JP2003274172A - Image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a high image quality processing system which suppresses the occurrence of a chained unique pattern in a low density area and reduces a blurred contour in an error diffusing method. <P>SOLUTION: As for an input image signal with little density change, a threshold matrix and a weighting factor matrix for making a dot generation locally uniform are used to eliminate biased dot generation and to prevent a unique pattern (texture) in a low density area. Besides, as for an input image signal with a large density change, more quantization errors are distributed to a pixel large in the density value of a distribution destination pixel to which quantization errors are distributed to make generate the dot easily. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing system using an error diffusion method as a method for converting a multi-tone input image signal used in a printer or a copying machine into a low tone image signal.

[0002]

2. Description of the Related Art An error diffusion method (Robert Floyd) is one of the pseudo gradation expression methods for recording a halftone image.
and Louis Steinberg, "An A
dap Algorithm for Spa
Tial Gray Scale ”, Proc.SI
D, Vol 17, No 2, pp. 75-77, 197
6) is known. The error diffusion method is a quantization process that performs density correction so that the density error between the input image and the output image is minimized, and the apparent resolution when printing a full-color high-definition image with a printer that can only record in binary. It is an extremely effective method that can record a good image with less deterioration of gradation expression.

A conventional error diffusion method will be described with reference to FIG. There are m pixels in the main scanning direction, n pixels in the sub-scanning direction, and m pixels in the main scanning direction and n pixels in the sub-scanning direction from the input image signal S having a density of 8 bits (0 to 255) per pixel. An output image signal O having a density of 1 bit (0 or 1) per pixel is generated. First, the density values of the x-th pixel from the left and the y-th pixel from the top of the input image signal S are S (x,
y), when the density values of the x-th pixel from the left and the y-th pixel from the top of the output image signal O are represented by O (x, y), the output image signal O
The density value O (x, y) and the quantization error ERR generated between the input image signal S and the output image signal O are calculated by the following equations 1 and 2.

When S (x, y) ≧ TH, O (x, y) = 1, ERR = S (x, y) −255 ... [Equation 1] When S (x, y) <TH O (x, y) = 0, ERR = S (x, y) −0 [Equation 2] Here, TH is a binarization threshold value of the maximum density value that can be expressed by the input image signal S. Half, that is, TH = 128 is used.

Next, the quantization error ERR generated between the input image signal S and the output image signal O is distributed to the peripheral pixels near the target pixel. FIG. 2 is a weighting coefficient matrix representing a distribution ratio when the quantization error ERR is distributed to the peripheral pixels. The quantization error ERR is calculated according to the weighting coefficient of FIG.
Are distributed to the peripheral pixels by the following equations 3 to 6.

S (x + 1, y) = S (x + 1, y) + (7 / (7 + 3 + 5 + 1)) × ERR ... [Formula 3] S (x-1, y + 1) = S (x-1, y + 1) + (3 / (7 + 3 + 5 + 1)) × ERR ... [Equation 4] S (x, y + 1) = S (x, y + 1) + (5 / (7 + 3 + 5 + 1)) × ERR ... [Equation 5] S (x + 1) , Y + 1) = S (x + 1, y + 1) + (1 / (7 + 3 + 5 + 1)) × ERR ... [Equation 6] By performing the above processing on all pixels in order,
It is possible to obtain the quantized output image signal O.

However, in the error diffusion method, a pattern (texture) having a unique structure with respect to image data having a constant density value.
However, there are drawbacks such as occurrence of white spots, delay of generation of dots for expressing low density in the rising portion of the low density region, and white spots. As a method of avoiding this problem, a method of adding random noise to the input image, a method of adding random noise to the threshold used in the quantization process, or
Threshold optimization error diffusion method that changes the threshold of quantization processing according to the density value of the input image (Shigeaki Sumiya, "High image quality technology of error diffusion method", Journal of the Photographic Society of Japan, Volume 60, No. 6, p.
p. 353-356, 1997) and the like have been proposed.

However, although these conventional methods improve the above-mentioned drawbacks, the dot dispersibility in the low density region is not sufficient, and a chain-like peculiar pattern as shown in FIG. 3 is generated. This is because the dot generation rate on the recording medium is uniform in a wide range, but locally uneven. At the same time, there is a problem that the contour is blurred in the low density region. This is because the accumulation rate of the quantization error is slow in the low density region and the response of dot generation is delayed. These cause image quality deterioration in the low density region. on the other hand,
In the medium density region and the high density region, even if the dots are unevenly generated, since there are many high frequency components in terms of spatial frequency, there is almost no problem in image quality for human eyes.

[0009]

SUMMARY OF THE INVENTION The present invention, among the drawbacks of the prior art, is an error diffusion method which suppresses the generation of chain-like peculiar patterns in the low density region and reduces the blurring of the contour. It is intended to provide a high quality image processing method.

[0010]

According to the invention described in claim 1, error diffusion for quantizing a multi-valued input image signal into an output image signal having a gradation number smaller than that of the input image signal. In the process, as a process to be performed on each pixel, a threshold value is extracted from a threshold matrix prepared in advance according to the position of the target pixel, and the density value of the target pixel is compared with the threshold value to obtain the target pixel of the output image signal. A target pixel density determining step of determining a density value of a position; a density value of the target pixel in the input image signal, and a density value of the target pixel in the output image signal quantized in the target pixel density determining step. In order to distribute the quantization error generated between them to the surrounding pixels that have not been quantized yet in the vicinity of the target pixel, the distribution destination pixel position and the distribution coefficient are stored in advance. A distribution ratio calculating step of calculating a distribution ratio using the lix and the density value of the input image signal of peripheral pixels of a distribution destination; and an error distribution for distributing the quantization error based on the distribution ratio calculated by the distribution ratio calculating step. It is achieved by providing an image processing method having steps.

Here, in the target pixel density determining step, for the target pixel of the input image signal, a threshold value is extracted from a threshold matrix prepared in advance according to the position of the target pixel, and the density value of the target pixel is calculated. The density value of the output image signal for the pixel of interest is determined from the result of comparison with the extracted threshold value.

On the other hand, in the distribution ratio calculating step, between the density value of the target pixel in the input image signal and the density value of the target pixel in the output image signal quantized in the target pixel density determining step. The generated quantization error is distributed to the peripheral pixels in the vicinity of the target pixel which have not been quantized yet. The position of the distribution destination pixel at that time is determined by a weighting coefficient matrix prepared in advance. Further, the distribution ratio of the quantization error to the distribution destination peripheral pixels is calculated using the distribution coefficient stored in the weighting coefficient matrix and the density value of the input image signal in the distribution destination peripheral pixels.

Further, in the error distribution step, the quantization error is distributed to the distribution destination pixels defined by the weighting coefficient matrix based on the distribution ratio calculated in the distribution ratio calculation step.

In the distribution ratio calculating step, the density value of the input image signal in the distribution destination peripheral pixels used for calculating the distribution ratio is the density value after the quantization error is added in the error distributing step. That is, the density value of the original input image signal is used instead of the density value after the error diffusion processing. For that purpose, the original input image signal is copied in advance as a duplicated image signal, and the original input image signal is used in the distribution ratio calculation step, and the duplicated image signal is used in the target pixel density determination step and the error distribution step. It is necessary.

The input image signal is quantized by the error diffusion process by sequentially performing the above process for all the pixels. In the quantization at that time, the distribution destination and distribution ratio of the quantization error can be controlled by the density value in the input image signal of the distribution destination pixel. That is, it is possible to prevent the blurring of the contour in the low density region by distributing the quantization error more to the pixels having a large density value in the input image signal among the peripheral pixels.

A second aspect of the present invention embodies a threshold matrix used in the step of determining the target pixel density according to the first aspect of the present invention, wherein the threshold matrix is
By arranging the threshold value by the method of the blue noise mask, the generation of dots in the low density region becomes like blue noise, and the generation of a pattern (texture) having a peculiar structure can be prevented. As a method of blue noise mask, (Robert
Ulichney, “Thevoid-and-cl
user method for dithera
"ray generation", SPIE, Vo
l. 1913, pp. 332-343) can be used. However, the method is not limited to this, and may be created by another known method.

The invention according to claim 3 is the same as claim 1, 2 or 3.
To realize the distribution ratio calculation step of the invention of
When the density difference between the maximum density value and the minimum density value of the input image signal for all the distribution destination peripheral pixels determined by the weighting coefficient matrix is larger than a predetermined reference density difference. , The distribution ratio for each peripheral pixel is the product of the distribution coefficient of the weighting coefficient matrix and the density value of the input image signal of the peripheral pixel, and when the density difference is smaller than the reference density difference, Let the distribution ratio be the distribution coefficient of the weighting coefficient matrix.
That is, in the input image signal with a small density change, by using the threshold value matrix and the weighting coefficient matrix for making the dot generation locally uniform, the bias of dot generation is eliminated and a unique structure in the low density region is obtained. It is possible to prevent the occurrence of the pattern (texture), and at the same time, in the case of an input image signal with a large change in density, check the density value of the distribution destination pixel that distributes the quantization error, and distribute more quantization error to the pixels with large density values. By facilitating the formation of dots, it is possible to reduce the blurring of the contour in the low density region and to improve the image quality in the low density region.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below based on examples with reference to FIGS. 2 and 4 to 14. In the present embodiment, for convenience of explanation, the input image signal has 8 bits (0 to 25) per pixel for convenience.
5) An image signal having a density is used and an output image signal after processing is a binary signal of 1 bit per pixel, but the present invention is not limited to this.

FIG. 4 shows an input image signal S, a duplicate image signal C thereof, and an output image signal O after the error diffusion processing of the present invention.
Is represented. Each of these images has a size of m pixels in the main scanning direction and n pixels in the sub scanning direction. The density values of the xth pixel in the main scanning direction and the yth pixel in the subscanning direction are S pixels in the input image signal. (X, y), C (x, y) in the duplicate image signal, and O in the output image signal after the error diffusion processing.
It is expressed by (x, y). Here, the input image signal S
Is used for calculating the distribution ratio of the quantization error in the distribution ratio calculation step described later. In addition, the duplicate image signal C is compared with a threshold value in a pixel-of-interest density determination step described later,
It is used for distribution of quantization error in the error distribution step, and the quantized binary image is stored in the output image signal O.

FIG. 5 is a threshold matrix used for locally uniforming dot generation. The threshold matrix of this embodiment uses a distributed dither matrix of 16 × 16 pixel size, but the present invention is not limited to this.
The image processing method of the present invention will be described below with reference to FIGS. First, in step S100 of FIG. 6, the input image signal S is copied to the duplicate image signal C. Next, in order to quantize the input image signal, step S1 is performed for all pixels.
The target pixel density determination step of 30, the distribution ratio calculation step of step S140, and the error distribution step of step S150 are executed. The iterative process for that is step S110 and step S120. In step S110, the iterative process is performed n times which is the image size in the sub-scanning direction, and in step S120, the iterative process is performed m times which is the image size in the main scanning direction.

Next, the target pixel density determining step of step S130 of FIG. 6 will be described with reference to FIG. First, the threshold value T for the target pixel is obtained based on the threshold value matrix of FIG. 5 (step S200). For example, when the coordinates of the pixel of interest are x in the main scanning direction and y in the sub scanning direction, the threshold value T is as shown in FIG.
It is obtained by B (x% 16, y% 16) of the threshold matrix of. Here, x% 16 represents the remainder when x is divided by 16. For example, when the coordinates of the pixel of interest are 20 in the main scanning direction and 40 in the sub scanning direction, the threshold T is B
(20% 16, 40% 16) = B (4,8) = 12.

Next, the density value C (x, y) of the target pixel of the duplicate image signal C is compared with the threshold value T (step S21).
0). If the density value C of the pixel of interest of the duplicated image signal C
If (x, y) is greater than or equal to the threshold value T, the density value O (x, y) of the pixel of interest of the output image signal O is set to 1 in order to generate dots in the pixel of interest (step S220), and The value obtained by subtracting 255 from the density value C (x, y) of the pixel of interest of the duplicated image signal C is substituted for the quantization error ERR at that time (step S240). On the other hand, when the density value C (x, y) of the target pixel of the duplicate image signal C is smaller than the threshold value T, the density value O (x, y) is set to 0 (step S
230), the duplicated image signal C is added to the quantization error ERR at that time.
The density value C (x, y) of the target pixel of is substituted (step S250). This completes the target pixel density determination step. At this point, the density value of the output image signal O for the pixel of interest and the quantization error ERR at that time are determined.

Next, the distribution ratio calculation step of step S140 of FIG. 6 will be described with reference to FIGS. 2, 8 and 9. In this step, the quantization error ER generated in the target pixel density determination step
This is a step of calculating a distribution destination pixel and its distribution ratio W when R is distributed to surrounding pixels which have not been quantized yet in the vicinity of the pixel of interest.

First, FIGS. 2 and 8 will be described. FIG. 2 is a weighting coefficient matrix used in this embodiment. This weighting coefficient matrix represents the distribution destination pixels of the quantization error ERR and the distribution coefficient thereof. That is, in the weighting coefficient matrix of FIG. 2, when the coordinates of the target pixel are (x, y), the distribution destination pixels are the four surrounding pixels, and the coordinates are (x
+1, y), (x-1, y + 1), (x, y + 1),
(X + 1, y + 1), and the distribution ratio for each distribution destination pixel is 7: 3: 5: 1. 8 is a distribution ratio matrix showing distribution destination pixels of the quantization error ERR and distribution ratio W thereof. The distribution destination pixels are determined by the weighting coefficient matrix of FIG. 2, and the coordinates thereof are those of the weighting coefficient matrix of FIG. As with the distribution destination pixel, (x +
1, y), (x-1, y + 1), (x, y + 1), (x
+1, y + 1). Further, the distribution ratio W for each distribution destination pixel is W (x + 1, y), W (x-1, y), respectively.
+1), W (x, y + 1), and W (x + 1, y + 1).

Next, the distribution ratio calculation step will be described with reference to the flow chart of FIG. First, in step S300, the density values of all the distribution destination pixels are checked to obtain the maximum density value Smax and the minimum density value Smin. That is, the coordinates of the distribution destination pixel are (x +
1, y), (x-1, y + 1), (x, y + 1), (x
+1, y + 1), each density value S (x +
1, y), S (x-1, y + 1), S (x, y + 1),
The maximum density value of S (x + 1, y + 1) is Smax,
The minimum density value is Smin. Next, step S31
At 0, the difference between the maximum density value Smax and the minimum density value Smin (Smax-Smin) is compared with the reference density difference Ds. Here, the reference density value Ds is a predetermined reference value for determining the degree of variation in the density value of the distribution destination pixels. In the experiment by the inventor, about 5% of the number of gradations of the input image signal showed a good result, so in the present embodiment, the number of gradations 256 of the input image signal is multiplied by 0.05 to be 13. It is not limited to. Next, in the comparison in step S310, when the difference (Smax-Smin) between the maximum density value Smax and the minimum density value Smin is larger than the reference density difference Ds, the distribution ratio W of the quantization error ERR to the distribution destination pixels is W.
Is a value obtained by multiplying the distribution coefficient of the weighting coefficient matrix of FIG. 2 by the density value of the distribution destination pixel as shown in Expressions 7 to 10 (step S320).

W (x + 1, y) = 7 × S (x + 1, y) ... [Equation 7] W (x−1, y + 1) = 3 × S (x−1, y + 1) ... [Equation 8] ] W (x, y + 1) = 5 × S (x, y + 1) ... [Equation 9] W (x + 1, y + 1) = 1 × S (x + 1, y + 1) ... [Equation 10] On the other hand, in step S310 In comparison, the maximum density value S
The difference between max and the minimum density value Smin (Smax-Smi
If n) is less than or equal to the reference density difference Ds, the quantization error ER
The distribution ratio W of R to the distribution destination pixels is the distribution coefficient of the weighting coefficient matrix of FIG. 2 as shown in Expressions 11 to 14 (step S330).

[0027] W (x + 1, y) = 7 [Equation 11] W (x-1, y + 1) = 3 [Equation 12] W (x, y + 1) = 5 [Equation 13] W (x + 1, y + 1) = 1 ... [Equation 14] This completes the distribution ratio calculation step.

Next, the error distribution process of step S150 of FIG. 6 will be described with reference to FIG. First, step S40
0, the total Ws of the distribution ratio of the quantization error ERR to all the distribution destination pixels is obtained as shown in Expression 15.

Ws = W (x + 1, y) + W (x-1, y + 1) + W (x, y + 1) + W (x + 1, y + 1) [Equation 15] Next, in Step S410, an equation from Equation 16 is obtained. As shown in 19, the quantization error ERR is distributed to the distribution destination pixels based on the distribution ratio W.

C (x + 1, y) = C (x + 1, y) + ERR × W (x + 1, y) / Ws ... [Equation 16] C (x-1, y + 1) = C (x-1, y + 1) + ERR × W (x−1, y + 1) / Ws [Equation 17] C (x, y + 1) = C (x, y + 1) + ERR × W (x, y + 1) / Ws ··· [Equation 18] C (X + 1, y + 1) = C (x + 1, y + 1) + ERR × W (x + 1, y + 1) / Ws (Equation 19) This is the end of the error distribution step, and at the same time, the end of quantization for one pixel. Is.

As described above, in the input image signal with a small density change, the threshold value matrix of FIG. 5 and the weighting coefficient matrix of FIG. 2 for locally uniforming the dot generation are used to generate the dot generation. It is possible to eliminate the bias and prevent the occurrence of a pattern (texture) with a unique structure in the low density range, and at the same time, in the case of an input image signal with a large density change, check the density value of the distribution destination pixel that distributes the quantization error, and check the density value. It is possible to blur the contour in the low density region by distributing the quantization error more to the pixels having a large number and making it easier to generate dots. Although the weighting coefficient matrix of FIG. 2 is used in this embodiment, the present invention is not limited to this. For example, the weighting coefficient matrix of FIG. 11 can be used. When the weighting coefficient matrix of FIG. 11 is used, more calculations are required, but since the quantization error is distributed over a wide range, the pattern (texture) having a more peculiar structure than the weighting coefficient matrix of FIG. 2 is used. Can be suppressed.

Next, a second embodiment of the present invention will be described with reference to FIGS. 12 and 13. The second embodiment differs from the first embodiment only in that a threshold matrix in which thresholds are arranged by the blue noise mask method is used in the target pixel density determination step. Less than,
A method of generating the blue noise mask will be described.

FIG. 12 shows a blue noise mask B to be created.
The matrix size is i in the main scanning direction and j in the sub scanning direction. Further, it is assumed that the coordinates of the threshold value of x in the main scanning direction and y in the sub scanning direction are represented by (x, y), and the threshold value of the coordinates is represented by B (x, y). Since the density value of the input image signal is 256 levels from 0 to 255, the threshold value to be arranged in the blue noise mask is 1 to 255.

Next, a method of generating a blue noise mask will be described with reference to the flowchart of FIG. First, the variables used are K, B (x, y), A (x, y), Ev.
a0 and majority are initialized in step S500. K is the density value of the blue noise mask currently created, and the initial value is K = 1. As described above, B (x, y) is the threshold value array at the coordinates (x, y) of the blue noise mask to be created, and the initial values are all 0. A (x, y) is an array indicating whether dots have already occurred at the coordinates (x, y) of the blue noise mask, and the initial values are all 0. E
va0 is a convergence evaluation value for determining the degree of dot dispersion at density 1, and the initial value is i * j. majority is a variable that indicates 1 when the number of dots already generated in the blue noise mask is larger than half the total number of pixels, and 0 when the number is small,
The initial value is 0.

Next, steps S510 to S5
At 80, a mask of density 1 (K = 1) is created. This is the initial pattern for creating masks with densities of 2 to 255. First, in step S510, dots for one density are randomly arranged in the mask. When arranging dots, 1 is substituted for A (x, y). In addition, the number of dots to be arranged should be (i × j /
255). Next, steps S520 to S570.
The dots randomly arranged in the mask are rearranged so that the spatial distribution becomes uniform. First, the expected densities of all pixels in the mask are calculated. The expected density is used to evaluate the degree of dispersion of dots, and when the coordinates of the target pixel are (x, y), the expected density of the pixel is represented by D (x, y) and calculated by Equation 20. To do.

Here, the majority in equation 20 is
The variable is 0 when the number of pixels generated in the mask is less than half of the total number of pixels, and 1 when the number of pixels is more than half. Further, u, v, and f (p.q) are calculated by Expressions 21 to 23.

The% in the expressions 21 and 22 is a symbol representing the remainder as described in the first embodiment, for example, 5% 3
= 2, 3% 5 = 3. Further, σ is a coefficient for adjustment, and in the experiment by the inventor, 1.5 showed a good result. By the way, the expected density is the sum of the peripheral pixels (u, v) existing around the target pixel (x, y), weighted according to the distance from the target pixel. f (p, q) is a weighting function according to the distance from the target pixel. Therefore, the more pixels are present around the target pixel, and the closer the distance is, the higher the expected density of the target pixel becomes.
This expected density is used in step S550 as a material for determining the degree of dot dispersion. However, majority
When the value of is 1, that is, when the number of pixels generated in the mask is more than half of the total number of pixels, the target pixel (x,
In order to calculate the expected density by searching for the peripheral pixel (u, v) around which the dot is not yet arranged in y), the more pixels exist around the target pixel, the closer the distance is. The smaller the expected density of the target pixel, the smaller. In addition, the search range of the peripheral pixels is x- (i
/ 2) to x + (i / 2), and the vertical direction is x- (j / 2) to
By setting x + (j / 2), all pixels in the mask can be considered. However, since the processing time increases as the mask size increases, it is also possible to adjust the search range according to the mask size and density.

Next, in step S530, among the pixels in which the dots are arranged, that is, the pixels of A (x, y) = 1,
The maximum expected density is calculated, the maximum expected density is Dmax, and the coordinates of the pixel are (xmax, ymax). Furthermore, the pixel where the dot is not arranged, that is, A (x,
y) = 0, the minimum expected density is obtained from the pixels, the minimum expected density is Dmin, and the coordinates of the pixel are (xmin, y
min). If the concentration K is 1, majorit
Since y = 0, the coordinates (xmax, ymax) are the most densely located dots, and the coordinates (xmin, ymi)
n) is the roughest place.

Next, in step S540, the difference between the maximum expected density Dmax and the minimum expected density Dmin is substituted into Eva. The Eva obtained here is a parameter indicating the current dispersion of dots.

Next, in step S550, this Eva is compared with the convergence evaluation value Eva0 of the degree of dot dispersion to determine whether or not the dot dispersion processing has converged. Eva is Eva
If it is less than 0, it is determined that the degree of dot dispersion is still insufficient, and in step S560, the pixel with the maximum expected density and the pixel with the minimum expected density are exchanged. That is, A (xmax, y
max) = 0 and A (xmin, ymin) = 1.
Further, in step S570, the convergence evaluation value Eva0 is set to Ev.
Substitute a for a new convergence condition, and control is returned to step S520 to continue the dot dispersion processing. On the other hand, if Eva is equal to or greater than Eva0 in step S550, it is determined that the dots are sufficiently dispersed, the process moves to step S580, and the contents of array A (x, y) are copied to B (x, y). This completes the creation of the blue noise mask of density 1.

Next, in steps S590 to S690, a blue noise mask having a density of 2 to 254 is created. Step S590 is a loop process for that purpose, and K = 2 to K = 2.
The processes of steps S600 to S690 are repeated until 54. In step S600, the gradation value K currently being processed is compared with 128 in order to check whether the number of dots generated in the mask has reached half (i × j / 2) of the total number of pixels. If K is less than 128, it is determined that the number of dots generated in the mask has not reached half of the total number of pixels (i × j / 2), and 0 is substituted for majority in step S610. Otherwise, in step S620, maj
Substitute 1 for ority. However, in this process, the number of dots generated in the mask is half the total number of pixels (i × j
It is not possible to strictly judge whether or not the value of / 2) has been reached, but the effect of this is that the processing time changes only slightly and there is no problem in mask making.

Step S630 is a loop process for creating a mask for one density, and steps S640 to S690 are repeated i × j / 255 times. As with the step S510, the number of loops is set to 25 for the total number of pixels in order to make the number of dots added in each density the same.
The value is divided by 5. Then, upon exiting the loop of step S630, the process returns to step S590 to create a mask of the next density. Next, the processing from step S640 to step S690 will be described. First, step S6
In step 40, as in step S520, expected densities of all pixels are obtained. Next, in step S650, it is determined whether or not the majority is 0. If it is 0, in step S660, the pixel where the dot is not yet arranged, that is, A (x,
The minimum expected density is obtained from the pixels of y) = 0, and the coordinates of the pixel are substituted into (xmin, ymin). Since the coordinates (xmin, ymin) are the coarsest coordinates of the dot density, dots are generated at the coordinates (xmin, ymin). Further, the density value K currently being processed is substituted into the mask. That is, A (xmin, ymin) = 1, B
(Xmin, ymin) = K (step S67)
0).

On the other hand, in step S650, majori
If ty is 1, in step S680, the maximum expected density is obtained from the pixels in which dots are not yet arranged, that is, the pixels of A (x, y) = 0, and the coordinates of the pixel are set to (x
(max, ymax). This coordinate (xmax,
ymax) is the coarsest coordinate of dot density, so
A dot is generated at the coordinates (xmax, ymax). Further, the density value K currently being processed is substituted into the mask. That is, A (xmax, ymax) = 1, B (xmax, y
max) = K (step S690).

In this way, when the repeating processing of step S630 and the repeating processing of step S590 are both completed, the masks of density 1 to density 254 are completed. After that, by substituting 255 for a pixel in which the threshold value is not yet filled, that is, a pixel of B (x, y) = 0, the mask of all densities is completed (step S70).
0).

By using the blue noise mask created by the above method as the threshold matrix, the unevenness of dots is reduced and at the same time, the feeling of roughness is eliminated, and the generation of a pattern (texture) having a unique structure can be suppressed.
FIG. 14 shows the result of quantizing a 5% density black image in this embodiment. As shown in FIG. 3, the chain-like peculiar pattern that occurred in the image quantized by the conventional method is eliminated in this embodiment.

[0046]

According to the conventional error diffusion method, the dot dispersibility in the low density region is not sufficient, and there is a problem that a pattern (texture) having a peculiar structure is generated or an outline is blurred. , For input image signals with little change in density,
By using the threshold value matrix and the weighting coefficient matrix for locally uniforming the dot generation, it is possible to eliminate the uneven dot generation and prevent the generation of a pattern (texture) having a unique structure in the low density region, For an input image signal with a large change in density, check the density value of the distribution destination pixel that distributes the quantization error, distribute more quantization error to the pixels with large density values, and make dots more likely to occur. It is possible to reduce the blurring of the contour of the image and improve the image quality in the low density region.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an image signal used in a conventional error diffusion method.

FIG. 2 is a first weighting coefficient matrix used in the conventional error diffusion method and the present invention.

FIG. 3 is a diagram showing an output result by a conventional error diffusion method.

FIG. 4 is a configuration diagram of an image signal used in the present invention.

FIG. 5 is a threshold matrix used in the present invention.

FIG. 6 is a flowchart showing the first embodiment of the present invention.

FIG. 7 is a flowchart showing a target pixel density determination process according to the first embodiment of the present invention.

FIG. 8 is a distribution ratio matrix for the first weighting coefficient matrix used in the first embodiment of the present invention.

FIG. 9 is a flowchart showing a distribution ratio calculation step of the first embodiment of the present invention.

FIG. 10 is a flowchart showing an error dispersion process according to the first embodiment of the present invention.

FIG. 11 is a second weighting coefficient matrix used in the present invention.

FIG. 12 is a blue noise mask of a second embodiment of the present invention.

FIG. 13 is a flowchart showing a blue noise mask creating method according to the second embodiment of the present invention.

FIG. 14 is a diagram showing an output result by the error diffusion method of the present invention.

[Explanation of symbols]

1 Input image signal S 2 Output image signal O 3 First weighting coefficient matrix 4 Input image signal S 5 Duplicate image signal C 6 Output image signal O 7 Threshold matrix B 8 Distribution ratio matrix W 9 Second weighting coefficient matrix 10 Blue Noise Mask B

Claims (3)

[Claims] 1. An image processing method using an error diffusion process for quantizing a multi-valued input image signal into an output image signal having a smaller number of gradations than the input image signal. A target pixel density determining step of determining a density value of the target pixel position of the output image signal by extracting a threshold value from the threshold matrix according to the target pixel position and comparing the density value of the target pixel with the threshold value; , A quantization error generated between the density value of the pixel of interest in the input image signal and the density value of the pixel of interest of the output image signal quantized in the pixel density determination step of interest,
In order to distribute to surrounding pixels that have not been quantized yet in the vicinity of the target pixel, a weighting coefficient matrix prepared in advance that stores distribution destination pixel positions and distribution coefficients, and density values of input image signals of distribution destination peripheral pixels An image processing method, comprising: a distribution ratio calculation step of calculating a distribution ratio using the above; and an error distribution step of distributing a quantization error based on the distribution ratio calculated by the distribution ratio calculation step. 2. The image processing method according to claim 1, wherein the threshold values are arranged in the threshold value matrix by a blue noise mask method. 3. In the distribution ratio calculation step, in the distribution destination peripheral pixels determined by the weighting coefficient matrix, the density difference between the maximum density value and the minimum density value of the input image signal for all the distribution destination peripheral pixels is predetermined. When the density difference is smaller than the reference density difference, the distribution ratio for each peripheral pixel is set as the product of the distribution coefficient of the weighting coefficient matrix and the density value of the input image signal of the peripheral pixel. The image processing method according to claim 1, wherein the distribution ratio for each peripheral pixel is used as a distribution coefficient of the weighting coefficient matrix.