Coastline Extraction Based on SAR images Combining Markov Random Field and Mixture Model
To efficiently and accurately extract the coastline based on remote sensing images,a new coastline extraction algorithm combining Markov random field and mixture model is proposed in this paper.The propose algorithm is based on the statistical model theory.Due to the asymmetric and heavy-tailed characteristics of the statistical distribution of pixel reflection intensity in the same ground object of synthetic aperture radar(SAR)images,Gamma mixture model is used to build the probability distribution of pixel intensity in SAR images.Considering the spatial correlation of local pixels,Markov random field is utilized to model the distribution of the component weight of gamma mixture model to overcome the effect of speckle noise.Then,SAR image segmentation model is built by combining MRF and gamma mixture model.Finally,parameter estimation is achieved using expectation-maximization to extract the coastline.To verify the performance of the proposed algorithm,the Sentinel-1 SAR image for Beibu Gulf is selected for experiment.The experimental results show that the proposed algorithm can accurately extract the coastline.
coastline extractionSAR image segmentationMarkov random fieldgamma mixture modelexpectation-maximization