New segmentation method of side-scan sonar image based on edge detection in NSCT domain

Aiming at the problems of strong speckle noise, low image contrast and segmentation difficulty of side-scan sonar image, according to the characteristics of side-scan sonar imaging, a new segmentation method of side-scan sonar image based on the nonsubsample contourlet transform (NSCT) domain edge detection and region growing is proposed in this paper. Firstly, the side-scan sonar image is decomposed in NSCT domain, the shadow is segmented by means of K-mean clustering algorithm in NSCT low-frequency domain. The difference of the two fine scale adjacent sub-band coefficients decomposed from the same coarse scale in NSCT high-frequency domain is searched, then the edge points of the image is selected according to the maximum absolute value of the difference; the edge fusion in intra-scale and inter-scale is carried out and the edge map is obtained. Finally, the target segmentation is completed using the edge based region growing method. The experiment results show that the proposed method has obvious advantages of fine noise immunity capacity, high classification accuracy and edge positioning accuracy in side-scan sonar image segmentation.