基于图上随机游走过程的SAR图像图谱分割方法

Spectral segmentation of SAR image based on random walk on graph

摘要:

基于图上的随机游走过程,提出了一种新的合成孔径雷达(SAR)图像图谱分割方法。首先根据SAR图像的统计性质,构造了适合于SAR图像的相似度度量函数;然后在正交谱分解和矩阵逼近理论的基础上,建立了SAR图像图谱分割方法中分类数目的确定准则。最后实现了基于谱聚类的SAR图像分割,并通过实验证明了所提方法的有效性和应用潜力。

Abstract:

A novel spectral segmentation method of SAR image is presented based on the multi-step transition probability matrix of random walk on graph.First,a similarity measure criterion that adapts to the peculiarity of SAR image is proposed.Second,a method to select the cluster number is developed using biorthogonal spectral decomposition and matrix approximation.The effectiveness of the proposed scheme is validated by applying it to real SAR images.