An H.264 Spatial-temporal Hierarchical Fast Motion Estimation Algorithm for High-Definition Video

Abstract

   H.264 advanced video coding exhibits much higher coding gain as well as computational complexity than previous video coding standards due to the utilization of coding tools such as variable block size and multi-reference frame in motion compensation process. There exist plenty of research outcomes that focus on the development of H.264 fast algorithms.

The limited bandwidth during the access between hardware components and the external memory often becomes the bottleneck of the system performance. One of the solutions of encoding the high-definition video in hardware with limited resources is to employ a hierarchical subsampling structure with the parallel-processing hardware architecture. The main objective of this work is to maintain both the video quality and bit-rate while pursuing the gain from computational complexity reduction.

This thesis proposes a hierarchical H.264 fast motion estimation algorithm by decreasing the coding complexity in both spatial and temporal domains. In spatial domain, we utilize the hierarchical search method to decrease the search points. In temporal domain, we utilize the linear motion model to reduce the search range. The simulation results show that the proposed algorithm can reduce the computational complexity to as low as 1.80% compared to JM12.4 with less than 0.10dB video quality degradation.