Fuzzy-Based Adaptive Denoising of Underwater Images

Image processing as an aid for underwater vision has been subject to intensified interest in the recent years. The major problem is that they are inherently affected by poor contrast and noise due to the attenuation of light, backscatter in the underwater environment and the quality of sensing elements in camera during acquisition. Image denoising as a preprocessing step is needed for extracting features and accurate object recognition. Adaptive filters are preferred because traditional techniques often result in excess smoothing and fail to preserve edges while removing noise. A fuzzy-based image denoising algorithm is proposed to retain edge information and as well remove noise for restoration of underwater images. The adaptive nature of the proposed algorithm was tested using varying degrees of Gaussian noise. Performance metrics like peak signal noise ratio, normalized mean square error and mean structural similarity index were used for evaluation. Experimental results show the proposed method can remove varying levels of Gaussian noise better than the traditional filters while still preserving 27% edges.