Abstract: We utilized the advantages of single-instruction-multi-data (SIMD) techniques of very-long-instruction-word (VLIW) processors, adopted data packing methods to speedup SIMD instructions and multi-direction parallel searching technique as well as the storage format of interpolated pictures suitable for picture data reuse and finally proposed an effective implementation of ME algorithm based on HW/SW combined framework.We implemented H.263 and H.264 ME algorithms on TI's TMS320C64××and LILY Processor developed by Institute of Microelectronics of Tsinghua University.Experimental results show the searching speed of ME rises to 3x~15x.