In a first time, we have led a comparative study of different spatio-temporal filtering schemes for denoising image sequences.  We found that an adaptive spatio-temporal Wiener filter computed via 3D DFT allowed a good trade-off between computation time and restoration quality.  Before applying this filter, the image sequences must be compensated for motion. For estimating motion,  we use a 2D parametric motion model estimated via an incremental multiresolution technique developped byVista. 

This filter has been shown to perform good results when integrated as a preprocessing step in a super-resolution scheme.

We now want to develop models for simultaneous displacement estimation and enhancement of noisy-blurred image sequences, so that each phase benefits from the increased quality of the other one.