Image smoothing using non-linear filters a comparative study

In this paper, four different non-linear filters, Kuwahara, Mean shift, Bilateral and SNN, in terms of noise immunity and edge preserving are compared. At first two types of noise (Salt and Pepper, Gaussian) are added to the images. Experimental results show that the mean shift can inhibit noisy edges, facilitated the detection for weak edges and have better result in images with non-uniform illumination compared with traditional methods such as: Canny, Sobel, Prewitt, Roberts. PSNR is also use for comparison four smoothing filters, the results show mean shift filter is better than another filters.