基于多尺度特征融合的SAR图像分割

摘要

SAR image segmentation is complicated due to the multiplicative nature of the speckle noise in SAR images.An SAR image segmentation method based on the multi-scale feature fusion is proposed in this paper. The fast discrete curvelet transform is applied to extract the image texture features,and the stationary wavelet transform is applied to extract the image statistical features. These two multi-scale features are fused to obtain a high dimensional feature vector. The fuzzy C-means clustering is used to segment the image. Experiments are carried out using typical noise-free image corrupted with simulated speckle noise as well as real SAR images,and the results show that the proposed method performs favorably in comparison to the methods based on the wavelet transform only. The proposed segmentation method can delete lots of small fragments in the homogeneous regions and obtain more accurate and smooth boundaries.%由于存在相干斑噪声的影响,给SAR图像分割造成很大的困难,提出一种基于多尺度特征融合的SAR图像分割方法.该方法利用快速离散curvelet变换提取图像的纹理特征,利用平稳小波变换提取图像的统计特征,将两种多尺度特征融合成高维的特征向量,采用模糊C均值聚类的方法进行分割.在仿真SAR图像和真实SAR图像的分割实验结果表明,提出的方法优于单独采用小波变换进行SAR图像分割的方法,在消除均质区内碎块的同时,使得边界更为精准和平滑.