Underwater image segmentation via dark channel prior and multiscale hierarchical decomposition

Underwater image segmentation is a key step for the analysis of the underwater target as segmentation quality will directly affect the stability and reliability of target recognition and tracking. A novel segmentation method is proposed in this paper that can help to solve the edge expansion and contour deformation problems in traditional segmentation methods. Firstly, the dark channel prior algorithm is applied to remove the “haze” and increase the object visibility in underwater images. Then, the multiwavelet kernels and multiscale hierarchical decomposition algorithm is used to transform the deblurred underwater images into binary images to reach the purpose of segmentation. The results show that our method is robust to noise and can capture more detailed information rather than object border or shape, which shows its feasibility and effectiveness for underwater image segmentation.