Abstract
In this paper, we propose a hierarchical three step searchalgorithm using pyramid hierarchy. HierarchicalMinkowski inequality is adopted to reduce the cost of thesum of absolute difference (SAD) computations. Atop-town procedure is developed to search motion vectorshierarchically and the search pattern used in thewell-known Three Step Search (TSS) is adopted. We alsopropose a hybrid motion estimation scheme by combiningthe proposed Hierarchical Block Sum Pyramid Three StepSearch (HBSPTSS) with the famous Block-BasedGradient Search (BBGDS) algorithms. Experimentalresults show that, compared to other existing searchalgorithms, the proposed hybrid search algorithm canachieve high computation efficiency for both slow-motionand fast-motion video contents while maintainingexcellent PSNR performance. It is thus suitable for wideapplication fields with real-time requirement.