KR100407488B1 - Method for processing error diffusion - Google Patents

Abstract

PURPOSE: A method for processing error diffusion is provided to improve the quality by reducing dot patterns generated in error diffusion by applying an appropriate weight value to the diffusion error. CONSTITUTION: An input unit obtains a color image. A quantizer(3) quantizes a corrected image to be fitted for a standard of an output device, and calculates a quantization value and a quantization error at the same time. An error weight value applying unit(7) stores a look-up table for compensating a diffusion error. The error weight value applying unit stores a weight value to a spreaded diffusion error, and diffuses the error value through an appropriate weight value, thereby reducing dot patterns. An error diffusion filter(5) spreads the calculated error value by putting the weight value to the original input image.

Description

Error diffusion processing method

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing for displaying a color close to an input color by diffusing an error in an output device using an arbitrary color input image using a limited number of colors. In particular, by appropriately applying a selective weight for diffusion error, The present invention relates to an error diffusion processing method for reducing a dot pattern of a sensitive part due to human visual characteristics.

Dithering and error diffusion are widely used as a representative method for displaying a suitable color in an output device having a limited output range and gradation. Since dithering only performs quantization processing on the unit region of the input image, the implementation is relatively simple. However, since the accurate representation of the input data and the resolution deterioration are in a trade-off, the resolution deterioration in the display device is not necessary for improving image quality. Inevitable Since the error diffusion method performs pixel unit processing, the resolution after processing is superior to dithering, and after quantizing each pixel, the error generated in the quantization step is propagated to adjacent pixels, thereby representing the input data of the unit region relatively accurately. However, the processing method has a disadvantage compared to dithering.

However, in the process of expressing the input color, a dot pattern is generated by repeatedly processing the same structure for each input data for each channel. The role of the dot pattern to express the color approaching the input value is very important. However, due to the influence of the dot pattern generated when the proximity color is expressed, the average effect of the error diffusion does not appear unless the distance for viewing the image is sufficiently secured. Therefore, when viewing an image, a soft low frequency component is reduced, and a high frequency corresponding to a dot pattern is increased, thereby degrading image quality.

This will be described in more detail with reference to FIG. 1 as follows. As shown in FIG. 1, the conventional error diffusion processing method is a simple extension of the black and white error diffusion, and cannot effectively reduce the dot pattern generated by applying error diffusion to three channels of red, green, and blue independently. This is because an image using a dot pattern is constructed because of the basic concept of error diffusion technology, and thus deterioration of image quality cannot be prevented. If the diffusion error value in error diffusion is extremely reduced, it is identical to the quantized image. On the other hand, no dot pattern occurs at all in the simple quantized image. However, since there are many low frequencies, the image is soft, but there are problems such as color distortion and resolution deterioration. On the other hand, if the quantization error value is propagated as it is, the resolution increases and the color distortion decreases. However, due to the occurrence of the dot pattern, there is a tendency for the image quality to deteriorate at close range. In the case of adjusting the amount of diffusion error, between the simple quantized image and the general error diffusion image, the advantages and disadvantages of each case are complementary to each other. Therefore, if the weight for the diffusion error is '0', it becomes a simple quantized image, and if it is '1', it is a general error diffusion image. Since the advantages and disadvantages of each case are complementary to each other, it is very difficult to produce an image having high resolution, small color distortion, and excellent image quality even at a short distance, which are composed only of the advantages of the two images.

Accordingly, an object of the present invention is to provide an error diffusion processing method for reducing a dot pattern of a sensitive part due to human visual characteristics by appropriately applying a selective weight for diffusion error.

In order to achieve the above object, the error diffusion processing method according to the present invention includes a) considering a visual sensitivity of a quantized output image with respect to a diffusion error value propagated in an input image of one of multi-tone colors and monochrome. Generating a numerically determined weight and storing the weight as a lookup table; b) generating a correction error by applying a weight to a diffusion error for a predetermined pixel of the input image; And c) diffusing the correction error to surrounding pixels of the predetermined pixel.

1 is a block diagram showing a conventional error diffusion processing apparatus.

2 is a block diagram for realizing the error diffusion processing method according to the present invention.

3 is a graph showing a quantization function used in the present invention.

4 is a structural diagram of a lookup table for diffusion error correction used in the present invention.

5 is a structural diagram of an error diffusion filter used in the present invention.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention.

2 is a configuration diagram for realizing an error diffusion processing method according to the present invention, and includes an input device (not shown), an adder 1, a quantizer 3, an error diffusion filter 5, a multiplier 6, and a diffusion. And an error weighting unit 7 for storing an error correction look-up table (LUT), a subtractor 8, an output device (not shown), and the like.

In FIG. 2, the input device is a device for acquiring a color image, and obtains a digital image through an analog-to-digital (A / D) conversion on an arbitrary analog input signal such as a TV or a video camera. For example. The corrected image 2 is obtained by adding the input signal value and the propagated error value by the adder 1 and serves as an input of the quantizer 3.

The quantizer 3 is a part that performs quantization on the modified image 2 in accordance with the standard of the output device. The quantizer 3 has a structure capable of simultaneously calculating a quantization value and a quantization error.

The output device is a device for displaying the output of the quantizer 3, and examples thereof include a monitor and a printer.

The error weight applying unit 7 storing the diffusion error correction lookup table (LUT) stores a weight for the spreading error value propagated, and reduces the dot pattern generated by spreading the error value through an appropriate weight. It is possible to improve the disadvantage that the image quality deteriorates even in the near field.

The error diffusion filter 5 serves to propagate the calculated error value by weighting the original input image. The error diffusion filter 5 serves as an input of the adder 1 to generate the corrected image 2.

The operation principle of the present invention is to reduce the dot pattern generated by adding a diffusion error correction LUT to the existing error diffusion algorithm.

The input image I _ij is an image composed of color vectors and has three or more elements for each pixel, and (i, j) means i rows and j columns in the input image. The corrected image M _ij ; 2 is the sum of the original input image I _ij and the error E ″ _ij propagated with respect to the center pixel to be processed, which is output from the adder 1 and can be expressed as Equation 1 below. have.

[Equation 1]

M _ij = I _ij + E " _ij

The modified image (M _ij ; 2) is quantized by the quantizer 3 so as to meet the output standard of the output device, and the transfer function Q _ij of the quantizer 3 is as shown in Equation 2 and FIG. .

[Equation 2]

Q (m _ij ) = q ×

Here, m _ij is one element of the M _ij color vector, and quantization is performed in the same manner for all elements. q is the quantization interval, (1 + p) is the number of desired output levels, u (.) is the unit step function, d is the width of one level.

The quantized data Q _ij is stored in an output image buffer (not shown) and transmitted to the output device. At the same time, the quantization error Eij is also calculated and input to the subtractor 8 to be used for the overall error calculation.

[Equation 3]

E _ij = M _ij -Q _ij

The quantized data Q _ij is input to a diffusion error correction LUT and used as an index for a weight to be applied to the diffusion error. The diffusion error correction value E ' _ij is obtained by multiplying the weight S _ij by the diffusion error E _ij and is expressed as in Equation 4 below.

[Equation 4]

E ' _ij = S _ij × E _ij

Here, S _ij represents a weight value to be applied to the quantization error for Q _ij . Since E ' _ij is determined by the value of Q _ij and its corresponding diffusion error weight S _ij , it is necessary to calculate and record the diffusion error correction weight for Q _ij in advance. Do not.

On the other hand, the calculation of the diffusion error weight S _ij is as follows. When observing an error-diffused image, the viewing distance is very important. If a sufficiently long distance is secured, the averaging phenomenon occurs according to the original concept of the error diffusion process, so that the dot pattern is not seen, and the image is of excellent image quality. However, when the observation distance is close, the dot pattern is not averaged and appears as it is, which causes deterioration of image quality. Due to human visual characteristics, when a dot pattern occurs in a visually sensitive portion, the effect of distortion is prominent, and even in the insensitive portion, it tends to be hardly felt. By using this phenomenon, it is possible to construct a method faithful to error diffusion while preventing the occurrence of dot patterns in the visually sensitive areas and maintaining the occurrence in the insensitive areas, while preventing deterioration of image quality as much as possible. The weight of the diffusion error can be calculated based on the sensitivity of the corresponding color for each quantization value. For example, based on the brightness value Y of the corresponding color, if the value of Y is large, it is sensitive when the dot pattern is generated, and the weight is determined by reducing the diffusion error value. If the Y value is small, the weight can be increased, and the processing can be performed to be faithful to the error diffusion method. That is, the weight is determined to be inversely proportional to the brightness value of the corresponding color. In addition, it is possible to reduce the dot pattern by determining the sensitivity for the color of the color as well as Y, and applying an appropriate weight according to the determined sensitivity. That is, the weight is determined by calculating the sensitivity to the wavelength and the intensity in consideration of the color property of the corresponding color. As an example of using the Y value, a diffusion error correction LUT storing a diffusion error weight value is shown in FIG. 4. In FIG. 4, when the input image is a color image, the lookup table for diffusion error correction includes a color index and a corresponding weight, and when the input image is a black and white image, the lookup table for diffusion error correction includes a gray scale index and a corresponding weight.

Since the correction error E ' _ij to propagate is a form in which the scale of the quantization error is adjusted, it performs a role of reducing the occurrence of the dot pattern while performing error diffusion. The calculated correction error E ' _ij is sent to the error diffusion filter 5. The error diffusion filter 5 serves to propagate the obtained correction error E ' _ij to adjacent pixels by applying an appropriate weight, and E " _ij added to each adjacent pixel is expressed as shown in FIG. Means the center pixel to be processed, W _i is the pixel-by-pixel weight to be.

The adder 1 adds an error E " _ij to be propagated to a pixel to be processed next in the input image and generates a corrected image 2, whereby the process of Equation 4 in Equation 1 mentioned above Iterate over the telephone office.

Through the error diffusion process of the present invention, the result of image quality improvement can be obtained by reducing the dot pattern by applying a weight to the diffusion error as well as the existing quantization error diffusion. At this time, Floyd & Steinberg proposed the weight for the correction error distributed to adjacent pixels.

During the process of Equations 1 to 4, if the dimension of the variable used is limited to 1, it can be seen that error diffusion is applied to the black and white image, and the effect of reducing the dot pattern is also maintained.

The present invention is used in color output devices such as color printers, color LCDs, color monitors and the like.

As described above, the present invention can improve the image quality by reducing the dot pattern generated during error diffusion by applying the weight of the diffusion error to the existing error diffusion processing method, and can prevent the image degradation at a short distance. There is an advantage. The present invention can be applied to systems such as color CRT monitors, LCDs, printers, LED display elements, and the like.

Claims (5)

a) generating a numerically determined weight value based on a visual sensitivity of the quantized output image with respect to a diffusion error value propagated in one of the multi-color color and the black and white image, and storing the weight as a lookup table; b) generating a correction error by applying a weight to a diffusion error for a predetermined pixel of the input image; And and c) diffusing the correction error to surrounding pixels of the predetermined pixel. The error diffusion processing method according to claim 1, wherein when the input image is a color, the lookup table includes a color index and a corresponding weight. The error diffusion processing method according to claim 2, wherein the weight is set in inverse proportion to a brightness value of a corresponding color. The error diffusion processing method of claim 2, wherein the weight is determined by calculating sensitivity to wavelength and intensity in consideration of color attributes of a corresponding color. The error diffusion processing method of claim 1, wherein when the input image is black and white, the lookup table includes a gray scale index and a weight.