一种改进的Ostu图像分割法  

An Improved Method of Ostu Image Segmentation

作　　者：覃晓[1] 元昌安[1] 邓育林[1] 石亚冰[1] 元建[1] 

机构地区：[1]广西师范学院计算机与信息工程学院,广西南宁530023

出　　处：《山西大学学报（自然科学版）》2013年第4期530-534,共5页Journal of Shanxi University(Natural Science Edition)

基　　金：广西自然科学基金重点项目(2011GXNSFD018025);国家自然科学基金(6136307)

摘　　要：针对经典的Ostu算法存在的算法效率低,单一阈值不利于观察多峰值直方图的图像前景背景分割的不足,提出了模拟退火优化的阈值图像分割算法(Ostusa),利用模拟退火算法优越的全局寻优能力,提高Ostu算法的效率.实验结果表明,Ostusa在充分保证图像分割效果的前提下,大大提高了阈值计算效率,平均阈值计算效率提高了93%.同时,对于多峰值直方图的图像,Ostusa算法还有利于观察到不同层次的前景分割,得到更细致的图像前景边缘信息.这对于后续的边缘提取工作,有着重要意义.Traditional Ostu threshold is selected by exhaustive search algorithm,which is high time cost and low efficiency,and the single threshold can not reflect the foreground and background of multi-peak Histo- gram images. We introduce an algorithm that about image segmentation^s threshold optimized by simulated annealing（Ostu_ sa）,which is using the superior global optimization ability of simulated annealing algo- rithm to improve performance of Ostu algorithm. The experimental results show that, Ostu_ sa maintain a good image segmentation effect, while greatly improving the efficiency of the threshold value calculation, the average efficiency of threshold calculation is improved by 93%. At the same time for multi-peak histo- gram image,Ostu_ sa algorithm is also conducive to observed different levels of prospects segmentation, and obtain more detailed foreground of the image edge information, which is of great significance to the edge extraction work.

关 键 词：类间方差法 模拟退火算法 图像分割 阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]