A refined motion estimation strategy for adaptive interpolation filter

摘要

Adaptively selecting the filter coefficients for interpolating fractional pixels (sub-pel) can significantly improve the quality of reference pictures. However, the complexity of implementation is greatly increased by performing motion estimation (ME) and mode decision (MD) for multiple passes. This paper briefly introduces the adaptive interpolation filter (AIF) technique and proposes a refined ME strategy for realizing the AIF process with low complexity. By using this method, the 2-pass encoding procedure can be avoided while the coding efficiency of the AIF is maintained. The proposed method is implemented and verified on the AVS video software platform.