KR20010036004A - Method of nonlinear interpolation of color images - Google Patents

Abstract

PURPOSE: A nonlinear interpolation method for a color video is provided to obtain a color difference signal with reference to a luminance signal when interpolating a sampled color difference signal again. CONSTITUTION: Pattern information of a luminance change is obtained from adjacent pixel values with reference to a luminance component. Obtained pattern information of the luminance change is reflected to a color difference signal for interpolating a desired color difference signal. Adjacent pixel values are two pixel values adjacent to the front and the rear to a pixel value of a certain position.

Description

Nonlinear interpolation method of color image {METHOD OF NONLINEAR INTERPOLATION OF COLOR IMAGES}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a nonlinear interpolation method of a color image, and more particularly, to a method for nonlinear interpolation of a color image in an image processing system such as display, storage, and compression storage of an image.

In particular, the present invention focuses on the fact that the luminance signal Y has more faithful information on the edge component because the luminance signal Y has a wider bandwidth and a large number of samples, and thus the number of samples and the information on the edge are relatively small. In particular, the present invention relates to a nonlinear interpolation method of a color image that enables reconstruction of a clear image at an edge part by obtaining an interpolation value of a color signal.

Digital color imagery consists of an appropriate combination of R, G, and B signal components.

This signal is converted into a luminance Y, color difference U, V (or Y, I, Q or Y, Cb, Cr) signal to remove the correlation between the components.

The Y, U, and V signals are more efficient than the R, G, and B signals having the same bandwidth, and are more efficient. The Y, U, and V signals can be obtained by using an appropriate matrix circuit (hardware or software implementation). .

1 is a block diagram showing an example of a conventional digital color image processing circuit, in which an input video signal (Video) is converted into Y, U, and V signals by passing through a 3x3 matrix circuit (101), where U, V The signals are passed through low pass filters (LPF) 102 and 103 to prevent aliasing, and then through the 2: 1 down samplers 104 and 105, respectively, to Y, U ', V. 'Getting a signal.

On the other hand, the human eye is more sensitive to the luminance signal than the chrominance signal, so the U and V signals are sampled to reduce the amount of data.The unsampled color format is obtained by sampling the color difference signal 4: 4: 4 once. The color format is 4: 2: 2 and the color format obtained by sampling it once is 4: 2: 0.

Where 4: 2: 0, 4: 2: 2, and 4: 4: 4 are the ratios of the sampling frequencies of the luminance signal and the two color difference signals, and 4: 2: 0 is alternately 4: in the odd and even lines. Since 2: 0 and 4: 0: 2, one of them is represented as 4: 2: 0.

'Luminance' is the degree to which the brightness of an image is expressed. In ITU-R Recommendation 601 (formerly CCIR Recommendation), luminance of a pixel is represented by 8 bits.

'Color difference' is information representing the color of an image. In Recommendation 601, two 8-bit bits are used to represent a color of a pixel.

The coordinate system representing the color is called a 'color space'. The general shape of the Y, Cb, and Cr systems used in MPEG is to express a pixel with three 8-bit information such as luminance Y and color difference Cb and Cr.

In this way, a means of allocating 24 bits of information per pixel, but reducing the color information using the human eye which is not very sensitive to color, is often used.

4: 4: 4 is the one that does not reduce color information, 4: 2: 2 is the half cut in the lateral direction, and 4: 2: 0 is the half cut in both the lateral and longitudinal directions.

Therefore, the color information becomes 1/4 of the luminance information at 4: 2: 0.

In order to prevent aliasing in the signal sampling process as described above, a filter is used as shown in FIG. 1 (mainly, a low pass filter is used to remove the high frequency altogether), and in the process of interpolating the signal in reverse, Use an appropriate filter.

However, since the linear low pass filter is mainly used in the sampling and interpolation process as shown in FIG. 1, in the case of an image having distinct edges of the image (signal), the image that is interpolated after sampling is merged with the edges, compared to the original image. It will look blurry.

Since human vision is sensitive to edge components, the edge-bound image becomes unobtrusive, as described above, and causes significant obstacles in discriminating and recognizing objects including image outlines.

The present invention proposes a method of using a nonlinear filter as an interpolation filter to solve a conventional problem.

The present invention provides a method for obtaining a chrominance signal by referring to a luminance signal when interpolating a sampled chrominance signal again under the assumption that the signal will change rapidly around the edge even if the luminance signal and the chrominance signal pattern are different. .

In the present invention, since the luminance signal (Y) in the color image has a wide bandwidth and a large number of samples, it focuses on the more faithful information on the edge component, from which a relatively small number of samples and less information on the edge color By obtaining an interpolation value of a signal, a nonlinear interpolation method of a color image is provided, which enables reconstruction of a clear image, particularly at an edge portion.

1 is a block diagram showing an example of a conventional digital color image processing circuit;

Fig. 2 is a diagram of pixel samples showing some examples of a method for restoring a color difference signal by nonlinear interpolation according to the present invention.

The present invention obtains the pattern information of the luminance change from neighboring pixel values by referring to the position of the luminance component, and interpolates the desired color difference signal by reflecting the pattern information of the luminance change into the color difference signal. Interpolation method.

According to the present invention, pattern information α of luminance change is determined from two neighboring pixel values Ay and By with reference to the luminance component Pref; Pref =? Ay + (1-?) By, for the color difference signal P_unknown for which interpolation of the pattern information? Of this luminance change is desired, respectively; A nonlinear interpolation method of a color image, characterized in that interpolation is performed with P_unknown =? A + (1-?) B with neighboring color difference signal pixel values (A, B).

In the present invention, the α value is a nonlinear interpolation method for color images, characterized in that it has a value between 0 and 1.

FIG. 2 is a diagram illustrating an example of a nonlinear interpolation method of a color image of the present invention. The luminance signals Y and the color difference signals U and V are shown, and are nonlinear interpolation methods in the horizontal direction (1 dimension). (Nonlinear interpolation can also be performed in the vertical direction (two dimensions).)

In Fig. 2, black circles are known pixel values, white circles are pixel values to be interpolated, and shaded circles represent pixel values interpolated according to the present invention.

Referring to Fig. 2, the nonlinear interpolation method of the color image of the present invention will be described in more detail.

First, the position of the luminance signal Y component is referred to.

The change pattern information α of the luminance signal is obtained using the pixel value Pref at the position referred to as the edge and two pixel values Ay and By adjacent to each other.

In other words, the change pattern information? Of the luminance signal is obtained from Pref =? Ay + (1-a) By.

Herein, the change pattern information α of the luminance signal is a parameter representing the tendency of the color difference signals U and V to follow the change of the luminance signal.

The change pattern information α of the luminance signal is used for interpolation of the color difference signals U and V. FIG.

In the case of the color difference signal U; If the pixel value of the position to be interpolated is called Pu_unknown, the change pattern information α of the luminance signal is applied to two neighboring pixel values Au and Bu, and Pu_unknown = αAu + (1-α) Bu. Obtain as

The pixel value Pu_unknown thus obtained is shown as a hatched circle in FIG.

That is, by setting the α value to an appropriate value with reference to the position of the luminance component, the color difference signal can be placed at a desired position (the position of the hatched circle).

Here, if the value of α is large, the interpolation signal will be placed at the level close to the Au position, and if the value of α is small, the interpolation signal will be placed at the level near the Bu position.

In FIG. 2, Pu_i shows the interpolation position when interpolated using a conventional linear filter, and in the present invention, if α = 1/2, it will be the same as the existing linear interpolation.

In the case of the color difference signal V; If the pixel value of the position to be interpolated is called Pv_unknown, Pv_unknown = αAv + (1-α) Bv by applying the change pattern information α of the luminance signal to two neighboring pixel values Av and Bv. Obtain as

The pixel value Pv_unknown thus obtained is shown as a hatched circle in FIG.

That is, by setting the α value to an appropriate value with reference to the position of the luminance component, the color difference signal can be placed at a desired position (the position of the hatched circle).

Here, if the value of α is large, the interpolation signal will be placed at a level close to the Av position, and if the value of α is small, the level close to the Bv position.

In FIG. 2, Pv_i shows the interpolation position in the case of interpolation using a conventional linear filter, and in the present invention, if α = 1/2, it will be the same as the existing linear interpolation.

In the case of interpolation using such a nonlinear method, the interpolation process of the proposed method can be performed after 2: 1 sampling without using a filter in the sampling process.

That is, in FIG. 1, encoding is performed without using the low pass filters 102 and 103. When interpolation, the interpolation of the color difference signals U and V is performed by looking at the tendency of the luminance signal Y as described above. .

In this case, a simpler hardware or software can be implemented without using a low pass filter in the encoder, and it can provide a clear image encoding / decoding technology base having the maximum edge components.

In the nonlinear interpolation method of a color image of the present invention, an image having a sharp edge can be secured compared to a conventional method for application fields such as a sharp edge image such as animation or computer graphics, an image including subtitles or characters, and the like.

In the encoding process, encoding can be performed without using a low pass filter to prevent an aliasing phenomenon, and encoding / decoding technology that can obtain the interpolation value of the chrominance signal by looking at the tendency of the luminance signal during decoding is implemented. Can be.

Claims (5)

Obtaining the pattern information of the luminance change from neighboring pixel values by referring to the position of the luminance component, and interpolating the color image signal by interpolating the desired color difference signal by reflecting the pattern information of the luminance change obtained in the step into the color difference signal. Nonlinear interpolation method of a color image, characterized in that. The non-linear interpolation method of claim 1, wherein the neighboring pixel values refer to two pixel values adjacent to each other back and forth with respect to the pixel value at a predetermined position. Pattern information α of luminance change from two neighboring pixel values Ay and By with reference to the luminance component Pref at a predetermined position; Obtaining Pref = αAy + (1-α) By, For a color difference signal P_unknown for interpolating the pattern information? Of luminance change obtained in the above step, respectively; A nonlinear interpolation method of a color image, wherein the color image is interpolated by interpolating the neighboring color difference signal pixel values (A, B) to P_unknown = αA + (1-α) B. 4. The nonlinear interpolation method of claim 3, wherein the α value has a value between 0 and 1. 5. The video signal according to claim 3 or 4, wherein the video signal comprises a luminance signal (Y) and a color difference signal (U, V), a luminance signal (Y) and a color difference signal (I, Q), a luminance signal (Y) and a color difference signal ( Cb, Cr) non-linear interpolation method of the color image, characterized in that according to any one of the signal formats.