.

In the conventional image area separation method, attention is paid to a difference in density change between a character image and a grayscale image, and the character image and the grayscale image are separated in a microscopic area by the difference in the density change. That is, binarization by a single threshold value is performed assuming that the density gradient is large as a character image area, and binarization by the halftone processing method is considered as a density image area where the density gradient is small.

However, in the method of performing image area separation by such processing, for example, when noise exists in the grayscale image area, the density gradient of the noise portion becomes large, so it is erroneously determined to be a character area. The problem that it is done occurs. FIG. 6 is a diagram showing an example of an image read by notch noise. 9A is an image of a document, FIG. 8B is an image read by a scanner, and FIG. 7C is a partially enlarged view of the read image. As described above, an image that is originally free from irregularities or omissions has irregularities or omissions due to noise, and the region is unclear, which causes a problem of erroneous determination.

It is inevitable that an image input by a scanner such as a facsimile machine or a copying machine will have noise, and if a density gradient is calculated from such noisy image information, an erroneous determination result is output. Resulting in.
Also, noise is likely to occur especially when the resolution is high.

On the other hand, in the image processing apparatus disclosed in Japanese Patent Laid-Open No. Hei 2-132968, the density Da within the predetermined range is determined by the average density Da within the predetermined range including the pixel of interest.
All pixels that are larger than the above are added, and pixels that have a density smaller than Da within the predetermined range are added, and the effect of noise is reduced by separating the character image area and the grayscale image area by the difference of their averages. doing.

[0008]

The image processing apparatus described in the above publication still has a problem. That is,
Also with this image processing apparatus, when noise is present, the noise itself affects the average density Da, and as a result, the regions cannot be accurately distinguished. Further, since the average is obtained only in the predetermined range including the pixel of interest, the character image area that is likely to include notch noise is
There is a problem that it is easy to erroneously determine a grayscale image area.

The present invention has been made in view of the above problems, and it is possible to appropriately separate the image areas of the character image area and the grayscale image area even in the case of an original containing noise, and as a result, the characteristics of the image are improved. The purpose is to create a beautiful image output by applying an adapted process.

[0010]

SUMMARY OF THE INVENTION According to the present invention, a first average of a plurality of pixels located around a pixel of interest excluding the minimum and maximum densities is calculated. The density calculating means and a plurality of pixels existing on the same line as the pixel of interest and located in the periphery of neighboring pixels located at a predetermined distance from the pixel of interest have the minimum and maximum densities. Second average density calculating means for calculating the average density of the excluded pixels, density gradient calculating means for calculating the difference D between the average densities calculated by the first and second average density calculating means, and the difference D An image determination unit that outputs data indicating a grayscale image area when the density gradient is within a preset range, and outputs data indicating a character image area when the density gradient is outside the range, From the image determination means When the data indicating the image area is output, the halftone processing method is selected to binarize the image information, and when the data indicating the character image area is output from the image determining means, the image information is converted by the single threshold value. A first feature is that the image forming apparatus is provided with a binarizing processing unit for binarizing.

Further, according to the present invention, the average density calculating means selects four pixels from the periphery of the pixel of interest and the neighboring pixels so that the number of pixels excluding the pixels having the maximum and minimum densities is two. A second feature is that the binarization processing unit is configured to binarize image information using an error diffusion method as a halftone processing method when the pixel of interest is determined to be a grayscale image area. There is.

Further, the present invention has a third feature in that a distance setting means for presetting a distance between the target pixel and the neighboring pixels is provided.

According to the first to third characteristics, among the pixels around the target pixel and the neighboring pixels, the density average value of the remaining pixels excluding the pixels having the maximum and minimum densities is used to determine the target pixel and the neighboring pixels. It is possible to calculate a density gradient between regions including pixels. Then, if this density gradient is within the predetermined range, it is determined to be a grayscale image area,
If it is outside the planned range, it is determined to be a character image area. The image information binarization processing method is selected according to the determination result.

Further, according to the second feature, the number of pixels is 2 except for the pixels of maximum and minimum densities, and the structure of the average density calculating means for calculating the average value can be simplified. Also,
According to the third feature, the distance between the target pixel and the neighboring pixels can be set arbitrarily.

[0015]

DETAILED DESCRIPTION OF THE INVENTION The present invention will be described in detail below with reference to the drawings. FIG. 2 is a schematic diagram for explaining the positions of pixels to be processed. As shown in the figure, the pixel of interest X and its peripheral pixels A1 to
A4 is defined, and another pixel (hereinafter, referred to as “neighboring pixel”) Y on the same line as the pixel of interest X and its peripheral pixels B1 to B4 are defined.

FIG. 1 is a block diagram showing a main configuration of an image processing apparatus according to an embodiment of the present invention. In the figure, the first line buffer 1 holds the error output by the error diffusion circuit 3 for one line. Details of the error diffusion circuit 3 will be described later with reference to FIG. The second line buffer 2 holds one line of image data including the pixel of interest X (see FIG. 2). The error diffusion circuit 3 selects either binarization after error diffusion or simple binarization based on the output of the image determination unit 7. The configuration of the image determination unit 7 will be described later with reference to FIG.

The pointer control circuit 4 generates a pointer for instructing the data to be sent to the error diffusion circuit 3 and the image determination section 7. The distance setting device 5 is means used for setting the distance S (see FIG. 2) between the pixel of interest X and the neighboring pixel Y, and R
It can be composed of a memory such as an AM. Pointer control circuit 4
Is used to obtain pixel data in the vicinity of a pixel that is a pixel distant from the pixel of interest X by a value S designated by the distance setting device 5, that is, a neighboring pixel Y. The setting of the distance S between the pixel of interest X and the neighboring pixel Y is performed by a CPU (not shown), and an appropriate value is designated according to the resolution handled in the system. Normally,
A higher resolution gives a larger value, and a lower resolution gives a smaller value.

The buffer 6 is large enough to hold the image data existing within the distance S.
Since the distance S is usually a small value, this buffer may be small. The image determination unit 7 detects whether the pixel of interest X is in the character image area or the grayscale image area. Reference symbol Pi is input data to the processing apparatus, reference symbol Po is a binarized output of the error diffusion circuit 3, and reference symbol Err is an error output of the error diffusion circuit 3.

Next, the image judging section 7 will be described in detail. FIG. 3 is a block diagram showing the configuration of the image determination unit 7. In the figure, the second line buffer 2 stores the image data of the first line (n-1th line) including the target pixel X and the neighboring pixel Y. On the other hand, the buffer 6 includes the pixels B2 to A among the image data of the line (nth line) next to the line including the target pixel X and the neighboring pixel Y.
Up to 4 image data are stored. Average concentration calculator 8
Each of a and 8b is composed of a 6-input adder and a circuit for arithmetically shifting the result by 1 bit to the right. The average density calculator 8a calculates the average density of the pixels B1 to B4 around the pixel Y, and the average density calculator 8b calculates the average density of the pixels A1 to A4 around the target pixel X.

The maximum / minimum detection circuit 9a is a circuit for detecting the maximum density and the minimum density from the pixels B1 to B4 around the pixel Y, and the maximum / minimum detection circuit 9b is the pixels A1 to around the pixel X.
This is a circuit for detecting the maximum density and the minimum density from A4.
When the maximum value output of the maximum / minimum detection circuit 9a is DMaxY and the minimum value output is DMinY, the average density calculator 8a calculates the average density by the following (Equation 1).

(B1 + B2 + B3 + B4-DMaxY-DMinY) / 2 (Equation 1) The average concentration detector 8b is also the same, and the maximum value output of the maximum / minimum detection circuit 9b is DMaxX and the minimum value output is DMin.
If X, the average concentration detector 8a calculates the average concentration by the following (formula 2).

(A1 + A2 + A3 + A4-DMaxX-DMinX) / 2 (Equation 2) As described above, in (Equation 1), the pixel having the maximum density and the pixel having the minimum density are excluded from the pixels B1 to B4, and the other pixels are excluded. The average value of the densities of the two pixels is calculated. Also,
In (Equation 2), the pixel having the maximum density and the pixel having the minimum density are excluded from the pixels A1 to A4, and the average value of the densities of the other two pixels is calculated.

Pixels A1, A2, A3, A for which image data, that is, density data, are input to the average density detectors 8a, 8b and the maximum / minimum detection circuits 9a, 9b.
4, B1, B2, B3, B4 are designated by the pointer control circuit 4 as shown in FIG.

The reason for selecting four points as shown in FIG. 2 when obtaining the average densities from the periphery of the target pixel X and the neighboring pixels Y is as follows. That is, even if error diffusion processing is performed, the line buffer is
This can be done with a book, and if the maximum and minimum pixels are excluded, the average value of 2 pixels will be obtained as a result, and the division will be a division by 2, so it can be done with a shifter and the circuit configuration will be Because it will be easier.

The difference between the average densities obtained by the maximum / minimum detection circuits 9a and 9b is calculated by the differentiator 10 to obtain the concentration gradient. This density gradient is input to the comparator 11 and compared with preset threshold values N1 and N2 to determine whether the target pixel X is in the character image area or the grayscale image area.

The determination algorithm is as follows.
At the threshold value N1 <N2, if N1 ≦ D ≦ N2, it is determined as a grayscale image area, and if D <N1, D> N2, it is determined as a character image area. The comparator 11 outputs the determination output 12, and if the determination output 12 is "0", it indicates that it is a grayscale image area, and if it is "1", it is a character image area.

When the image determination unit 7 determines that the pixel of interest X is a grayscale image region, the error diffusion circuit 3 biases the density data of each pixel by the error diffusion method, and then 2 If the pixel of interest X is binarized and is determined to be the character image area, simple binarization is performed without applying a bias.

FIG. 4 shows the error diffusion circuit 3 concretely, and FIG. 5 shows the error and the coefficient used when obtaining the weighted average of the error data. In FIG. 4, input data Pi ′ is pixel density data supplied from the second line buffer 2 and is, for example, 8-bit 256 gradation data. The first line buffer 1 holds the error data for one line as already described. The binarization circuit 13 compares the input data Pie biased by the adder 14 with a threshold value, for example, “128”, and outputs “1” if the biased input data Pie is larger, and outputs the smaller value. If so, “0” is output as the output data Po.

The level converter 15 converts the output data Po of the binarization circuit 13 into multivalued data. The level converter 1
Reference numeral 5 multiplies the output data Po of the binarization circuit 13 by the maximum value of 8 bits, that is, "255" and outputs it. The level conversion circuit 15 may be composed of a memory which generates multi-valued data "255" or "0" with an input of "1" or "0" as an address. The error calculator 16 calculates the level converter 1 from the output data Pie of the adder 14.
The output multivalued data Po ′ of 5 is subtracted to obtain error data. The error data is input to the weighted average circuit 17 and also to the first line buffer 1. The weighted average circuit 17 is a circuit which weights the error generated in the error diffusion process around the pixel of interest X by appropriately weighting it in advance.

The multiplexer (MUX) 18 is controlled by the judgment output 12 of the image judgment unit 7. When the judgment output 12 is “0”, the multiplexer 18 selects the output of the weighted average circuit 17 and outputs it to the adder 14,
When the judgment output 12 is "1", "0" data is selected and output to the adder 14. That is, the input data P
i ′ is binarized by being biased by the weighted error data output from the weighted average circuit 17 or biased by “0” according to the value of the determination output 12 of the image determination unit 7. It is an input to the circuit 13.

As described above, when the image determination unit 7 determines that the image is a gray image region, the input data is biased by the weighted average value of the error data and then binarized. After the data is biased by "0", it is binarized.

Next, an example of how to obtain a weighted average of error data will be described. As shown in FIG. 5A, it is performed for five pixels a, b, c, d, and e from the previous line (n-1 line) of the target pixel X, and f and g2 pixels in front of the target pixel X. . When the weighted average is calculated for each pixel by the error filter of FIG. 5B, the weighted averaged error data can be calculated by the following (formula 3). Error data = 1 /
16 x a + 2/16 x b + 4/16 x c + 2/16 x d
+ 1/16 × e + 2/16 × f + 4/16 × g (Equation 3) In this way, when it is determined that the target pixel X is in the grayscale image area, the target pixel X is binarized by the following algorithm. Be seen. First, according to (Equation 4), the weighted averaged error data is biased to the target pixel X to obtain output data (denoted as Q). Q = 1/16 x a + 2/16 x b
+ 4/16 × c + 2/16 × d + 1/16 × e + 2/1
6 × f + 4/16 × g + X (Equation 4) If Q ≧ 128, the binarized data “1” is output. If Q <128, binarized data “0” is output.

The output of the error calculator 46 is Q ≧ 12.
If 8, error data (Q-255), if Q <128,
The error data (Q) is output. The error data is stored in the first line buffer 1.

As described above, in the present embodiment, it is determined whether the pixel of interest is in the grayscale image area or the character image area.
The judgment is made based on the average density excluding the maximum and minimum density of the peripheral pixels. When it is determined that the image is a grayscale image region based on the result of this determination, binarization is performed by the error diffusion method. However, this is not limited to the error diffusion method, but a dither processing method, etc.
It can be modified to use other halftone processing schemes.

[0035]

According to the first to third aspects of the invention, among the pixels around the target pixel and the neighboring pixels, the average value of the remaining pixels excluding the pixels having the maximum and minimum densities is used to detect the target pixel. It is possible to calculate a density gradient between regions including pixels and neighboring pixels. Therefore, even when the image contains noise, the image area separation between the character image area and the grayscale image area can be appropriately performed by the density gradient. As a result, an appropriate processing method can be selected according to the image area, and a beautiful binarized image can be obtained.

In particular, according to the second aspect of the invention, the number of remaining pixels excluding the pixels of maximum and minimum densities is two, and the structure of the average density calculating means for calculating the average value can be simplified.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a binarization processing device according to an embodiment of the present invention.

FIG. 2 is a diagram showing positions of a target pixel and peripheral pixels used for determining an image region.

FIG. 3 is a block diagram showing a configuration of an image determination unit.

FIG. 4 is a block diagram showing a configuration of an error diffusion circuit.

FIG. 5 is a diagram showing error distribution pixels used for weighted average and coefficients of error distribution.

FIG. 6 is a diagram showing an example of notch noise of a read image.

[Explanation of symbols]

1 ... 1st line buffer, 2 ... 2nd line buffer, 3 ... Error diffusion circuit, 5 ... Distance setting device, 6 ...
Buffer, 7 ... Image determination unit

Claims (3)

[Claims] 1. A binarization device for image information obtained by reading an original image in which a character image and a grayscale image are mixed, wherein a plurality of pixels located around a pixel of interest have a minimum density and a maximum density. And a first average density calculating means for obtaining the average density of several pixels except for the ones, which are located on the same line as the target pixel, and which are located in the vicinity of neighboring pixels located at a predetermined distance from the target pixel. Second average density calculating means for calculating the average density of several pixels excluding the minimum density and the maximum density of a plurality of pixels, and the first and second average density calculating means respectively. Concentration gradient calculating means for obtaining the difference D between the average densities, and N1 ≦ with respect to the preset values N1 and N2.
If D ≦ N2, output data indicating a grayscale image area, and if D <N1 or D> N2, output image data indicating a character image area; and the grayscale image area from the image determination means. Is output, the halftone processing method is selected and the image information is set to 2
When data indicating the character image area is output from the image determination means, the image information is converted into 2 by a single threshold value.
A binarization processing device comprising: a binarization processing unit for binarizing. 2. The average density calculating means selects four pixels from the periphery of the target pixel and neighboring pixels so that the number of pixels excluding the pixels having the maximum and minimum densities is two, and the binarization is performed. The processing means is configured to binarize image information by using an error diffusion method as a halftone processing method when the pixel of interest is determined to be a grayscale image area. Binarization processing device. 3. The binarization processing apparatus according to claim 1, further comprising a distance setting unit that sets a distance between the pixel of interest and the neighboring pixel in advance.