JPH01147961A - Image processing device - Google Patents

Abstract

PURPOSE:To prevent granular noise at a high light part by using a threshold value matrix fluctuated periodically as a threshold value in case of quantization and varying the threshold value matrix in response to the information of an original picture. CONSTITUTION:A density fmn on a coordinate (m, n) is read from a scanner 301 and the result is given to an adder 302. The density fmn and an error xn weighted on an error diffusion table 308 by using a diffusion matrix are given to the adder 302, which outputs a sum gmn. On the other hand, pulses Pm, Pn representing positions from the scanner 301 are given to a threshold value table 305. Similarly, the density information fmn of the original picture is given to the threshold value table 305. They are used as the information to select the threshold value table and to select a threshold value in the selected threshold value table. The threshold value decided in this way is given to a binarizing circuit 303, compared with the value gmn and binarized.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing method for quantizing image data.

[Conventional technology]

Conventionally, the density pattern method, the Bayer method, the error diffusion method, and the minimum average error method, which is substantially the same as the error diffusion method, have been known as methods for quantizing image data, particularly as a binary power method.

[Problems to be Solved by the Invention] The density pattern method and the Bayer method compare the data values of the original image of interest and the threshold of interest in a threshold matrix to generate binarized data.

However, these methods have the drawback that the overall image density is different between the original image and the reproduced image after binarization. On the other hand, the error diffusion method and the average value minimum method have the advantage that the density is preserved because the error between the generated original image data and the output data is propagated to surrounding pixels when each pixel is binarized.

However, this error diffusion method has the following drawbacks.

(1) When the error diffusion method is applied to the highlight part of the original image,
Since the concentration data is generally low, even if errors are added,
Since it takes time to exceed the threshold, there is a drawback that there is a delay in the appearance of dots. For example, when an image such as a density pattern patch is output, a delay occurs in the peripheral area, and a printing delay occurs in the shaded area as shown in FIG. 6, degrading the image quality.

(2) In the error diffusion method, minute noise components in the original image are accumulated as errors, so the positions of printed dots are printed relatively randomly. However, humans perceive that the dots printed in the highlight areas are more grainy when printed with regular periodicity.However, when printing using the error diffusion method, the graininess noise is noticeable in the highlight areas. There are drawbacks.

[Means and effects for solving the problem] According to the present invention, in an image processing method that performs quantization while preserving the density of an original image, a periodically varying threshold value matrix is used as a threshold value during quantization. By using this method and changing the threshold matrix according to the information of the original image, both high resolution and high gradation are maintained, and granular noise and delay in the appearance of dots in highlight areas are prevented.

[Embodiment] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram for explaining an error diffusion method that performs binarization processing while preserving the density of an original image. In FIG. 1, first, a general method in which the threshold value during binarization is fixed will be explained. In FIG. 1, reference numeral 201 is an input unit composed of a scanner or the like for reading original image data. The output f0° from the scanner 201 indicates the density data of the pixel at the coordinate (m, n) point. Reference numeral 202 denotes an adder, in which data weighted and cumulatively added in the error diffusion table 208 is stored in the line buffer memory 209, and the cumulative error x from the line buffer 209 and f4° are added. Letting this cumulative error be X9, the concentration data f□
0 is the cumulative error X. are added g, n, = xn + f,,
, , are input to the binarization circuit 203. 203 are threshold data ①, g, . Comparison with the value is performed, and when g ff1Tl≧vlh, it is 1 again, and when <Vlh, it is 0.

Bind output. Do. The result is multiplied by 206 coefficients. This k changes depending on how many bits the data read by the input section 201 is quantized.
When handling data in bits, the density of one printer dot is 255. The error e0 is 207 and e, = to -D-,
−g, , is calculated and transmitted to the error diffusion table 208. The error diffusion table 208 uses a diffusion matrix to give a predetermined weight to the error eo, and stores it in the line buffer memory 209. For example, if the errors up to now are stored as shown in the line buffer 209, x
The error when processing position n++ is new. When one line of the original image has been scanned,
The first line of 209 contains the data of the second line, the second line contains the data of the third line, and the third line contains O.
enters. By repeating the process in this manner, binarization processing is performed using the error diffusion method. 1 of the output section 204 beams
, 0 values, and outputs a reproduced image by controlling the dots to turn on or off.

Next, with reference to FIGS. 2 to 5, a case will be described in which a periodically varying threshold matrix is used as the threshold value, and the threshold matrix is changed in accordance with the density of the original image. From the scanner 301, the density f m on the coordinates (m, n)
n is read and entered into adder 302. This f and the error X0 weighted using the diffusion matrix in the error diffusion table 308 are input to the adder 302. Therefore, the addition value gm++ in the adder 302 is g m++−f , n
n+x. On the other hand, p□ and pl are input to the threshold table 305 as pulses or address data indicating the position from the scanner. Similarly, density information f of the original image.

is also input into the threshold table 305. These are used as information for selecting a threshold table and selecting a threshold in the selected threshold table. The threshold determined in this way is input to a binarization circuit 303, compared with gIn11, and binarized. Thereafter, the same processing as the error diffusion method shown in FIG. 1 is performed.

Figure 3 shows the data g
0 is used to select a threshold pattern. It should be noted that the same processing as in FIG. 2 is performed in the parts labeled with the same numbers as in FIG.

FIG. 4 shows the threshold table 305 in more detail. 501 and 502 are step counters in the X-axis direction and the y-axis direction. For example, 503 threshold pattern memory TR
If OMI is as shown in FIG. 5(A), both CNTl and CNT2 become octal counters. The inputs prn and p are x- and y-axis pixel step pulses obtained from the input unit 301 or from a file storing image data. Therefore, C
The threshold data of the address determined by NTl and CNT2 is TR
Selected from OMI and output through DPI line.

Similarly, the memory 507 (TR0M2) 1, 1 corresponding to the threshold pattern memory in FIG. 5(B) is controlled by a quaternary counter 9-(7)CNT3°CNT4. Similarly, the threshold pattern memory of FIG. 5(C) is
1 (TR0M3), so CNT5. Each CNT6 becomes a binary counter. The data output from each counter is output through DPI to DP3. 513 is a fixed threshold value, and contains, for example, 128 data.

The above data are selected by gates 504, 508, 512, and 514. On the other hand, 516 is the density f mn of the original image data from the input unit 301 in FIG. 2, and data g to which an error has been added in FIG.

For example, signals classified into four levels are output according to the information. For example, input data O~63°64~127
, 128-191, 192-255, the A1 port of 516 goes high for bright input data of 0-63, and the A2 port goes high for input data of 64-127. becomes 128~1
The A3 port is selected at 516 so that it becomes high for input data of 91, and the A4 port becomes high for input data of 192 to 255. In other words, the dark areas of the original image have high density data.

Make the size of the threshold matrix smaller than the bright areas where the density data is low.

FIG. 4(B) shows a different threshold selector 30 from FIG. 4(A).
5 is a diagram showing the configuration of No. 5. 520° and 521 are counters for calculating two-dimensional coordinates. 523 f. .

or gffi, 2 bits divided into 4 stages according to
A signal is generated, CNTl.

It is input to CNT2. As a result, CNT1, CN
T2 becomes an octal counter as in Figure 4 (A),
It can be a quaternary counter or a binary counter. The addresses input by counters 520 and 521 and the data divided into four stages by 523 are entered into TR0M522, and as in FIG. 4(A), a specific threshold value in the threshold value matrix is selected by the density of the input data and the coordinate position counter. Ru.

FIGS. 5(A) to 5(D) are examples of threshold matrices, and the fifth
Figure (A) is selected when the input data value is 0 to 63,
FIG. 5(B) is selected when 64 to 127, and FIG.
C) is a matrix selected when 128 to 191,
FIG. 5(D) shows the threshold values selected in the case of 192 to 255.

Note that the threshold value data is displayed in hexadecimal notation. Moreover, each threshold value matrix is set so that the average value is 128 (when the input is 0 to 255), which is the intermediate value of the inputs. This threshold value may be set to be larger than the average value (for example, 128) when the threshold matrix size is large, and may be set smaller than the average value as the size becomes smaller. As a result, it is possible to make the highlight part more periodic and the dark part more random. Figure 5 (B)' is a modification of Figure 5 (B) with an average value of 128 and a threshold value of 128. An example is shown in which the width is smaller than (B).

As described above, according to the above-mentioned embodiment, grainy noise in highlight parts, which was a drawback of the error diffusion method, can be prevented by using a threshold matrix for the binarization threshold and creating a periodic structure. Moreover, in dark areas, the high resolution that is an advantage of the error diffusion method can be maintained by using a relatively small matrix size.

Furthermore, the delay in dot appearance in highlight areas that occurs when processing using a constant value can be prevented since the binarization is performed using a threshold matrix and also a threshold matrix larger than that for dark areas. This is because a larger matrix can generate smaller thresholds.

In this embodiment, a method of binarizing an image has been described, but the method can be similarly applied to a multi-value method.

Further, when applied to a color image, it can be realized by processing each color of Y, M, and C respectively.

〔Effect of the invention〕

As described above, according to the present invention, in an image processing method that performs quantization while preserving the density of an original image, a periodically fluctuating threshold matrix is used as a threshold during quantization, and the threshold matrix By changing the value according to the information of the original image, it is possible to maintain both high resolution and high gradation, and prevent grainy noise and delay in the appearance of dots in highlight areas.

4. Detailed explanation of the invention Figure 1 is a block diagram of binarization processing using a fixed threshold value, and Figures 2 and 3 are binarization processing using a threshold matrix that is an embodiment of the present invention. Figure 4 is a diagram showing the details of the threshold table in Figures 2 and 3, Figure 5 is an example of a threshold matrix, and Figure 6 is a diagram showing the conventional problems. FIG.

In the figure, 301 is an input section, 302 is an adder, 303 is a binarization circuit, 304 is an output section, 305 is a threshold table, 306
is a coefficient unit, 307 is an error calculator, 308 is an error diffusion table, and 309 is a line buffer memory.

8ηh 1 inch mortar 4X(B) 日°°゛竿57 6th recess Kunio Ogawa, Commissioner of the Patent Office 1, Indication of the incident Patent Application No. 307285 of 1988 2, Title of the invention Image processing method 3, Amendment Relationship with the case of a person who does
) Canon Co., Ltd. Representative Ryuzo Kaku
Part 4, Agent address: Canon Co., Ltd., 3-30-2 Shimomaruko, Ota-ku, Tokyo 146 (telephone: 758-2111) 5, Specification subject to amendment and drawings 6, Contents of amendment (1) Page 12 of the specification Amend “invention” in line 5 to “drawings”.

(2) Correct Figure 3 as shown in the attached sheet.

Claims (1)

[Claims] In an image processing method that performs quantization while preserving the density of an original image, a periodically varying threshold matrix is used as a threshold during the quantization, and the threshold matrix is used as information on the original image. An image processing method characterized by changing the image according to.