Transform Domain Motion Estimation

摘要: This paper introduces an algorithm for estimating the displacement of moving objects in a television scene from spatial transform coefficients successive frames. The works recursively such way that estimates are updated coefficient to coefficient. A promising application this is motion-compensated interframe hybrid transform- dpcm image coding. We give statistical analysis domain estimation and prove its convergence under certain realistic conditions. An analytical derivation presented gives sufficient conditions rate be independent type. result supported by number simulation examples using Hadamard, Haar, Slant transforms. also describe extension adaptively updates according local features objects. Simulation results demonstrate adaptive has good properties even very noisy images.