Adaptive wiener filter based chrominance up-sampling enhancement method for video coding

In order to improve the quality of the reconstructed video, the Wiener filter is employed in video coding to reduce the distortion caused by chrominance down/up sampling and encoding. Based on the mathematical analysis, it is verified that the distortion introduced by chrominance sampling and coding can be approximately treated as a wide sense stationary noise, which can be suppressed by the Wiener filter. Accordingly, in this paper, a two-dimensional Wiener filter is designed for reducing the noise of the reconstructed chrominance components, which further improves the video quality in the RGB color space as finally perceived by the users. Experimental results show that compared with bi-linear and inter-color up-sampling, the proposed method achieves the average gains of 0.42 dB and 0.30 dB, respectively. Meanwhile, the reconstructed video obtained by the proposed method is more close to the original one in terms of subjective quality¹.

I. Introduction

Most existing image/video coding standards, such as JPEG [1], MPEG [2] and H.26x [3] series, are based on the YUV4:2:0 format which is converted from the original RGB signal. This is because that the RGB color space has high inter-color correlations which is considered to be unsuitable for coding, whereas the RGB to YUV conversion has a good de-correlation property and therefore the YUV color space is consider appropriate for coding. Since the human visual system (HVS) is much more sensitive to the variation of luminance than chrominance, the chrominance signal is reduced by down-sampling to obtain the YUV4:2:0 signal. That is, the chrominance components are down-sampled with the ratio of 4:1, which means that four Y (luminance) samples are corresponding to each U/V (chrominance) sample.