An efficient dual-interpolator architecture for subpixel motion estimation

New generation of video codecs typically adopts a sophisticated interpolation filter for sub-pixel motion estimation (ME). For embedded applications, it is crucial to reduce both the memory requirement and the computational complexity for sub-pixel ME. The key observation in this paper is that the interpolation filter for motion estimation and the filter for coding do not have to be the same. By adopting a simpler on-the-fly interpolation filter during ME process and a standard-compliant filter for coding purpose, both memory and complexity can be reduced with very little coding performance degradation. Since Hadamard transformed-SAD is often used in high quality codecs as a better matching measure, we further show that the ME interpolation filer can be combined with Hadamard transform for efficient VLSI implementation. Initial results show very promising performance of the proposed system.