A Level Set Method With Heterogeneity Filter for Side-Scan Sonar Image Segmentation
Meiyan Zhang, Wenyu Cai, Wang Yu-hai, Jifeng Zhu
Abstract
Small underwater target detection from sonar images remains a challenging task. In this article, a novel level-set-based image segmentation algorithm combined with heterogeneity filter is proposed to segment target from original sonar images. The proposed method first uses nonlocal means filter to remove speckle noise of sonar image, and then applies super-pixel method to aggregate areas with similar texture, thus reducing computational complexity. In addition, two heterogeneity filters are used to eliminate heterogeneity in sonar images and enhance target contours. Moreover, the adaptive threshold is provided to obtain the rough contours of highlight and shadow areas. The level set method is further evolved on the basis of rough contours to obtain fine contours of underwater targets. Extensive experimental results verify that the proposed method has a better performance than that of the traditional sonar image segmentation algorithms in terms of false alarms, missing alarms, etc.