Shadow regions detection algorithm by adaptive narrowband two-phase Chan-Vese model

To obtain fast and highly accurate shadow regions detection results of sonar image, a detection algorithm of adaptive narrowband two-phase Chan-Vese model is proposed in the paper. The anisotropic second-order neighborhood MRF (Markov Random Field, MRF) is used to describe texture feature parameters of sonar image, and complete the noise smoothing. Initial two-class segmentation is determined by the block mode k-means clustering algorithm, to preliminarily estimate the approximate position of the shadow regions. Then, in order to reduce human intervention and improve the detection speed, zero level set function is adaptively initialized by approximate position of shadow regions. On this basis, the narrowband level set of two-phase Chan-Vese model is proposed to detect the sonar image and complete local optimization, which eliminates the global image's interference in detection results, makes shadow regions detection results more accurate. Experimental results demonstrate that the proposed algorithm can remove partial noise, improve the detection speed and accuracy, it has certain automaticity and adaptability.