Adaptive Search Algorithm Based on Double-initial-cross Search Pattern

This paper proposes an adaptive search algorithm based on double-initial-cross. The algorithm employs big-small cross search pattern, octagon and quasi-cross search pattern for adapting to different block content. Median prediction and early skip strategy are used to reduce unnecessary search. Compared with UMHexagonS, simulation results of the sequences present that this algorithm can save motion estimation time of average 15.88% and 15.94% while obtaining similar PSNR and Bitrate.