Parallel H.264/AVC Fast Rate-Distortion Optimized Motion Estimation using Graphics Processing Unit and Dedicated Hardware

Heterogeneous systems on a single chip composed of CPU, Graphical Processing Unit (GPU), and Field Programmable Gate Array (FPGA) are expected to emerge in near future. In this context, the System on Chip (SoC) can be dynamically adapted to employ different architectures for execution of data-intensive applications. Motion estimation is one such task that can be accelerated using FPGA and GPU for high performance H.264/AVC encoder implementation. In most of works on parallel implementation of motion estimation, the bit rate cost of motion vectors is generally ignored. On the contrary, this paper presents a fast rate-distortion optimized parallel motion estimation algorithm implemented on GPU using OpenCL and FPGA/ASIC using VHDL. The predicted motion vectors are estimated from temporally preceding motion vectors and used for evaluating the bit rate cost of the motion vectors simultaneously. The experimental results show that the proposed scheme achieves significant speedup on GPU and FPGA, and has comparable ratedistortion performance with respect to sequential fast motion estimation algorith