Saliency-Based Adaptive Object Extraction for Color Underwater Images

Because of the special optical underwater imaging environment, the contrast and quality of images are affected severely, causing it difficult to extract objects from underwater images. An adaptive underwater object extraction method based on the saliency maps is proposed in this paper. Firstly, preprocessing method is utilized to improve the color contrast and quality. Then multi-scale image features are combined into a single topographical saliency map. The most salient image location and primary object are directed by the saliency map. By calculating the Bhattacharyya distance of salient features between the primary object and background, the adaptive weights of conspicuity maps can be obtained, as well more accurate object can be extracted from the fuzzy images. Experiment results show that the proposed method can not only detect the objects effectively but also extract more accurate object areas. It has a better performance compared with other algorithms.