Novel methods in denoising, resolution enhancement and object reconstruction of multidimensional signals

The presence of noise produces deficiencies during acquisition, broadcast or storage of the color image sequences. A principal problem here consists in a design of the noise reduction techniques while image content (edges, fine details, chromaticity characteristics, etc.) should be unchanged. There are many filters designed that are based on order statistics technique, on fuzzy logic theory, etc. The proposed technique in difference to other state-of-the-arts approaches employs the RGB channels data and fuzzy logic description of semantic properties of image features, processing several pixel gradients together in two temporal neighboring frames. A 3×3 sliding window located into a bigger 5×5 window novel framework is employed in an approach, applying the gradient values for neighboring pixels in eight different directions γ = (NW, N, NE, E, SE, S, SW, W) with respect to a central pixel. Two hypothesizes are resolved: the central pixel is a noisy or it is a free-noise pixel. The LARGE and SMALL fuzzy sets are introduced with an objective to estimate the noise contamination employing the Gaussian membership functions for membership degrees of gradient values.