基于改进PSO算法的Otsu快速多阈值图像分割  

Fast Multilevel Threshold Method for Image Segmentation Based on Improved Particle Swarm Optimization and Maximal Variance

作　　者：彭正涛[1] 方康玲[1] 苏志祁[1] 

机构地区：[1]武汉科技大学信息科学与工程学院过程控制与信息技术研究室,湖北武汉430081

出　　处：《现代电子技术》2011年第6期10-14,共5页Modern Electronics Technique

基　　金：湖北省教育委员会基金项目(Q20091112)

摘　　要：为了确定图像分割的最佳阈值,基于改进粒子群优化算法,提出了一种快速多阈值图像分割方法。首先引入独立峰值将直方图划分为若干区域,然后在各个区域使用最大类间方差法得到优化的目标函数,用具有非均匀变异特性和雁群飞行启示的线性递减惯性权重粒子群算法对目标函数进行优化,得到分割的最佳阈值,并用该阈值对图像进行分割。将分割结果与常规最大类间方差法的多阈值分割结果相比较,证明该算法不仅可以正确地实现图像分割,并可使分割速度大大提高。To determine the optimal threshold in image segmentation, a new multilevel threshold method based on improved particle swarm optimization （PSO） is proposed. Firstly, the histogram was divided into several areas by the conception of independent peaks. Secondly, the maximum between-class variance （MV） method was used to get the optimization object function in each area. Thirdly, the object function was optimized by the non-uniform mutation and geese-LDW PSO, the optimal thresholds was obtained, and the image was segmented by the threshold. Compared with the basic MV algorithm, the experimental results show that the new method can realize the image segmentation well and improve the speed greatly.

关 键 词：图像分割 粒子群算法 非均匀变异 线性递减惯性权重 独立峰值 多阈值 最大类间方差 

分 类 号：TN911-34[电子电信—通信与信息系统]