Motion estimation using hierarchical phase plane correlation and block matching

摘要

Motion estimation comprising performing hierarchical phase plane correlation 602 on image data of video frames to generate plurality of candidate motion vectors, performing block matching 614 to determine matching block, and determining motion vector. Two level hierarchical phase plane correlation may be used comprising global level and local level correlation 604. Correlation may comprise dividing image data into regions 603. Global correlation may comprise four regions and involve downscaling global regions to standard region size. Phase plane correlation may comprise applying windowing function 605, Discrete Fourier Transform 606, cross power spectrum 607, inverse DFT 608, and Fast Fourier Transform to generate a correlation plane, and determining correlation of peaks 609 to determine candidate motion vector 610. The DFT may be implemented using a radix-2 FFT. Block matching may comprise a motion vector selected in a previous motion estimation iteration, possibly modified by a modification vector based on a heuristic algorithm, or MV selected by blocks neighbouring the individual block based on median filter of three MVs. Block matching may comprise evaluating sum of absolute differences (SAD) between blocks. Consecutive video frames may comprise luminance data. This method and system may be used in High Efficiency Video Coding (HEVC).