N T I. Introduction. II. Proposed Adaptive CTI Algorithm. III. Experimental Results. IV. Conclusion. Seo Jeong-Hoon

Transcription

1 An Adaptive Color Transient Improvement Algorithm IEEE Transactions on Consumer Electronics Vol. 49, No. 4, November 2003 Peng Lin, Yeong-Taeg Kim Seo Jeong-Hoon CONTENTS N T I. Introduction III. Experimental Results IV. Conclusion 1

2 I. Introduction In color TV broadcasting standards, the transmitted signals consist of Y, U, V. In comparing with the luma signal bandwidth, the chroma signal bandwidth is rather narrow Causing smeared color edges in the received image. Various methods have been proposed to enhance the sharpness of color transitions : Such Techniques are referred to as Color Transient Improvement. CTI is to add a correction signal to the original chroma signals to restore the components lost, followed by a post-processingprocessing to remove the overshoot. Usually the correction signal is multiplied by a control parameter. Most CTI methods use a constant control parameter.(a cause of overshoots.) Overshoot make the appearance of some color edges unnatural. Adaptive CTI makes the gain control parameter depend on the local image feature Different regions of the image can be treated a different correction signal. This correction signal is able to prevent the overshoot happening. It eliminates the post-processing for overshoot removal. The existing noise in the homogeneous region will not be amplified. 2 I. Introduction C(color)TI L(luminance)TI 3

3 The algorithm will be presented for processing horizontal color transient. The algorithm needs to compute the gain control function and to generate the correction signal. The gain control function is computed using the second derivative of the signal. The correction signal is generated from the local minimal or maximal of the signal and sign of second derivative. Fig. 1. Block diagram of the proposed adaptive color transient improvement algorithm. 4 In order to compute the gain control function, The input signal U is smoothed by a Gaussian filter standard deviation σ = 1. : smoothed signal From the first smoothed signal, the second derivative and its sign. The second derivative is approximated by central difference. The second derivative is calculated from a smoothed version of the input signal. It is less sensitive to noise and is more accurate in reflection the geometric property of the transition curve of the input signal The sign of the second derivative block. is supplied to the correction signal 5

4 The absolute value of the second derivative control function as following : is used to compute the gain : Positive constant (1) The gain control function in (1) gives different gains for different image areas. Making the color transient t enhancement processing adaptive to local l image feature. The range of the gain control function is restricted between 0 and 1. This property will be used together with the correction signal to avoid overshoots. 6 In order to generate the correction signal, The input signal U is supplied to the distance to local minimum block and distance to local minimum block. The distance to local minimum(maximum) block computes the local minimum U min(max) by searching a neighborhood of radius r centered at the current signal sample position. It then computes the distance between the input signal U and the local minimum (maximum) U min(max) for the current position, which h is U U min (U max U ). Both the distance are passed the the correction signal block. And the sign of the second derivative is used to determine which correction signal to use. Correction signal : Correction signal : Correction signal : 0 7

5 After the correction signal has been generated, it is multiplied by the gain control function computed in (1) and then added to the original signal to yield the enhanced signal W. (2) The gain control function is computed from the smoothed signal, but correction signals are computed from the original signal. Since, it is to see that The enhanced signal will never go below(over) its local minimum(maximum). There is no overshoot in the enhanced signal. 8 Fig. 2. The impact of the proposed algorithm on color transition curve. The solid line is the transition curve of the input chroma signal U. According to the geometric property of the second derivative, The second derivative is positive The curve is concave up the curve will be pushed down towards its local minimum by subtracting. The second derivative is negative The curve is concave down the curve will be pushed up towards its local maximum by adding. The result curve(the dotted line) is an enhanced chrominance signal with steeper transitions, which does not have overshoot. 9

6 III. Experimental Results The proposed adaptive color transient improvement algorithm has been tested using different color images with band-limited chrominance components. There are two parameters used in the proposed algorithm. The parameter C in the gain control function. The search range r for finding the local minimal and maximal. In simulation, C = 1.0, and r = 3 are used. The first example of results is for the color bar image with and-limited chroma components. The second example of our result is for the pepper color image with both the U and V components being low-pass filtered to simulate the band-limited chroma signals. 10 III. Experimental Results Fig. 3. Results on color bar Fig. 4. Results on color bar Fig. 4. Results on pepper image (yellow-cyan color edge) (green-magenta color edge) (Results on pepper image(color edge transition from pixels(160,15) to (160,209)) 11

7 IV. Conclusion In this paper, an adaptive color transient improvement algorithm for enhancing the color edge transition of band-limited color images have presented. The proposed method is able to produce steeper yet smooth enhanced color edge transitions without introducing unpleasant overshoot effects. Therefore eliminates the post-processing for removing overshoots. 12