Removal of High Density Gaussian and Salt and Pepper Noise in Images with Fuzzy Rule Based Filtering

What is it about?

This paper presents a robust and detailed approach to design a fuzzy filter for the reduction of noise in colored as well as black and white images. The filter consists of two stages. In the first stage, fuzzy derivative values for all the eight directions that is E, W, N, S, NE, NW, SE, SW with reference to the central pixel are calculated for determining noisy pixels. In the second stage, another fuzzy rule based system is employed. It uses the output of the previous fuzzy system to perform fuzzy smoothing by weighting the contribution of neighboring pixels. For a particular value of an adaptive parameter K, the fuzzy logic is iteratively used on the corrupted image till a desired value of PSNR comes. Experimental results have shown that the proposed algorithm works well not only for high density salt and pepper noise, but also for high variance Gaussian noise. The proposed filter outperforms the median filter for large density salt and pepper noise. Its performance in terms of PSNR values is comparable to Wiener filter, but it takes less time to show the same results, and hence provides less time complexity than Wiener filter. The results are compared using numerical measures (like PSNR, and MSE) and also through visual inspection.

Featured Image