23 June 2015 Fuzzy c-means clustering algorithm incorporating neighborhood relations for synthetic aperture radar image segmentation
Author Affiliations +
Abstract
Bright and dark spots in synthetic aperture radar (SAR) images, a phenomenon called speckle noise, make SAR images difficult to segment. A robust fuzzy c-means clustering algorithm incorporating neighborhood relations (FCMNR) for SAR image segmentation is presented. First, target information is extracted based on the principle of probability maximization and the spatial correlation between neighboring pixels to estimate the degree of noise influence on image pixels, and then a tradeoff weighted factor incorporating neighborhood relations is built, which improves the performance of FCMNR by cutting down on the impacts of noisy pixels and preventing the misclassification of central pixels. It is worth noting that the new method avoids the selection of the empirically adjusted parameters (α,λgs) incorporated into all the other fuzzy c-means algorithms mentioned in the literature. Finally, in comparison with robust fuzzy local c-means clustering (FLICM) and fuzzy c-means clustering with local information and kernel metric (KWFLICM), experiments performed on synthetic and X-band TerraSAR images illustrate the excellent performance of the new method.
© 2015 Society of Photo-Optical Instrumentation Engineers (SPIE)
Jian Ji, Zhipeng Wu, Jingjing Huang, "Fuzzy c-means clustering algorithm incorporating neighborhood relations for synthetic aperture radar image segmentation," Journal of Applied Remote Sensing 9(1), 095076 (23 June 2015). https://doi.org/10.1117/1.JRS.9.095076 . Submission:
JOURNAL ARTICLE
12 PAGES


SHARE
Back to Top