Improved Deep Learning Approach For Underwater Salient Object Detection Using Spectral Residual and Fuzzy C-Means Clustering

The novel analysis that the underwater salient detection is the act of recognizing and emphasizing prominent and visually distinctive elements or objects within underwater images or films, assisting in tasks like marine research, underwater navigation, and resource prospecting. In this paper, we present a novel fuzzy c-means clusteringintegrated convolutional neural network (FCMC-CNN) for analysing the accuracy and robustness of saliency detection in difficult underwater situations. To enhance the object detectionperformance below the water’s surface, this method involves the application of both high-level CNN representations and low-level parameters. Spectral residual analysis (SR) approachand the proposed method are used for precisely locating salient objects in underwater photos. Although the difficulties imposed through variable spectral properties and low light conditions, this method requires to increase the accuracy of underwater prominent item detection. To gauge how well the suggested algorithm works, we simulate experiments using Python software. We assess the experimental results in terms of PSNR (44.2658%), SSIM (0.7726), FSIM (0.8369), and Average time (1.0852).This method demonstrated significant improvements in accurately recognizing underwater objects, enhancing the performance in detecting objects under water and possibly obtaining greater accuracy rates.