A new cross diamond search algorithm for block motion estimation

In block-matching motion estimation, search patterns with different shapes and sizes can very much impact the search performance. Based on a study of the search pattern and the motion vector distribution, a new cross-diamond search (NCDS) algorithm using the asymmetric diamond search pattern is proposed. In it, two asymmetrical search patterns are employed instead of other popular square-shaped or diamond-shaped search patterns. Simulation results demonstrate that the proposed NCDS algorithm outperforms other well-known fast block matching algorithms greatly. The performance is also better than the most recent cross diamond search (CDS) algorithm in terms of mean absolute distortion and the number of the search points.