INTERPOLATION-FREE SUBPIXEL MOTION ESTIMATION TECHNIQUES IN DCT DOMAIN

Currently existing subpixel motion estimation algorithms require inter polation of interpixel values which undesirably increases the overall complexity and data flow and deteriorates estimation accuracy. In this paper, we develop discrete cosine transform (DCT)-based techniques to estimate subpel motion at different desired subpel levels of accuracy in the DCT domain without interpolation, We show that subpixel motion information is preserved in the DCT of a shifted signal under some co ndition in the form of pseudophases, and we establish subpel sinusoida l orthogonal principles to extract this information. The proposed subp ixel techniques are flexible and scalable in terms of estimation accur acy with very low computational complexity O(N-2) compared to O(N-4) f or the full-search block-matching approach and its subpixel versions. Above all, motion estimation in the DCT domain instead of the spatial domain simplifies the conventional hybrid DCT-based video coder, espec ially the heavily loaded feedback loop in the conventional design, res ulting in a fully DCT-based high-throughput video codec. In addition, the computation of pseudophases is local, and thus a highly parallel a rchitecture is feasible for the DCT-based algorithms. Finally, simulat ion on video sequences of different characteristics shows comparable p erformance of the proposed algorithms to block-matching approaches.