2D Image pre-processing — Part II: Fuzzy 3D filtering

This paper proposes a VLSI circuit for the fuzzy 3D filtering algorithm based on two-stages processing of the image. In first stage, an average value is computed for all noisy pixels from the pre-processing window, which define the central pixel of this window. In same manner, a central pixel is computed for each window defined in given image. The proposed fuzzy 3-D interpolation algorithm (see Part I) is applied in second stage for all these central pixels. Using a knowledge-based model and taking account of its simplicity in implementation, the performances of fuzzy 3-D filtering are good. The filtered image obtained by interpolation of the central pixels can be computed in each node of grid used for image sampling or a high definition image can be implemented using image zooming algorithm. This filtering operation can be performed by a spatial smoothing circuit based on MOS transistors that have for all gates a spatial-dependent bias.