关键词: 谱聚类算法, 联合模型, K-调和平均算法, 合成孔径雷达图像分割

Abstract: Due to the initialization-dependence of original spectral clustering, an initialization-independent spectral clustering on the joint model is proposed and then is applied to image segmentation. The joint model can make full use of the information, spatial adjacency information and feature similarity information included in the data and then a more precise clustering result can be obtained. And the introduction of the K-Harmonic Means algorithm (KHM) can overcome the initialization-dependence of original spectral clustering and thus a more robust clustering result can be obtained. Experiments on textural images and Synthetic Aperture Radar (SAR) images verify the validity of the proposed algorithm.

Key words: spectral clustering, joint model, K-harmonic means algorithm, SAR image segmentation