基于改进粒子群和自适应滤波的快速模糊聚类图像分割

Image Segmentation of Fast Fuzzy Clustering Based on Improved Particle Swarm Optimization and Adaptive Filtering

摘要:

为改进传统模糊C均值聚类(FCM)算法对初始聚类中心敏感、易陷入局部收敛、抗噪性差、计算量大的问题,提出一种新的基于改进粒子群算法的快速模糊聚类图像分割方法(PSOFFCM);方法首先利用自适应中值滤波对图像进行滤波处理,增强算法的鲁棒性；然后,将图像像素灰度值映射到二维直方图特征空间,作为聚类样本,优化FCM的目标函数,减少图像分割的计算量；最后,利用PSO算法代替FCM的梯度迭代过程,减弱了算法对初始聚类中心的依赖,同时增强全局搜索能力；实验结果表明,该方法不仅克服了FCM算法对初始聚类中心的依赖,而且抗噪能力强,收敛速度快,分割精度明显优于传统FCM。

Abstract:

The traditional fuzzy C-means clustering (FCM) algorithm is sensitive to the initial clustering center and is easy to fall into local convergence,and lacks enough robustness,and also has big computational cost. A fast fuzzy clustering image segmentation method based on Improved Particle Swarm Optimization (PSOFCM) is proposed to solve those problems. Firstly,the image is filtered by the adaptive median filter and the robustness of the algorithm is enhanced. Secondly,the gray value of the image pixel is mapped to the two-dimensional histogram feature space in order to reduce calculation and optimize the algorithm objective function of FCM. Finally,the PSO algorithm is used to replace the FCM gradient iterative process,which enhances the global search ability,and reduces the dependence of the algorithm to the initial clustering center. Experimental results show that this method overcomes the dependence on initial clustering center of FCM algorithm,which brings high robustness and segmentation accuracy,and has faster convergence speed.