Spatio-temporal correlation based fast motion estimation algorithm for MPEG-2

A fast motion estimation algorithm having computational complexity independent of search range is presented for MPEG-2 video encoding. The algorithm is based on spatial and temporal correlation of motion vectors. The algorithm yields good tradeoff between motion estimation distortion and number of computations. The speed up achieved by the algorithm is in the order of 160-250 and the SAD (sum of absolute difference) increase is 8-15 % compared to full search. The number of bits required to code motion vectors is less as the algorithm is based on correlation of motion vectors. Overall performance of the algorithm is compared with full search and other fast algorithms using speed-up, average SAD per MB, and PSNR measure.