One pass enhanced adaptive interpolation for high-resolution video coding

Recently, many adaptive interpolation filter (AIF) algorithms have been proposed to improve the coding performance especially for the high-resolution video coding. However, most of the algorithms are two-pass encoding which needs to calculate the motion estimation twice and brings much computation burden to the encoder. This paper presents a one pass Enhanced AIF algorithm. It allows optimal selection among previous frames' AIF with the fixed H.264 high precision filter while a sum of absolute differences(SAD) based fast decision criterion is used to select the best one. Because the Wiener filter derived from the current frame will be used by the subsequent frames, not the current frame, there is only one encoding pass for each frame, which is a significant benefit for hardware implementation. Moreover, the proposed one pass AIF can reach almost the same compression performance when compared with two-pass AIF algorithms. The evaluation results shows that for most sequences the proposed algorithm can achieve around 35% total encode time saving with average gain around 5%.