Open AccessUnderwater Sonar Image Segmentation by a Novel Joint Level Set Model
Author(s) -
Yue Wang,
Kefa Zhou,
Wei Tian,
Zhe Chen,
Daoyong Yang
Publication year - 2022
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2173/1/012040
Subject(s) - sonar , underwater , artificial intelligence , computer science , computer vision , speckle noise , segmentation , markov random field , synthetic aperture sonar , shadow (psychology) , image segmentation , noise (video) , joint (building) , pattern recognition (psychology) , image (mathematics) , geology , engineering , architectural engineering , psychology , oceanography , psychotherapist
This paper proposes a novel joint level set model for underwater sonar image segmentation. Combining features in points and regions in our novel joint level set (LS), it can achieve excellent performance for underwater sonar image segmentation. Regional information guides the model to locate the object of interest, whereas the point information accurately delineates contours. In addition, the unified Markov random field (UMRF) is taken to measure the neighboring relation between points and regions, which can overcome the problems of the high speckle noise, strong bias and low resolution of underwater sonar images. Our novel model can segment underwater sonar images into three partitions, such as the objects of interest, shadow and backgrounds. In contrast to current segmentation methods, outstanding results are demonstrated by our model. Moreover, another advantage of our model lies in its high efficiency.