基于灰色Tsallis熵的SAR图像快速分割*

针对SAR图像斑点噪声及分割速度慢的问题，提出一种基于灰色理论和Tsallis熵的SAR图像快速分割方法。该方法首先对待分割图像进行小波变换，将表征图像概貌信息的低频部分重构为概貌图像，表征图像细节和边缘的高频部分重构为细节图像，并建立了相应的概貌—细节共生矩阵模型；然后利用灰色理论和Tsallis熵设计了基于该共生矩阵的灰色Tsallis熵模型，用于求解最优分割阈值；同时，为加快阈值搜索速度，引入群体智能中的粒子群优化算法。实验结果显示，新方法在抗噪性、分割速度和灵活性三个方面均有明显提高。

Fast SAR image segmentation method based on grey Tsallis entropy

Aiming at the speckle noise in SAR image and slow segmentation speed, the paper suggested a fast SAR image segmentation method based on grey theory and Tsallis entropy. In the method, after deduced an approximation image and a gradient image respectively from the origin image via wavelet transform, constructed their approximation-gradient cooccurrence matrix. On the basis of the matrix, designed a 2D grey Tsallis entropy model to locate the best threshold value via grey theory and Tsallis entropy. Additionally, introduced particle swarm optimization (PSO) to speed up the segmentation procedure. Some experimental results indicate that the new algorithm not only shortens the segmenting time obviously, but also ignores the disturbance of inherent speckle in SAR image and illustrates some flexibility in segmenting different objects.