JPH11136510A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a processor which can reduce the generation of snake line texture without spoiling the resolution and gradations that an error diffusing method originally has by using a random number generated by a random number generating means for the operation of a diffusion density value. SOLUTION: A random number generation part 23 is further provided which adds a random number to a High reference value HB and a Low reference value LB. In this device, the values obtained by adding the random number to the High reference value HB and Low reference value LB are inputted to a selector 19, which selects and outputs one value as a reference value B'xy for binarized error Exy calculation to a subtracter 18. Thus, the reference value B'xy for calculating the binarized error E'xy is calculated by using the random number, so a unique stripe pattern called snake line texture which is apt to be generated when an image of uniform density level with small dispersion is binarized can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus capable of reproducing an M-value gradation image having M of 3 or more into a binary gradation by using an error diffusion method.

[0002]

2. Description of the Related Art As an image processing apparatus of this type, there is an image processing apparatus capable of reproducing an input multi-value gradation image into a binary gradation by using an error diffusion method.
The error diffusion method used in this type of conventional image processing apparatus is based on the difference between the density of the pixel of interest in the read image and the display density (2
This is a method of diffusing the binarization error of a predetermined peripheral pixel into a density of a target pixel by a predetermined ratio at a predetermined rate. This is a method of calculating the display density after weighting is performed. The outline is described below.

FIG. 11 is a block diagram showing a conventional error diffusion method. The density data of each pixel constituting the multi-value gradation input image is represented by Ixy (0 ≦ Ixy ≦ 1). The density data of each pixel constituting the binarized image output in accordance with Ixy is represented by Bxy (Bxy = 0 or 1). In this case, when the correction by the weighted average error Eavexy described later is not considered, the binarization error Exy is calculated by the following equation (1).

[0004] Exy = Ixy-Bxy (1) The error diffusion method is to diffuse a binarization error to peripheral pixels. This is a method of averaging and reducing the binarization error.
1, reference numeral 1 denotes an error weighting filter 15 used for error diffusion of a binarization error of a peripheral pixel to a pixel to be processed. * Is a pixel to be processed (this pixel is particularly referred to as a target pixel), and pixel data Ixy of the target pixel
In contrast, the binarization error Exy of another processed peripheral pixel at a position belonging to the dotted frame of the error weighting filter 15 is
The weighting factors (1/16, 2/16,
3/16), and the density data of the target pixel Ixy is corrected. Therefore, actually, the binarization error Exy shown in the above equation (1) is calculated based on the corrected data I'xy of the target pixel data Ixy, and the error diffusion is performed as shown in FIG. , An error feedback loop is formed. Attention pixel data Ixy
Is the data correction value of Eavexy, the corrected data I'xy is calculated by the following equation (2).

I′xy = Ixy + Eavexy (2) Here, the correction value Eavexy is a weighted average error with respect to the input target pixel data Ixy, The weighting coefficient of the error weighting filter 15 is K _{i, j} (i is a matrix size in the main scanning direction, for example, i = 5, j
Is the matrix size in the sub-scanning direction, for example, j
= 3,), it is calculated by the following equation (3).

Eavexy = Σ (K _{i, j} × E _{i, j} ) (3) In FIG. 11, reference numeral 9 denotes density data of a target pixel and its peripheral error. Is an adder for adding the weighted average error Eavexy to the data I ′ after the correction by the equation (2).
xy is output to a comparator 10 for binarization and a subtractor 11 for calculating a binarization error.

[0007] The comparator 10 compares the corrected data I'xy obtained by the above equation (2) with a predetermined threshold value Th, and outputs the result. Binarization is performed. Thereby, the density data Bxy of the pixels constituting the binarized image is determined as shown in the following equation (4).

Bxy = 1 (I′xy ≧ Th) or 0 (I′xy <Th) (4) The comparison result of the comparator 10 is also input to the selector 12. The selector 12 selects a reference value B′xy for calculating the binarization error Exy as shown in the following equation (5) according to the input result from the comparator 10, and sends the selected value to the subtractor 11. Is output.

B′xy = HB (Bxy = 1) or LB (Bxy = 0) (5) In the above equation (5), HB (High reference value) and LB (Low reference value) Are the upper and lower limits of the output pixel density dynamic range, for example, HB = 1 and LB = 0.

Next, in the subtracter 11, the binarization error Exy
Is calculated by the following equation (6) with respect to the above equation (1).

Exy = I′xy−B′xy (6) The calculation result of the subtractor 11 is stored in the error storage unit 14 and added.
Each time the next Ixy is input to the arithmetic unit 9, the product-sum operation unit 13
The stored data is output to. The product-sum operation unit 13
Stored in the error storage unit 14 according to the force data Ixy.
Binarization error E _{i, j}Are sequentially read and the weighting coefficient K_{i, j}
And the binarization error E_{i, j}And the weighted average error Eav
xy is calculated. Then, the result is output to the adder 9.
I do.

The error diffusion method outlined above is characterized in that excellent resolution and gradation can be obtained. However, in the case of an image having a uniform density level, error diffusion is performed in a certain pattern.
It is known that there is a disadvantage that a unique stripe pattern called an ake-line (or worm-line) texture is generated.

FIG. 12 shows an example of an image processing result in which a Snake-Line texture appears remarkably by the conventional error diffusion method of this kind. The multi-level gradation input image to be processed, which has obtained the image processing result, is originally a character "
Except for 5 ″, it is composed of four regions where the pixel density is uniformly distributed, but as a result of the binarization process, Sna
A unique stripe pattern called a ke-Line texture appears, indicating that the reproducibility of the image is extremely poor.

To solve such a problem, for example,
Japanese Patent Application Laid-Open No. Hei 3-243063 discloses that a direction of a main scanning direction in a binarization process for binarizing a target pixel density of a read image and an error distribution process for diffusing and distributing an error is determined at a predetermined cycle or at random. By reversing to
It is disclosed to reduce the occurrence of e-Line texture.

In addition, Japanese Patent Laid-Open No. Hei 5-37781 discloses that the periodicity of dot connection is cut off by randomly switching the reading direction of a memory for storing error data to be distributed and distributed in accordance with the output of a random number generation circuit. It is disclosed that this reduces the occurrence of Snake-Line texture.

[0016]

However, Sna
Any of the conventional techniques for reducing the occurrence of the ke-Line texture includes a feedback element for error diffusion (a binarization threshold, a coefficient of a marginal error weighting filter,
Raster-scan direction) is controlled by a random number or the like.
Although effective in reducing the occurrence of texture, the binarization error value and the error diffusion direction calculated prior to the pixel of interest are greatly changed. For this reason, the resolution and the gradation inherent in the error diffusion method are impaired, and the processing system in the feedback loop is further complicated.

The present invention has been conceived in view of the above circumstances, and has as its object to reduce the generation of Snake-Line texture without impairing the resolution and gradation inherent in the error diffusion method as much as possible. An object of the present invention is to provide an image processing device.

[0018]

According to the first aspect of the present invention, there is provided an image processing apparatus capable of reproducing an M-value gradation image in which M is 3 or more into a binary gradation by using an error diffusion method. Diffusion density value calculation means for calculating a diffusion density value to be diffused to a peripheral pixel of a target pixel which is a binary tone reproduction target among the pixels constituting the M-value tone image; random number generation means for generating a random number; Wherein the diffusion density value calculating means includes: a density value of the target pixel before binary tone reproduction; a reference value determined on the basis of a tone of the target pixel after binary tone reproduction; The diffusion density value can be calculated using a random number generated by a random number generation means.

According to the first aspect of the present invention, since a random number generated by the random number generating means is used in the calculation of the diffusion density value, an image of a uniform density level with small variance is processed. In addition, it is possible to prevent the error diffusion from being performed in a certain pattern, and the Snake-Li
Ne texture can be reduced.

According to a second aspect of the present invention, in addition to the configuration of the first aspect, the diffusion density value calculating means includes a random number adding means capable of adding the random number to the reference value,
The diffusion density value is calculated by calculating a difference between a density value of the pixel of interest before binary tone reproduction and an addition result of the random number adding means.

According to the second aspect of the present invention, the diffusion density value is calculated by calculating the difference between the density value of the target pixel before binary tone reproduction and the addition result of the random number addition means. .

According to a third aspect of the present invention, in addition to the configuration of the first aspect, a pixel density of a predetermined pixel area including the target pixel in a pixel area constituting the M-value gradation image is provided. A reference value setting means for setting the magnitude of the reference value based on the reference value.

According to the third aspect of the present invention, the magnitude of the reference value is set based on the pixel density of a predetermined pixel area including the target pixel among the pixel areas constituting the M-value gradation image. Therefore, a gamma characteristic according to the binarization target area can be further obtained.

According to a fourth aspect of the present invention, in addition to the configuration of the first aspect, a pixel density of a predetermined pixel area including the target pixel in the pixel area forming the M-value gradation image is provided. Based on the random number, determining whether or not to calculate the diffusion density value using the random number.The diffusion density value calculation means should calculate the diffusion density value using the random number. The diffusion density value is calculated using the random number, on condition that the determination of the effect is made by the determination means.

According to the fourth aspect of the present invention, on the condition that the determination unit determines that the diffusion density value should be calculated using the random number, the diffusion density value can be calculated using the random number. Since the value is calculated, the calculation of the diffusion density value using the random number is performed only in a necessary area.

According to a fifth aspect of the present invention, in addition to the configuration of the fourth aspect of the present invention, the determining means performs the determination based on a maximum value and a minimum value of the pixel density in the pixel area. It is characterized by the following.

According to the fifth aspect of the present invention, since the determination is performed based on the maximum and minimum values of the pixel density in the pixel area, the determination process is facilitated.

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing apparatus according to one embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram for explaining the configuration of the image processing apparatus 1 and the outline of processing. The image processing apparatus 1 includes an MPU 2, an image input device 3 including a photoelectric conversion element such as a CCD and a drive system for scanning the same, an A / D conversion device 4, a Log conversion device 5, a sharpness correction device (MTF correction device) 6 , A gamma correction device 7, an image binarization device 8, and an image recording device 9.

The image input device 3 reads a mixed document composed of, for example, a continuous tone image, a line drawing, and the like, and generates a sampled analog signal. The A / D conversion device 4 quantizes the sampled analog signal generated by the image input device 3 as continuous tone reflectance data in which one pixel has a value of, for example, 8 bits (256 tones). The Log converter 5 performs logarithmic conversion of the quantized continuous tone reflectance data to calculate 8-bit continuous tone density data. The sharpness correction device (MTF correction device) 6 performs sharpness correction of an 8-bit continuous tone density data image using a digital filter such as a Laplacian filter, for example. The gamma correction device 7 performs gamma correction in order to correct a difference in a gradation curve between the image input device 3 and the image recording device 9 and realize a desired gamma characteristic of the image processing device 1 as a whole. This gamma correction is
For example, using a RAM that functions as an LUT (look-up table) of about 256 words and 8 bits,
This is performed by setting non-linear gamma correction data according to 2. The image binarization device 8 converts the gamma-corrected 8-bit continuous tone density data into 1-bit binary data corresponding to light and dark by using an error diffusion binarization method described later. The converted 1-bit binary data is printed on a predetermined recording medium by an image recording device 9 such as an electrophotographic printer or an ink jet printer.

FIG. 2 is a block diagram showing a first embodiment of the error diffusion binarization process performed by the image processing apparatus 1. FIG. 3 is a flowchart showing a processing procedure of the error diffusion binarization processing of FIG. The processing contents of the error diffusion binarization processing according to the first embodiment will be described below in detail with reference to FIGS.

Referring to FIG. 2, the first embodiment differs from the block diagram of FIG. 11 described as the prior art in that a high reference value HB and a low reference value L
The point is that a random number generator 23 for adding a random number to B is further provided. A value obtained by adding a random number to the High reference value HB and the Low reference value LB is input to the selector 19, and one of the values is converted to a binarization error Exy.
The subtractor 1 is selected as a reference value B'xy for calculation.
8 will be output.

The processing will be described with reference to the flowchart of FIG. 3. First, the target pixel data Ixy is added to the adder 1.
6 is determined (S1). The target pixel data Ixy is one of the pixels constituting the above-described gamma-corrected 8-bit continuous tone density data, and is density data of a pixel to be binarized.

If there is no input of the pixel of interest, it is determined whether or not the processing has been completed, that is, whether or not the binarization processing has been completed for all the pixels of the multi-value gradation input image (S13). .

Then, as long as the binarization processing has not been completed for all the pixels, the processing returns to S1 again.

When the target pixel is input, the weighted average error Eavexy to be added to the density data of the target pixel
Is calculated (S2), and this weighted average error Eavex is calculated.
y is added to Ixy to obtain I'xy (S3). E
The avexy is a value calculated by a process using the error weighting filter 22, as shown in FIG. 2, and is calculated by feeding back the input value.

Next, in the comparator 17, this I'xy
Is greater than or equal to the threshold value Th of the binarization process (S4). If I'xy is greater than Th, the binarized output pixel density data Bxy is set to 1 (S5). On the other hand, if it is equal to or smaller than the threshold Th, the output pixel density data Bxy is set to 0 (S6). The set Bxy is also output to the selector 19.

Next, the random number generator 23 outputs a random number ZRand
(Randmax) is extracted (S7). Random number ZRa
The range of possible values of nd (hereinafter abbreviated as random number Z) can be appropriately determined within a range that does not impair the image quality of an image obtained by the error diffusion method. It may be about 20%. Now, assuming that the dynamic range is 0 to 1,
The value of the random number Z ranges from 0.1 to 0.2.

Next, Bxy set in S5 or S6
Is determined to be 1 (S8). When Bxy is 1, the value obtained by adding the random number Z to the High reference value HB is selected by the selector 19, and is output to the subtractor 18 as the reference value B'xy (S9). On the other hand, when Bxy is not 1, that is, Bxy = 0, and the low reference value L
The value obtained by adding the random number Z to B is selected by the selector 19,
The reference value B'xy is output to the subtractor 18 (S1
0).

Next, the corrected pixel data I'xy and 2 '
A binarization error Exy is calculated from the reference value B'xy obtained from the binarization result and stored in the error storage unit 21 (S1).
1).

Next, the target pixel position is updated (S1).
2) If the binarization process has not been completed for all the pixels, the process returns to S1 again (S13), and the above process is repeated.

As described above, in the first embodiment, the reference value B'xy for calculating the binarization error Exy is used.
Is calculated using the random number Z, so that a unique stripe pattern called a snake line texture, which is likely to occur particularly when binarizing an image having a uniform density level with small variance, is achieved. be able to. Further, since the processing is performed such that only the binarization error Exy is changed at random using a random number without changing the threshold value Th in the binarization processing, any one of the error diffusion feedback elements is used. As compared with the case where is controlled by random numbers, it is possible to prevent as much as possible the deterioration of the excellent gradation expression ability, which is a feature of the error diffusion method.

FIG. 4 is a diagram showing an example of an image processed and output by the error diffusion binarization method shown in this embodiment.
The multi-value gradation input image to be processed is the same as the multi-value gradation input image from which the processing result shown in FIG. 12 is obtained. As is apparent from the drawing, as a result of processing the same multi-valued gradation input image, in FIG. 4, the snake line texture is compared with the image processed and output by the conventional error diffusion binarization method (FIG. 12). It can be seen that the uniform pixel distribution is more faithfully reproduced. On the other hand, for the character "5",
Although error dispersion was performed using random numbers, FIG.
It can be seen that the line drawing is clearly reproduced without inferior to the reproducibility of the image as compared with FIG.

FIG. 5 is a block diagram showing a second embodiment of the error diffusion binarization process performed by the image processing apparatus 1. The second embodiment differs from the block diagram of FIG. 2 described as the first embodiment in that an area determining unit 24 to which target pixel data Ixy is input is further provided.

FIG. 6 shows the structure of the area determining section 24. The area determination unit 24 includes a line memory 27, a maximum / minimum value detection filter 28, and LUTs (look-up tables) 29 and 30. The input target pixel data Ixy is sequentially stored by the line memory 27. The line memory 27 can store four lines.
Four are provided. Then, the new target pixel data I
Each time xy is input, 5 including the target pixel * at the center
25 pixels from a to x in the x5 local area are extracted, and the maximum / minimum value detection filter 28 determines the “maximum value”, “minimum value”, and “maximum value−minimum value” of the pixel density of 25 pixels. Are respectively detected. The “maximum value” and “maximum value−minimum value” of the 25 pixels are output to the LUT 29, and the “minimum value” and “maximum value−minimum value” are output to the LUT 30. The LUT 29 is a maximum / minimum value detection filter 28
The high reference value HB corresponding to the value input from the table is selected from the table data stored therein, and an adder 25 is selected.
Output to The LUT 30 has a low reference value L corresponding to a value input from the maximum value / minimum value detection filter 28.
B is selected from the internally stored table data and output to the adder 26.

FIG. 7 is a diagram illustrating a relationship between the “maximum value−minimum value” input to the LUT 29 and the High reference value HB selected by the LUT 29. FIG. 8 shows “maximum value−minimum value” input to the LUT 30.
And a diagram illustrating a relationship between the low reference value LB selected by the LUT 30 and the low reference value LB. In these figures, the vertical axis represents the High reference value HB (Max) or the Low reference value LB.
(Min), and the horizontal axis indicates “maximum value−minimum value” (Max
−Min). In addition, the gradation of the input multi-valued gradation image is set to 8 bits (0 to 255).

As the value of “maximum value−minimum value” becomes larger, the local area needs to be binarized so that the characteristics as a character area, for example, appear more strongly. Is smaller, the local area needs to be binarized so that the characteristics as a photographic area, for example, appear more strongly. If the tables of the LUTs 29 and 30 are set so that the High reference value HB and the Low reference value LB are changed and selected according to the relationships shown in FIGS. 7 and 8, a gamma characteristic suitable for a character area or a character area can be obtained. Can be.

Referring to FIG. 5 again, the High reference value HB and the Low reference value LB set by the area determination unit 24 are
These are input to the adders 25 and 26, respectively. Thereafter, the random number generated by the random number generator 23 is added to the High reference value HB and the Low reference value LB and input to the selector 19 as in the first embodiment. Since the following processing is the same as that of the first embodiment, the detailed description of the subsequent block diagram is omitted here.

As described above, according to the second embodiment, for example, a gamma characteristic suitable for a character area or a photograph area can be obtained, and the high reference value H
B, the Snake-Line texture can be reduced by adding a random number to the Low reference value LB as in the first embodiment. Therefore, for example, when the original multi-valued gradation image from which the output image of FIG. 4 is obtained is processed in the second embodiment, the second embodiment is more character-readable than the first embodiment. It is expected that the reproducibility of the area “5” will be high.

FIG. 9 is a block diagram showing a third embodiment of the error diffusion binarization process performed by the image processing apparatus 1. Note that, in FIG. 9, a portion having the same configuration as that of FIG. 5 of the second embodiment is not illustrated. Specifically, the random number generator 23 shown in FIG.
A configuration in which the components including the adder 25 and the adder 26 are replaced with the configuration of FIG. 9 is an overall block diagram according to the third embodiment.

Compared to the block diagram of FIG. 5 described as the second embodiment, the third embodiment is operatively different from the block diagram of the pixel of the 5 × 5 local area including the target pixel * at the center. In accordance with the “maximum value−minimum value” of the density, the selector 19
(See FIG. 5) High reference values HB and Lo output to
The point is that it is determined whether to add a random number to the w reference value LB. That is, unlike the second embodiment or the first embodiment in which a random number is always added to the High reference value HB and the Low reference value LB, in the third embodiment, it is necessary to add a random number in accordance with the state of the 5 × 5 local area. Sex (Snake-
The presence or absence of the occurrence of Line texture) is determined.

The error diffusion binarization processing according to the third embodiment will be described with reference to FIGS. 9 and 10 and FIG. 5. First, the pixel data of interest Ixy is added to the adder 1.
6 is determined (S21). If the pixel data of interest is input, the weighted average error Eavexy is determined.
(S22) and I′xy (S23). Next, I'xy is compared with the threshold value Th1 of the binarization processing (S4), and output pixel density data Bxy is calculated (S25, S26).

Next, the area discriminating section 24 calculates the "maximum value-minimum value" (Max-Min) of the pixel density of the pixels of the 5 * 5 local area including the target pixel * at the center (S2).
7). The area determining unit 24 has a maximum / minimum value detection filter 28 and a line memory 27 similar to those of the second embodiment. The maximum / minimum value detection filter 28 determines the value of S27
Detects the “maximum value−minimum value” of the pixels in the 5 × 5 local area. Note that, unlike the second embodiment, the area determination unit 24 does not include the LUTs 29 and 30.

Next, "maximum value-minimum value" (Max-Mi
n) is smaller than the threshold value Th2.
Is determined. This Th2 is whether or not the 5 × 5 local area is an area having a uniform density distribution characteristic with a particularly small variance, in other words, when the binarization processing is performed without using the random number Z, the Snake-Line This is a threshold value for determining whether or not the area is where a texture can occur.
If the “maximum value−minimum value” is smaller than the threshold value Th2, the 5 × 5 local region is considered to be a region having a uniform density distribution characteristic with a particularly small variance.
From 1, a signal indicating this is output to the selectors 32 and 33.

The selector 32 changes the High reference value HB to the random number generator 23 according to the signal input from the comparator 31.
Is selected, and a High reference value HB is output to the selector 19 (see FIG. 5). On the other hand, the selector 33 selects whether or not to add the random number Z input from the random number generator 23 to the Low reference value LB according to the signal input from the comparator 31 and selects the selector 19 (see FIG. 5). Output the Low reference value LB.

Therefore, if YES is determined in S28, the random numbers Z are added to the High reference value HB and the Low reference value LB by the selectors 32 and 33 (S29), and the addition result is the High reference value HB and the Low reference value. The data is output to the selector 19 as LB (S30). On the other hand, S2
If NO is determined in Step 8, the High reference value HB and the Low reference value LB are output from the selectors 32 and 33 to the selector 19 without any addition of the random number Z (S30).

The range of the value of the random number Z is, for example, 10 pixels of the dynamic range of the pixel as in the first embodiment.
% To about 20%.

Next, according to the value of the output pixel density data Bxy, the reference value B'xy for calculating the binarization error is set to Hi.
gh reference value HB or Low reference value LB (S
31, S32, S33), a binarization error Exy is calculated (S34). Thereafter, the target pixel position is updated (S3
5) The above processing is repeated until the binarization processing is completed for all the pixels (S36).

As described above, according to the third embodiment, the necessity of random number addition (the possibility of generation of a Snake-Line texture) is determined according to the state of the 5 × 5 local area. When the 5 × 5 local area is, for example, a character area, no random number is added. Therefore, it is possible to prevent the inconvenience that the random number addition processing is performed to the area where the possibility of the occurrence of the Snake-Line texture is low and the image quality is rather deteriorated. Further, it is possible to prevent the random number addition processing from being performed wastefully and lowering the speed of the binarization processing. This has an effective effect particularly in the error diffusion method having a difficult processing speed.

Next, modifications of the above-described embodiments will be listed. (1) The random number Z is added to the High reference value HB and the Low reference value LB, but may be configured to be subtracted from the High reference value HB and the Low reference value LB.

(2) As a procedure for calculating the binarization error, the random number Z is added to the High reference value HB or the Low reference value LB, and then this value is subtracted from I'xy. , I′xy after subtracting the random number Z,
The High reference value HB or the Low reference value LB may be subtracted. In this case, for example, in FIG.
A subtractor may be provided between the subtractor 18 and the random number generator 26 to output the random number Z to the subtractor.

(3) The random number generator 23 may be configured to generate different random numbers for the High reference value HB and the Low reference value LB. In this case, the range of random numbers may be changed with the same range of random numbers, or the ranges of random numbers may be different.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image processing apparatus 1.

FIG. 2 is a block diagram for explaining control processing of the image processing apparatus 1 (first embodiment).

FIG. 3 is a flowchart illustrating a control process of the image processing apparatus 1 (first embodiment).

FIG. 4 is a diagram illustrating an example in which Sna is performed by binarization processing performed by the image processing apparatus 1.
FIG. 8 is a diagram illustrating an example of an output image in which ke-Line texture is reduced.

FIG. 5 is a block diagram for explaining control processing of the image processing apparatus 1 (second embodiment).

FIG. 6 is a block diagram for explaining control processing of the image processing apparatus 1 (second embodiment).

FIG. 7 shows a High reference value H selected by the image processing apparatus 1.
It is a figure which illustrates the relationship between B (Max) and "maximum value-minimum value" of the pixel of a 5x5 local area.

FIG. 8 is a low reference value LB selected by the image processing apparatus 1.
It is a figure which illustrates the relationship between (Min) and "maximum value-minimum value" of the pixel of a 5x5 local area.

FIG. 9 is a block diagram for explaining control processing of the image processing apparatus 1 (third embodiment).

FIG. 10 is a flowchart illustrating a control process of the image processing apparatus 1 (third embodiment).

FIG. 11 is a block diagram for explaining control processing of a conventional image processing apparatus.

FIG. 12 is a diagram showing a Snake-Line texture that appears conspicuously by the binarization processing by the conventional image processing apparatus.

[Explanation of symbols]

 Reference Signs List 1 image processing device 8 image binarization device 17 comparator 19 selector 21 error storage unit 22 error weighting filter 23 random number generator 24 area discriminating unit 27 line memory 28 maximum / minimum value detection filter 29, 30 LUT (lookup table) ) 31 Comparator 32, 33 Selector

Claims (5)

[Claims] 1. An image processing apparatus capable of reproducing an M-value gradation image having an M of 3 or more into a binary gradation by using an error diffusion method, wherein two of pixels constituting the M-value gradation image are included. A diffusion density value calculating means for calculating a diffusion density value to be diffused to a peripheral pixel of the pixel of interest to be reproduced, and a random number generating means for generating a random number; Using a density value of the pixel of interest before tone reproduction, a reference value determined on the condition of the tone of the pixel of interest after binary tone reproduction, and a random number generated by the random number generating means, An image processing apparatus capable of calculating a diffusion density value. 2. The method according to claim 2, wherein the diffusion density value calculation means includes a random number addition means capable of adding the random number to the reference value, and adds a density value of the target pixel before binary tone reproduction and the random number addition means. The image processing apparatus according to claim 1, wherein the diffusion density value is calculated by calculating a difference from the result. 3. A reference value setting means for setting a magnitude of the reference value based on a pixel density of a predetermined pixel region including the target pixel in a pixel region constituting the M-value gradation image. The image processing apparatus according to claim 1, wherein: 4. A method for calculating whether or not to calculate the diffusion density value using the random number, based on a pixel density of a predetermined pixel area including the target pixel in a pixel area forming the M-value gradation image. Determining means for calculating the diffusion density value using the random number, provided that the determination means has determined that the diffusion density value should be calculated using the random number. The image processing apparatus according to claim 1, wherein the diffusion density value is calculated by the calculation. 5. The image processing apparatus according to claim 4, wherein the determination unit performs the determination based on a maximum value and a minimum value of a pixel density in the pixel area.