关键词: 图像分割, 非下采样Contourlet变换, 模糊c均值聚类

Abstract:

In order to realize the unsupervised and automatic segmentation of SAR（Synthetic Aperture Radar） ima-ges and improve the accuracy and computational efficiency of segmentation,a segmentation method of SAR images based on NSCT（Non-subsampled Contourlet Transform） and FCM（Fuzzy c-Means） clustering is proposed.In this method,first,a denoising method based on NSCT is used to preprocess SAR images,which may protect the details of texture information.Then,the gray and texture features of SAR images are extracted by an edge-preserving extraction method of gray feature and gray-level co-occurrence matrix.Finally,the improved method of fast determining the clustering number is combined with the FCM clustering algorithm to realize the automatic segmentation of SAR images.Experimental results show that the proposed method is precise and effective in the unsupervised and automatic segmentation of SAR images.

Key words: image segmentation, non-subsampled Contourlet transform, fuzzy c-means clustering