Application of the SUSAN filter to wavelet video-coding artifact removal

This paper proposes the application of the Single Univalve Segment Assimilating Nucleus (SUSAN) algorithm for the reduction of coding artifacts in wavelet video-coding schemes. The SUSAN filter is as an image-structure preserving non-linear filter with low computational complexity. The filter adapts its bandwidth and stop-band attenuation based on a simple pixel classification procedure. In addition to this initial proposition, the paper provides two extensions of the original 2-D filter to the temporal domain: a two-stage 2-D+T SUSAN filter and a pure 3-D SUSAN filter. Experimental results provided in the paper provide substantial evidence of the effectiveness of the proposed approaches in terms suppression of artifacts that are typical in wavelet coding. Comparative tests show that SUSAN and its proposed extensions outperform state-of-the-art post-processing algorithms both in terms of objective (PSNR) and perceptual (PQS) quality comparisons.