(A) computationally efficient motion search method for H.264/AVC

Abstract

H.264/AVC achieves higher coding performance than the previous video coding standards such as MPEG-2, MPEG-4, and H.263. Especially, its motion estimation (ME) accomplishes considerable coding performance. However, the H.264/AVC encoder requires heavy computational complexity due to the ME with multiple reference frames and variable block sizes. Therefore, in order to reduce the computational complexity, this thesis analyzes statistical characteristics of early termination (ET) methods in unsymmetrical-cross multi-hexagon-grid search (UMHS) and modifies the previous ET methods by using adaptive thresholds, which are based on rate-distortion (R-D) cost. In addition, the thesis proposes a new ET method and multi-octagon search pattern to decrease search points of the multi-hexagon-grid search. Experimental results show that the proposed algorithm reduces computational complexity of encoder while keeping R-D performance of UMHS in various encoding conditions.