An evidence theory supported expectation-maximization approach for sonar image segmentation

Abstract

In this paper an expectation-maximization (EM) approach assisted by Dempster-Shafer evidence theory (DST) for image segmentation is presented. The images obtained by a synthetic aperture sonar (SAS) are segmented into highlight, background and shadow regions for the purpose of shape feature extraction, which requires highly correct and precise segmentation results. The EM method of Sanjay-Gopal et al. is improved by using the gamma mixture model. Moreover, an intermediate step (I-step) based on DST is introduced between the E- and M-steps of the EM to consider the spatial dependency among pixels. Two combination rules of DST are adopted and compared, i.e. the Dempster rule and the cautious rule. Finally, numerical tests are carried out on both synthetic images and SAS images. The results are compared to those methods from the literature. Our approach provides segmentations with less false alarms and better shape preservation.