A three-dimensional extension of the SUSAN filter for wavelet video coding artifact removal

This paper proposes two extensions of the smallest univalue segment assimilating nucleus (SUSAN) filter to three dimensions. The obtained filters, i.e. 2D+T and 3D SUSAN provide a viable solution to video post-processing with the goal of wavelet coding artifact removal. Both filter extensions incorporate motion estimation in order to apply the temporal filtering stage along the motion trajectories in order to take advantage of inter-frame correlation. Empirical results reveal the effectiveness of the proposed approach.