基于模拟退火算法的多阈值图像分割  

Multi-threshold image segmentation based on simulated annealing algorithm

作　　者：赵于前[1] 李慧芬[1] 王小芳[1] 

机构地区：[1]中南大学信息物理工程学院,长沙410083

出　　处：《计算机应用研究》2010年第1期380-382,共3页Application Research of Computers

基　　金：湖南省自然科学基金资助项目(09JJ3119);中国博士后科学基金特别资助项目(200902482);湖南省博士后科研资助计划资助项目(2008RS4026);湖南省国土资源厅资助项目(200718);湖南省科技计划资助项目(2009FJ3015)

摘　　要：针对多阈值图像分割时Otsu算法计算量过大的问题,提出基于模拟退火算法的阈值选取方法。首先依据最大类间方差准则,通过对直方图分析处理得到初始阈值向量,然后将该阈值向量作为初始解,利用改进的模拟退火算法逼近最优阈值向量。较之Otsu算法,该算法的计算量大幅减小,如实验中对图像的三阈值分割,运行效率提高了400多倍。结果表明,该算法能够快速、准确地实现多阈值图像分割。This paper proposed a threshold segmentation algorithm based on simulated annealing （SA） considering the vast computation of Otsu algorithm for multi-threshold image segmentation. At first,obtained an initial threshold vector through analyzing and processing of histogram. Then made an improvement of SA by setting this vector as the initial solution and approached an optimum threshold vector well applying this improved SA. Compared with Otsu algorithm,the proposed algorithm decreased the computation greatly. For instance, it increased the efficieney of tri-threshold image segmentation by over 400 times in the experiment. The experimental results demonstrate that this algorithm can cope with multi-threshold image segmentation efficiently and rapidly.

关 键 词：OTSU算法 初始阈值向量 模拟退火 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]