一种两阶段区域生长的遥感图像分割算法

提高图像分割算法的精度与速度对遥感影像的解译工作具有重要意义。提出了一种基于区域生长的遥感图像分割算法,包括的两个步骤分别是局部最优合并和全局最优合并。第一步注重提高图像分割的速度,在其具体实现中引入了一个局部最优合并的阈值,以减少错误的合并;第二步侧重提高图像分割的精度,在其实现过程中利用了一种高级数据结构红黑树来提高搜索最优合并区域的速度。最后,利用模拟的遥感影像和Orbview\|3高分辨率影像开展了图像分割实验。利用一种监督的图像分割精度评价方法,定量评价了该方法的性能。实验结果表明:该方法在分割精度和分割速度方面都取得了令人满意的效果。

A Two Stage Region Growing Method for Remote Sensing Image Segmentation

How to improve the accuracy and speed of image segmentation algorithm that has a significant effect upon the interpretation of remote sensing images.In this paper,a region\|growing\|based method is proposed for remote sensing image segmentation.This method is composed of two procedures which are local the best merge and global best merge.The first step focuses on performing high\|speed image segmentation.In implementation,a threshold is introduced into the first step to reduce erroneous region merges.The second step aims at increasing segmentation accuracy.In order to raise the running speed,an advanced data structure and red\|black tree are utilized to implement the second step.Finally,simulated remote sensing image and Orbview\|3 high resolution images are used to carry out segmentation experiment.A supervised image segmentation accuracy evaluation method is utilized to quantitively validate the performance of our algorithm.The experiment result indicates that the proposed method can achieve satisfactory segmentation in terms of segmentation accuracy and speed.