JPH01157166A - Image processing device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field] The present invention relates to an image processing apparatus, such as a digital copying machine and a digital facsimile, that digitally handles images.

[Prior art]

In general, digital copying devices that sample images using a CCD sensor, etc., output the digitized data from a digital printer such as a laser beam printer, and reproduce the image have become widely used in place of conventional analog copying devices due to the development of digital devices. It's coming.

In order to reproduce halftone images, this digital copying apparatus generally employs a method of performing gradation reproduction using a dither method or a density pattern method. However, this method has the following problems.

(1) When the original is a halftone image such as a print, a periodic striped pattern that is not present in the original appears in the copied image.

(2) If the original contains line drawings, characters, etc., the edges will be cut off due to dithering and the image quality will deteriorate.

The phenomenon (1) is called a moiré phenomenon, and the cause of its occurrence is: A) Beat between the halftone original and input sampling B) Beat between the halftone original and the dither threshold matrix In particular, the phenomenon of (B) is generally When the dither threshold is arranged in a dot-concentrated manner, the output image also has a pseudo-halftone structure, which causes beats with the input document and causes moiré phenomena.

On the other hand, there is an error diffusion method as another binarization method.

This method calculates the density difference between the original image density and the output image density for each pixel, and distributes the error amount, which is the result of this calculation, after applying specific weighting to surrounding pixels. Regarding this, see the literature R, W, F]oyd an
dL.

Steinberg, S.I.D., 17. pp75-77
(1976).

Since this method has no periodicity, no moiré occurs in the halftone dot image. Furthermore, the image resolution is also superior to that of dither.

However, for images with little variation in shading, such as photographs, a unique striped pattern may appear in the output image, or the highlights of the image,
There were drawbacks such as noticeable grainy noise in shadow areas.

As a countermeasure to this problem, a method can be considered to make the noise and striped pattern less noticeable by using a multilevel printer with an intermediate density. For this reason, a method can be considered in which the output level of error diffusion is set to a multi-value level of three or more values, but there is a problem in that false contours are likely to occur if the multi-value processing is simply performed.

[verbal]

The purpose of the present invention is to eliminate the drawbacks of the above-mentioned conventional examples.The present invention uses an error diffusion method to prevent false contours that occur when processing with multiple values, and to produce high-quality and high-definition images of any original. The present invention provides an image processing device that reproduces images.

FIG. 1 is a block diagram showing one embodiment of the present invention.

The image data read by the input sensor section 10, which has a photoelectric conversion element such as a COD and a drive system that scans it, is
/D converter 11. Here, the data of each pixel is converted into 8-bit digital data. This results in 25
This means that the data has been quantized to have six levels of gradation.

Next, the correction circuit 12 performs shading correction and the like to correct unevenness in sensitivity of the sensor and unevenness in illumination due to the illumination light source, etc., by digital calculation processing.

Next, this correction signal 100 is input to the multilevel error diffusion circuit 13. The error diffusion circuit 13 obtains a signal 101 requantized by an error diffusion method described later. This signal 101 is input to the multilevel printer 14.

What is a multilevel printer? A binary printer is a dot 0N10FF.
While only two states (black and white) can be expressed, gray, which is an intermediate level, can be expressed. A printer with one level of gray is called a ternary printer, and one with two levels of gray is called a four-level printer.

In the multilevel printer 14, the amount of ink is controlled,
An image is formed by outputting this onto a predetermined sheet of paper.

FIG. 2 is a block diagram showing details of the multilevel error diffusion circuit of FIG. 1.

The image signal 100 is converted to correction data 11 by an adder 70, which will be described later.
4 and becomes a correction signal 109. This signal is quantized into multilevel levels by a requantizer 72. here,
Taking the case of requantization into three values (0,128°255) as an example, the result is as follows.

Here, T, , T2 are set in the requantizer 72 by setting signals 131 and 132 from the threshold setter 71.

The output signal 110 is sent to the multilevel printer 1 in the output buffer 73.
7 and is converted into a signal 101 sent to the multilevel printer 17.

On the other hand, the signal 110 and the signal 109 are subtracted by a subtracter 74, and this is stored as an error (111) in a location corresponding to m in the error buffer memory 75. In the error buffer memory, error data generated by requantization is stored for a total of three lines, including the current line and two lines before.

The correction data 114 is calculated by calculating the data of 12 pixels of error data a - 1 of surrounding pixels by predetermined weighting and weighting the sum, assuming that the current pixel position to be processed is the position m of the error buffer memory 75. Attached are signals processed by the arithmetic circuit 76.

Writing this as a formula, image data X15. Correction data X,
′, requantized data y+1+weighting is coefficient α kl+error ε1”, then ε,,=y,,−x,.

It is. An example of weighting coefficients is shown in FIG.

FIG. 3(a) is a typical example in which a correction value is created by referring to an error from a wide range, and FIG. 3(b) is a typical example in which a correction value is created by referring to an error from a narrow range.

FIG. 4 is a block diagram showing details of the requantizer 72 of FIG. 2.

Two threshold signals 131 from the threshold setter 71.
132 and the correction signal 109 are each sent to a comparator 50.5.
1 is input. In comparator 50゜51, input terminals A and H
The output is set so that when A>B holds true for the data, the output D=1, and otherwise, D=O. signal 1
31 is a threshold value T1. It is assumed that a threshold value T2 is set for the signal 132.

In the encoder 52, the output 133. of the comparator 50.51.
134 and the signal X coded according to this value.
get. This coded signal X is shown below.

However, M=0. N=1 cannot actually exist because the threshold values 131 and 132 are set such that T+<'l'.

The encoder output X (135) is supplied to a comparator 54° selector 55. The comparator 54 connects the value (136) set in the setter 53 to the terminal A and the signal (135) to the terminal B, and outputs D=1 when A=B holds true. This signal (140) is input to the weighted calculation circuit 140 and is used as a data area matrix size switching signal from the error buffer.

Also, the selector 55 selects the signal (
137), the signal from the gray level setter 57 (138), and the signal from the black level setter 58 (139).

The relationship between the select signal S and the output X is shown below.

The signal selected by the selector 55 becomes the signal (110). By setting the value 1 corresponding to the intermediate level (gray level) in the setting device 53, the output 1 of the comparator 54 is set when the output is gray (when the signal 138 is selected).
40 will be set to 1.

FIG. 5 is a circuit block diagram for weighting data from the error buffer memory 75 and switching the matrix size based on the signal 140 in the coefficient circuit 76.

Error data of error buffer memory 75-a ~ 75
-1 is LUT (lookup table) 80-a~8
0-1! and each weighting is multiplied according to the coefficient.

An adder 81 calculates the sum of each result.

This LUT can switch weighting coefficients using a signal 140.

FIG. 3 shows a correspondence table of weighting coefficients for two types of pixel positions (a-11). When the signal 140 is 0, the weighting coefficient corresponds to the large matrix size shown in FIG. 3(a), and similarly when the signal 140 is 1, the weighting coefficient corresponding to the small matrix size shown in FIG. 3(b) is selected.

Through the above operations, it is possible to make the size of the diffusion matrix of the error diffusion method variable according to the value of the signal 140.

In other words, in this embodiment, a small matrix is selected when the multivalued data is at a gray level, a large matrix is selected at other times, and the error is diffused based on that matrix.

This makes it possible to prevent false contours that would otherwise occur when simply using the same matrix to diffuse errors. This false contour is caused by the creation of a periodic structure in the image when a single matrix is used, and by changing the matrix, the periodic structure is no longer created and the false contour is created. It is possible to prevent the occurrence of

Furthermore, by using a small matrix for the gray level, that is, the intermediate density area, and using a large matrix for the other areas (highlights and shadows), it is possible to obtain an intermediate density image with high resolution in the intermediate density area. It is possible to obtain smooth images without grainy noise in highlight and shadow areas. This is because when using the error diffusion method, the smaller the matrix is used, the higher the resolution can be because the density can be stored in a narrower area. This is because an image can be obtained.

FIG. 6 shows an embodiment in which the present invention is applied to a color image reproducing device.

A color image (not shown) is read by a color human sensor 60 and output as an RGB signal. This signal is
The A/D converter 61 converts it into an 8-bit digital signal. Next, the correction circuit 62 performs shading correction and gradation correction to correct non-uniformity of the light source of the reading system and uneven sensitivity of the sensor. Next, in the color conversion circuit 63, RG
From the B signal, Y (which corresponds to the amount of ink in a color printer)
yellow), M (magenta), C (cyan), and Bk (black) signals. The Y, M, C, and Bk signals enter multi-value error diffusion circuits 64a to 64d, undergo requantization processing, and enter a multi-value color printer 65, where a color image is reproduced.

〔effect〕

As explained above, in the present invention, the generation of false contours in images is prevented by varying the diffusion matrix size of the multi-value error diffusion method by the multi-value level values, thereby making it possible to obtain high-quality images. .

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a block diagram showing details of a multilevel error diffusion circuit, and Fig. 3 shows two types of weighting coefficients with different diffusion matrix sizes. 4 is a block diagram of the configuration of the requantizer 72, FIG. 5 is a block diagram of a circuit that switches the diffusion matrix based on the level signal 140, and FIG. 6 shows an embodiment of the present invention that is compatible with color images. This is a diagram. 10 is an input sensor, 11 is an A/D converter, 12 is a correction circuit, 13 is a multi-value error diffusion circuit, and 14 is a multi-value printer.

Claims (1)

[Claims] quantization means for requantizing image data to a smaller number of bits than the input data; storage means for holding error data generated when quantized by the quantization means; and correction means for correcting the error data; An image processing apparatus characterized in that the correction means makes variable the size of a matrix used for distributing error data to surrounding images according to the requantized value.