基于引导滤波和Tsallis熵的SAR图像分割算法

SAR image segmentation algorithm based on guided filtering and Tsallis entropy

针对合成孔径雷达(SAR)图像易受噪声干扰、分割方法精度低的问题,提出了一种基于频域引导滤波和Tsallis熵的SAR图像多阈值分割算法。利用非下采样Contourlet变换(NSCT)对图像多尺度分解,提取图像各方向的高频信息;通过引导滤波增强高频分量的边缘信息,在保持边缘的同时抑制了相干斑噪声;利用改进的二维Tsallis熵多阈值对增强图像精确分割。实验结果表明:分割算法对噪声不敏感,分割精度和适应性明显提高。

Aiming at problem that synthetic aperture radar（ SAR） image is susceptible to noise,precision of segmentation method is low,a novel SAR image segmentation algorithm based on multilevel threshold of Tsallis entropy in frequency domain guided filtering is proposed. Nonsubsampled contourlet transform（ NSCT） is used to decompose the source image into multi-scale to extract high frequency. Information guide filtering is used to enhance edges information of high frequency component and suppress speckle noise while preserving edges. Multithreshold selection algorithm based on improved two-dimension Tsallis entropy is used to make a precise segmentation for enhancement images. Experimental results show that the proposed method is insensitive to noise and effectively improves precision of image segmentation and adaptability.