JPH06189119A - Image processing method - Google Patents

Abstract

PURPOSE:To process a dynamic range of halftone to be handled by expanding by NXM times compared with the one in pixel unit by dividing an image targeted to process into NXM areas, and setting each of the NXM areas as one unit in processing. CONSTITUTION:Assuming area of interest in NXM(N=2, M=3) target areas to be the four picture elements in a broken line, total pixel data in the broken line are represented in A=Di, j+2+Di, j+3+Di+1, j+2+Di+1, and j+3 in the pixel area A of interest. Quantization is performed by applying binarization to the four picture elements in the area of interest by an error diffusion method.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method, and more particularly to an image processing method suitable for application to an apparatus such as a digital copying machine or a family of equipment which needs to quantize multi-valued image data. .

[0002]

2. Description of the Related Art Hitherto, various methods such as a dither method and an error diffusion method have been proposed as a method for quantizing multi-valued image data. It was to convert.

Explaining a concrete example of quantization by the error diffusion method, when the pixel of interest A is quantized by x, the error is expressed by the following equation. A- [x] = ε (1) Here, [x] is multi-valued data corresponding to the quantized data. For example, if the pixel data A is 8 bits, 0
˜255, x = 0 or 1 when the quantization is binary with 0, and [x] = 0, when x = 0.
When x = 1, [x] = 255.

The error .epsilon. Is dispersed in peripheral pixels B, C, D and E by a predetermined method in order to preserve the density, and is added to each of the peripheral pixels B, C, D and E as shown in FIG. That is, the error is ε = ε
_B ε _C ε _D ε _E , B = B + ε _B , C = C + ε _C , D
= D + ε _D , E = E + ε _E , which is added to each of the peripheral pixels. Next, the above process is performed with the B pixel as the pixel of interest, and thereafter, the C and D pixels are processed in the same manner.

[0005]

However, according to the above-mentioned conventional method, since the target pixel is set on a pixel-by-pixel basis, the dynamic range of the halftone to be handled is limited to the gradation range of one pixel at the maximum, and the dynamic range beyond that. I couldn't have a range. For example, if the multi-valued data is 8-bit pixel data, it can have a dynamic range of 256 gradations, but on the contrary, it cannot be quantized with a range of more than that.

[0006]

In order to solve the above-mentioned problems, the present invention is characterized in that (1) an area to be quantized is N × M pixels, and quantization is performed for each pixel. Further, (2) the quantization is performed by an error diffusion method, or (3) the N is used in the error diffusion method.
The feature is that the weighting of the quantization is performed in advance for the × M pixels.

[0007]

When the image data is quantized, the target image to be processed is not limited to one pixel, but N × M area pixels are targeted, and the N × M area is regarded as one unit of processing. , The halftone dynamic range to be handled is expanded by N × M times as compared with that of one pixel unit for processing. Here, N and M are integers and N × M means 2 or more.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION In the present invention, in quantization of image data, a target (target) pixel is divided into N × M areas, and each area is treated as one unit.
If the image data before quantization is 8 bits and the halftone dynamic range is 256 gradations, the dynamic range can be dramatically increased to 512 gradations only by setting N = 1 and M = 2. Therefore, a significant improvement in image quality can be achieved. The effect of this image quality improvement is that the processing is performed without changing the dynamic range of the original image data.
This is an effect when xM, but conversely, NxM is 2,
Further, in order to set the dynamic range after quantization to 256 gradations, the original image data becomes 256/2, which is 128.
This means that a gradation, that is, a resolution of 7 bits is sufficient, and there is also an effect that the load on the hard circuit such as A / D conversion can be reduced while maintaining the dynamic range of halftone.

An embodiment in the case of N = 2 and M = 2 in N × M will be described below. Each pixel data is represented by 8 bits and has a dynamic range of 256 gradations per pixel. Further, the quantization is binary, the result is expressed by 0 or 1, and the threshold of quantization is 128 which is the median value.

In FIG. 1, assuming that the target area corresponding to the N × M target area is four pixels in the broken line, the total pixel data in the broken line is A = Di, j + 2 + Di in the target pixel area A. , j + 3 + Di + 1, j + 2 + Di + 1, j + 3 (2) Similarly, assuming that the target pixel area around the target pixel area A is B, C, D, and E, similarly to A, B = Di, j + 4 + Di, j + 5 + Di + 1, j + 4 + Di + 1, j +5 (3) C = Di + 2, j + 4 + Di + 2, j + 5 + Di + 3, j + 4 + Di + 3, j + 5 ... (4) D = Di + 2, j + 2 + Di + 2, j + 3 + Di + 3, j + 2 + Di + 3, j + 3 (5) E = Di + 2, j + Di + 2, j + 1 + Di + 3, j + Di + 3, j + 1 (6) Here, the range that each pixel area can take from the above equation is 0 to 255 for each pixel,
It means that the value of X255 × 4 = 1020 can be taken.

Quantization is performed by dividing four pixels in the area of interest into two pixels.
It is calculated by the following procedure by expressing it as a value (0 or 1). When the value of A is 0 to 128, it is assumed that all four pixels are 0 as shown in FIGS.
(A)), assuming that the quantization pattern is 0, the quantization error at this time is A. Note that these relationships are shown in a graph in FIG. 2 and in a table in FIG. 3 as concrete numerical data.

Next, when the value of A exceeds 128 and is 383, one of the four pixels is set to 1, the remaining 3 pixels are set to 0 (FIG. 4B), and the quantization pattern is set to 1. ,
The quantization error at this time is A-255 (see FIGS. 2 and 3).

Similarly, when A exceeds 383 and is between 638, 2 out of 4 pixels are set to 1 and the remaining 2 are set to 0 (FIG. 4 (c)), and the quantization pattern is 2. , The error is A-510 (see FIGS. 2 and 3). When A exceeds 638 and is between 893, 3 out of 4 pixels are set to 1 and the remaining 1 pixel is set to 0 (see FIG.
(D)), assuming that the quantization pattern is 3, the error is A-7.
65, and when A exceeds 893 and is between 1020,
If all four pixels are set to 1 (FIG. 4E) and the quantization pattern is set to 4, the error becomes A-1020 (FIG. 2, FIG.
(See FIG. 3).

Next, a specific method of corresponding binary data to each pixel based on the quantization pattern will be described. Figure 5
Shows 2 × 2 pixels in the attention area A, which shows the priority of pixels to be printed. That is, the printing priority is given to the pixels having smaller numerical values. The numerical value 1 is Di, j + 3 in FIG. 1, the numerical value 2 is Di + 1, j + 3 in FIG. 1, the numerical value 3 is Di, j + 2, and the numerical value 4 is Di in FIG. The pixels of +1 and j + 2 are shown. Therefore, in the quantization patterns 0, 1, 2, 3, and 4 shown in FIG. 3, the pixel data of the area of interest after quantization is (a) (b) (c) (d) in FIG.
It is possible to correspond in the order of (e).

The quantization procedure of the attention area A has been described above. As a result, the generated error is the pixel area, B,
It is distributed to each block of C, D, and E. Regarding the method of distribution, it is possible to perform proportional distribution as in the conventional error diffusion, or to change the weight of distribution by random numbers,
Here, the method does not matter. In any case, all the error components are distributed to the peripheral pixels.

As described above, the processing for the area centered on A is completed, and then the target area is shifted by 2 pixels in the main scanning direction. Considering FIG. 1 as the center, the pixel area of B is set as the target area. Quantization is performed in the same procedure as the method of A above. Further, when the main scanning direction is completed, the target pixel is shifted by two pixels in the sub scanning direction, so that the next line is also quantized.

As described above, all pixels in the main / sub-scan are quantized. In this quantization,
By setting the target pixel to 4 pixels, 4 × 255 = 1020
This means that the gradation density has been saved. In terms of area, although it has 255 gradations per pixel, it has a display capability of 1020 gradations, and it is possible to perform precise and expressive image processing on halftone images. .

FIG. 6A shows the order of priority of quantization when 4 × 4 pixels are set as the target area on a concentric circle centered on a specific pixel. By weighting, it becomes possible to process halftones by the method of halftone dot expression. In addition, as shown in FIG. 6B, by weighting in a specific scanning direction, it becomes possible to express a halftone. Therefore, if a large number of such weighted conversion tables are prepared and these can be arbitrarily switched by software, they can be used for various types of image processing, and great advantages can be obtained. .

[0019]

As is apparent from the above description, according to the present invention, when the image data is quantized, the pixel area to be processed is expanded to a plurality of N × M pixels for processing. The dynamic range of the tone can be increased N × M times, and the quantization is performed by the error diffusion method, and further, in the error diffusion method, predetermined quantization is performed in N × M pixels. By performing the weighting of, it is possible to perform fine and expressive image processing on the halftone image.

[Brief description of drawings]

FIG. 1 is a diagram for explaining an embodiment of the present invention, in which target pixel areas (A, B, ... E) are designated as N (= 2) and M (=).
It is a figure which shows the example at the time of expanding 4 times of 2).

FIG. 2 is a graph showing an example of a quantization calculation method.

FIG. 3 is a table showing the quantization calculation method shown in FIG. 2 as numerical data.

FIG. 4 is a diagram showing pixel data of an attention area after quantization of quantization patterns 0, 1, 2, 3, 4.

FIG. 5 is a diagram showing pixels (2 × 2) in a target pixel area.

FIG. 6 is a diagram showing an example of weighting for quantization.

FIG. 7 is a diagram for explaining an example of an error diffusion method.

Claims (3)

[Claims] 1. An image processing method, characterized in that a quantization target area is N × M pixels, and quantization is performed for each pixel (where N and M are integers and N × M ≧ 2). . 2. The image processing method according to claim 1, wherein the quantization is performed by an error diffusion method. 3. The image processing method according to claim 2, wherein, in the error diffusion method, quantization weighting is performed on the N × M pixels in advance.