High performance hardware architecture for half-pixel accurate H.264 motion estimation

In this paper, we present a high performance and low cost hardware architecture for real-time implementation of half-pel accurate variable block size motion estimation for H.264 / MPEG4 Part 10 video coding. The proposed architecture includes a novel half-pel interpolation hardware that is shared by novel half-pel search hardwares designed for each block size. This half-pel accurate motion estimation hardware is designed to be used as part of a complete H.264 video coding system for portable applications. The proposed architecture is implemented in Verilog HDL. The Verilog RTL code is verified to work at 85 MHz in a Xilinx Virtex II FPGA. The FPGA implementation can process 30 HDTV frames (1280x720) per second.