Data-dependent α-trimmed mean filters for image restoration

Data-dependent α-trimmed mean filters based on local statistics of signal are introduced and analyzed in this paper. The resulting filter is then time/space-varying, allowing it to adapt to different parts of the signal, e.g., it can effectively remove background noise from smooth areas of the signal, while preserving edges (with different filter parameters) in detail areas. The properties of the proposed filter are clarified, by comparing it with the modified trimmed mean filter. Simulation results are included to assess the effectiveness of the proposed adaptive structures