基于改进粒子群的模糊聚类超声图像分割

医学超声图像由于存在斑点噪声等模糊和不确定性的特点使得分割一直是一个难题。模糊C-均值聚类算法是一种结合无监督聚类和模糊集合概念的技术,广泛应用于图像分割,但存在着受初始聚类中心和目标函数高度非线性影响,极易收敛到局部极小的缺点。将集群智能的粒子群优化算法(PSO)与模糊C-均值聚类算法相结合,实现了基于粒子群模糊C-均值聚类的图像分割算法。实验结果表明,该方法具有搜索全局最优解的能力,因而可得到很好的图像分割结果。

Image segmentation based on improved Particle Swarm Optimization fuzzy C-means clustering

Segmenting ultrasound image is a difficult problem because of its intrinsic speckle noises and tissue-related tectures. Fuzzy C-means clustering algorithm (FCM) is an efficient algorithm and is commonly used in image segmentation. However it is sensitive to initialization and is prone to local minimum. This drawback can be alleviated by Particle Swarm Optimization (PSO), which possesses the effective ability of searching global optimal solution. Thus we propose a hybrid method combing the FCM and PSO. The numerical experiment results show that better results are obtained.