Directional adaptive diamond search for motion estimation

Based on the research of motion vector distribution, a novel directional adaptive diamond search (DADS) algorithm was proposed. This algorithm adopted three strategies: the initial search point prediction strategy, the early search termination strategy, and the directional adaptive diamond search strategy. The initial search point prediction strategy made the predicted motion vectors close to true values and avoided them falling into the local optimal. The early search termination strategy detected the matching macroblock according to two fixed thresholds, and terminated the current search immediately. The directional adaptive diamond search strategy adopted the small diamond search pattern (SDSP) and four novel directional adaptive diamond search patterns (DADSP), and the search patterns were utilized adaptively based on the type of motion activity. Experimental results show that the proposed algorithm not only significantly reduces the average search points, but also improves the search accuracy to a certain extent, and it excels other traditional fast block matching algorithms in speed and veracity.