Nonrigid motion estimation from a sequence of degraded images

In this paper the problem of online motion and deformation estimation from a sequence of degraded images is considered. More specifically we have considered the superposition of a rigid motion and an affine deformation, while the considered degradation consists in a blurring effect and an additive noise. The measurement equation is heavily nonlinear with respect to motion parameters. Therefore, in order to process the pixels of each frame of the sequence, a filtering procedure is proposed, which includes a linear approximation. The proposed procedure has been widely tested against simulated data and also with reference to an experiment with real data. The obtained results turn out to be quite satisfactory.