A Highly Data Reusable And Standard-Compliant Motion Estimation Hardware Architecture

Xing Wen,Oscar C. Au,Jiang Xu,Lu Fang,Run Cha,Jiali Li
DOI: https://doi.org/10.1109/ICME.2010.5583300
2010-01-01
Abstract:Motion Estimation (ME) is the most computationally intensive part in the whole video compression process. The ME algorithms can be divided into full search ME (FS) and fast ME (FME). The FS is not suitable for high definition (HD) frame size videos because its relevant high computation load and hard to deal with complex motions in limited search range. A lot of FME algorithms have been proposed which can significantly reduce the computation load compared to FS. Though many kinds of hardware implementations of ME have been proposed, almost all of them fail to consider about the motion vector field (MVF) coherence and rate-distortion (RD) cost which have significant impact to the coding efficiency. In this paper, we propose a hardware friendly ME algorithm and corresponding highly data reusable hardware architecture. Simulation results show that the proposed ME algorithm performs better RD performance than conventional FME algorithm. The proposed reconfigurable ME hardware is implemented in VHDL and mapped to a low cost Xilinx XC3S1500 FPGA. It works at 100MHz and is capable to process 1920 x 1080 of 30fps video format in real time and have very high data reuse ratio.
What problem does this paper attempt to address?