基于改进粒子群优化算法的Otsu图像分割方法  

Otsu Image Segmentation Method Based on Improved PSO Algorithm

作　　者：刘申晓[1] 王学春[1] 常朝稳[2] 

机构地区：[1]黄河科技学院信息工程学院,郑州450006 [2]解放军信息工程大学电子技术学院,郑州450004

出　　处：《计算机科学》2013年第8期293-295,共3页Computer Science

基　　金：河南省教育厅科学技术研究重点项目(12B510018);郑州市嵌入式系统应用技术重点实验室(121PYFZX177)资助

摘　　要：Otsu算法分割图像时不依赖于图像的内容,具有较好的适应性,但计算量大和实时性差的缺点限制了其应用。针对这一问题,提出一种基于改进粒子群优化算法的Otsu分割方法。该方法以Otsu算法中的类间方差作为粒子群优化算法的适应度函数,以当前分割阈值作为粒子的当前位置,以阈值更新速度作为粒子的当前速度,以粒子最优适应值的改进量作为惯性权重,在灰度空间动态搜索使类间方差最大的阈值。实验结果表明:该方法能获得与经典Otsu相当的分割效果,而且显著地缩短了分割时间,算法效率更高。The Otsu image segmentation algorithm has good adaptability due to its contents-independent characteristics. However, its shortcomings like large amount of computation and poor real time quality have limited its application. To solve this problem,we proposed a new segmentation algorithm using the principle of Otsu based on an improved PSO algorithm. Taking the class-between variance of Otsu as the fitness function of PSO, the current segmentation threshold as the particle＇s current location, and the updating speed of threshold as the particle＇s current speed, and using the im- provement of particle＇s best fitness value as the inertia weight of PSO, the proposed algorithm searches for the thre- shold which makes the maximum value of the class-between variance in grey space dynamically. The experimental re- sults show that the new algorithm can get segmentation result which is equal to the classic Otsu, significantly reduces the time of segmentation process and also has higher efficiency.

关 键 词：图像分割 OTSU 粒子群优化算法 惯性权重 适应值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]