Sar Image Segmentation Based On Triplet Markov Field

Synthetic Aperture Radar (SAR) images are high-resolution images. Because of the speckle noise in Synthetic Aperture Radar (SAR) images, it is more difficult to segment SAR image than the conventional optical images. Recently, segmentation methods for SAR images are feature-based method, structure-based method, model-based method, and so on, but there is no common one in segmenting different SAR images.In this paper, we mainly research the approaches of SAR image segmentation, which are based on the recent Triplet Markov Field (TMF) model, and the three aspects including are as follow:(1) We present a SAR image segmentation method based on the Gabor feature and Triplet Markov Field, which extracts the texture feature with Gabor wavelet transformation as the additional field in TMF to solve the issue that existing methods are so sensitive to the speckle noise. The experiments show that we can not only retain the accuracy of the image edges, but also improve the region consistency in the SAR image with simple edges and textures.(2) We present a fuzzy based Triplet Markov Field for SAR image segmentation, which use the fuzzy property of each pixel’s classification for fuzziness of label field in TMF. The segmentation results show that this method is effective especially for the segmentation of SAR images in improving the accuracy of edges.(3) We present a SAR images fusion segmentation method based on Triplet Markov Field, which segments images by merging the multi-scale segmentation result of the traditional Wavelet transformation-HMT model and the pixel one. It makes full use of spatial and multi-scale information of to improve the accuracy of segmentation results.