Motion compensated video interpolation calculation that is time shifted (scaled) relative to desired output interpolation timings

摘要

In a method and apparatus for motion compensated video interpolation, each field or frame in a video image sequence is subdivided into plural blocks, and a motion vector field is derived using block (best) matching for a current video field against a previous video field or frame. A first time instance (e.g. t=-0.5, t=0.5) is determined at which an interpolated block is to be displayed, i.e. the `ideal' display time as set by the interpolation scheme required; and a second time instance (e.g. t=-0.75) is determined at which a corresponding interpolated block is to be actually created, i.e. a "working time instance", for improved interpolation accuracy, depending on motion vector errors. Video data is interpolated at its second (optimum calculation) time then displayed at the first (expected, ideal) time. Creation times may be a scale factor of the ideal, display times shifting the interpolation toward input frames - and determined independently for each block; scale factors may depend on motion estimator performance and scale factor clamping or thresholds (Figures 13, 14). Scale factor products or averages may be used. For very poor motion estimation, frame repetition to be used.