基于改进的Otsu算法的遥感图像阈值分割

Remote sensing image thresholding segmentation based on the modified Otsu algorithm

传统的Otsu算法仅仅适用于目标与背景分布均匀的图像,在处理遥感图像时具有一定的局限性。本文在分析传统的Otsu算法原理的基础上,结合遥感图像灰度级多、信息量大和边界模糊的实际情况,提出了一种改进的Otsu算法,用图像的方差信息代替均值信息计算最佳的分割阈值,实现遥感图像阈值分割。实验仿真结果表明,与传统的Otsu算法以及其它的一些改进的Otsu算法相比,本文算法具有明显的优越性。

The traditional Otsu algorithm only suit well-distributed images in target and background.When used in processing the remote sensing images,it exists some limitations.Based on the traditional Otsu＇s principles,this paper proposes a modified Otsu algorithm which combines the characters of the remote sensing image：more gray-scale,great information and fuzzy boundaries.In order to calculate the optimum threshold of the remote sensing image,this paper uses the variance information instead of the mean information.Compared with the traditional Otsu algorithm and other modified algorithms,the experimental result show that this algorithm has obvious advantages.