Abstract:

The latest H.264/AVC video compression standard promises significantly higher compression efficiency than prior standards. H.264/AVC achieves coding gains through a rich set of advanced coding tools including variable block size motion compensation, quarter-pel motion compensation and long-term memory motion compensation [1]. However, with so many coding options available, it has become extremely challenging to efficiently choose the coding parameters, including motion vectors and prediction modes, such that near optimal compression efficiency is achieved [2,3]. In literature, various attempts have been made in order to reduce the complexity of mode decision and motion estimation. Zhou et al. [5] proposed to determine initial search center based on the correlation between motion vectors of different block sizes. Fast motion estimation algorithms such as EPZS [6], UMHexagonS [7], and SEA [8] have been proposed to reduce the number of searching points in motion estimation, while the recent-biased search [9] and forward motion trace [10] have been introduced to reduce the complexity of long term memory motion compensation. A mode decision algorithm based on a coarse-to-fine approach [11] has been proposed assuming a monotonic rate-distortion (RD) relation across block sizes. Despite these efforts, further complexity reduction is still very desirable for practical applications. In this work, we propose a new hierarchical complexity control framework to efficiently control the complexity of encoding process, with focus on motion estimation and mode decision algorithms. The goal is to provide a complexity scalable encoder that may be applied to various applications.