The hardware architecture of a novel motion estimator with adaptive crossed quarter polar search patterns for H.264 encoding

The advanced algorithms and corresponding hardware architectures are very demanded for the low-cost and high-performance motion estimation solutions. A hardware implementation for a novel H.264 motion estimator, with adaptive crossed quarter polar search patterns, is presented in this paper. Design trade-offs, including search patterns and memory accesses, have been made to target at very low implementation complexity. This hardware architecture is optimized for variable block sizes utilized in H.264 motion estimation. The architecture is mapped and verified with co-design techniques. The experimental results show that the proposed hardware motion estimator can sufficiently support the real-time 4CIF @ 30 fps video encoding running at 50 MHz, and yield an average PSNR of -0.05 dB, +0.34 dB and +0.11 dB when compared to the full search, diamond search and adaptive rood pattern search algorithms, respectively.