A Single-Pass-Based Localized Adaptive Interpolation Filter for Video Coding
Kai Zhang,Xun Guo,Jicheng An,Yu-Wen Huang,Shawmin Lei,Wen Gao
DOI: https://doi.org/10.1109/TCSVT.2011.2157194
IF: 5.859
2012-01-01
IEEE Transactions on Circuits and Systems for Video Technology
Abstract:Recently, an efficient coding tool named adaptive interpolation filtering (AIF) has been proposed to hybrid video coding scheme. By introducing Wiener filter into the fractional-pixel interpolation procedure, AIF can reduce the inter-prediction error and improve coding efficiency significantly. However, the training-based Wiener filter mechanism brings AIF an inherent multi-pass encoding structure, which imposes big burdens on the encoder in terms of huge computational complexity and memory access. In this paper, we propose a single-pass-based localized adaptive interpolation filtering (SPL-AIF) algorithm for video coding, which can reduce the complexity of AIF dramatically without sacrifice of its outstanding coding performance. The proposed SPL-AIF algorithm is based on the observation that there is a high correlation among optimal interpolation filters of consecutive frames, and different regions in a frame often possess different statistical characteristics. Accordingly, the proposed algorithm can be designed including two major parts. First, a competitive filter set which includes the optimal interpolation filters of several previous frames as well as the fixed H.264/AVC interpolation filters is built up for the coding of the current frame. Then a rate-distortion optimization criterion is used to select the best one at macroblock (MB) level. In order to reduce overhead, a predictive coding method is used to compress the filter signaling flag for each MB. Experimental results show that, by using the proposed algorithm, the encoding complexity can be reduced significantly while the average coding gain in Bjöntegaard distortion bit-rate reduction can be improved about 1% compared with the multi-pass AIF. The proposed method has been adopted into the Video Coding Expert Group Key Technology Area software.