Nonuniform Image Motion Estimation from Noisy and Blurred Image Sequences

AbstractThis paper develops, presents, and evaluates a new iterative algorithm to estimate image motion from noisy-blurred image sequences. This algorithm seeks the maximum-likelihood estimate of image motion coefficients in an arbitrary transformed domain directly from noisy-blurred data. The range of the convergence parameters that control the speed of convergence is also presented. Simulation experiments demonstrate the validity of the motion estimator. Additional experiments investigate the algorithmu0027s sensitivity to the extent of blur and noise. We also show the potential application of the estimated motion to image sequence restoration.