Otsu多阈值算法推广实现

遥感图像处理中的地物分割是分析和解译遥感图像的必要环节。本文分析了Otsu算法进行阈值分割的原理和有效性,针对Otsu多阈值分割算法计算效率低的缺点,提出了用Nelder-Mead单纯形法对Otsu的多阈值分割进行优化。将优化的Otsu算法与传统方法进行了比较,从理论分析和实验结果来看,此方法不仅能获得与传统方法分割一致的结果,而且应用到大尺寸遥感图像的多阈值分割中大大提高了分割效率。

Otsu multi-threshold promotion and realization of Otsu multi-threshold segmentation method

Landscape segmentation is necessary for analyzing and interpreting the remote sensing images. As the development of remote sensing technology, remote sensing images have higher resolution and more rich gray value. One way that we adequately use the information in the high resolution images is doing multi-threshold segmentation. An effective method for image segmentation, Otsu algorithm, based on threshold was analyzed in this paper. While the Otsu algorithm is very slow in processing the multi-threshold segmentation. So we optimized the Otsu algorithm with Nelder-Mead method. We compared the improved method with the traditional one. The method in this paper can speed up the segmentation of the big size remote sensing images