An Automated Level-Set Model Fusing Saliency Information for Sonar Image Segmentation

Abstract

Affected by the influences of various marine environments, sonar image is generally characterized by blurred target edges and uneven gray scale. Aiming at the difficulties of segmentation caused by such reasons, an automated level-set model integrating saliency information is proposed in this paper. This model includes two important parts: an automatic shadow removal algorithm based on pixels and a composite iterative segmentation strategy based on an improved level set method (LSM). First, shadows of the targets are extracted by color space transformation and replaced manually with pixels of the background area. Next, shadow removal is finished automatically by fusing saliency information from sonar image to reduce time complexity. Finally, a composite iterative strategy is proposed for sonar image with complex contents and blurred edges, where the initial contour of target is gradually optimized to the boundary of the target to achieve accurate segmentation. Qualitative and quantitative analysis experiments demonstrate that the proposed model has accurate target segmentation capability and is superior to other existing methods.