基于多尺度图像块的SAR图像无监督分割

The Unsupervised Segmentation of SAR Imagery Based on Multiscale Image Block

提出了一种基于多尺度图像块的SAR图像无监督分割方法。在利用高斯混合模型进行图像分割时,大多采用的是基于单个像素的分割方法,这种方法由于未考虑像素周围邻域结点的信息,分割精度往往不高。论文考虑到SAR图像具有很强的斑点噪声,为了更好地抑制斑点噪声对分割结果的影响,在多分辨分析的基础上提出了一种基于多尺度图像块的图像分割新方法。实验表明,这种基于多尺度图像块的分割较在单个像素下多尺度Markov模型的MPM分割好,分割精度有了较大的提高。

An unsupervised segmentation of SAR imagery based on Muhiscale image block is proposed.ln the segmentation of Gaussian mixture model,the method of single pixel value has been widely used,since it does not consider the neighborhood information of pixels and always gets poor precision.It is considered that there are a lot of speckles in SAR imagery,for segmenting imagery well,we propose a new method of imagery segmentation based on Multiscale block.Experimental results demonstrate the effectiveness of this new method is better than MPM segmentation of single pixel based on Multiscale Markov model.