Multiframe resolution enhancement methods are used to estimate ahigh-resolution video still (HRVS) from several low-resolution imagesequence frames, provided that objects within the video sequence movewith subpixel increments. Estimating accurate subpixel-resolution motionvectors is a challenging, albeit critically important component ofsuper-resolution enhancement algorithms. A Bayesian motion estimationtechnique is proposed which models the motion field with adiscontinuity-preserving prior. The method is related to Horn-Schunck(1981) optical flow estimation, except that the discontinuity-preservingprior can allow abrupt changes within the motion field without the useof line processes. Simulations compare the high-resolution video stillswhich result from using the subpixel motion vectors calculated by blockmatching and the proposed Bayesian motion estimation technique
展开▼
