Adaptive enhancement filtering for motion compensation

Abstract

This paper proposes an enhanced motion compensated prediction algorithm for hybrid video coding. The algorithm is built upon the concept of applying adaptive enhancement filtering at the motion compensation stage. For luma, a high-pass filter is applied to the motion compensated prediction signal to recover distorted high-frequency information. For chroma, cross-plane filters are applied to enhance the motion compensated signals by restoring the blurred edges and textures of the chroma planes using the high-frequency information of the luma plane. To verify the effectiveness, the proposed algorithm is implemented on the HM Key Technology Area (HM-KTA) 1.0 platform. Experimental results show that compared to the anchor, the proposed algorithm achieves average Bjentegaard delta (BD) rate savings of 0.4%, 8.8% and 7.4% for Y, Cb and Cr components, respectively.