Reconfigurable hardware implementation of an improved parallel architecture for MPEG-4 motion estimation in mobile applications

摘要

A reconfigurable hardware implementation of a high-parallel architecture for MPEG-4 motion estimation is proposed in this paper. It possesses the characteristics of low power dissipation and low cost, thus primarily aiming at video-based mobile applications. The architecture employs a dual-register/buffer technique to reduce preload and alignment cycles and a high-parallel pipeline to reduce power consumption of redundant memory access. As an example, a content-based full-search block-matching algorithm has been mapped onto this architecture using a 16-PE array. This has the ability to calculate the motion vectors of 20 fps QCIF video sequences in real time at 8.2 MHz clock rate with 36.7 mW power dissipation using Xilinx Spartan II FPGA.