Fast Motion Estimation Using Spatio Temporal Filtering

In this paper a fast algorithm for motion estimation is presented. It models the temporal averaging of a group of frames as the spatial filtering of the reference one with a suitable Dirac comb function. This equality allows us to estimate a constant affine motion by comparing the phases of the FFT. Experimental results show that the proposed algorithm outperforms the available fast motion estimation techniques in terms of both quality and computational effort.