Adaptively weighted update steps using chrominance for scalable video coding

Scalable video coding using motion-compensated temporal filtering is one of the latest trends in video coding standardization. In the lifting based motion-compensated temporal filtering framework, such as that in the joint scalable video model (JSVM), the data used in the update steps are basically the residuals from the motion-compensated prediction. When the motion model used in the prediction steps fails to capture the true motion, energy in the high-pass temporal frames becomes substantial and strong ghosting artifacts may be introduced to the low-pass frames during the update steps. In this paper we propose a new block-based update approach, which takes advantage of the chrominance information of the video sequence to further reduce ghosting artifacts in low-pass temporal frames. We adaptively weight the update steps according to the energy not only of luminance pixels, but also of chrominance pixels in the high-pass temporal frames at the corresponding locations. Experimental results show that the proposed algorithm can significantly improve the quality of the reconstructed video sequence, in PSNR and visual quality.