Modified adaptive interpolation filtering algorithms in video coding

摘要

In this paper, we propose two algorithms to improve the rate-distortion performance and reduce the computational complexity, respectively. Due to the observation that content does not change dramatically from frame to frame in majority of cases and the effect of aliasing stays almost constant between consecutive frames, we assume that adaptive filters for the previous frame could accurately compensate the effect of aliasing for the current frame. In the first proposed algorithm, the interpolation filters are initialized with the adaptive filters for the previous frame instead of static ones. The second encoding pass of the existing adaptive filtering schemes is eliminated conditionally in our second algorithm. Experimental results show that our first algorithm produces significant bit-rate reduction (3.23% on average), while maintaining the same level of peak-signal-to-noise-ratio (PSNR). The execution time is reduced about 26.73% by the second algorithm, compared to KTA.