摘要
许多传统的图像分割方法都需要输入用户难以理解的参数,而且这些参数对于结果的影响又比较大。基于聚类分析技术的方法对参数不敏感,简单而且高效,但因其专注于对单一特征空间的划分而无法同时保持区域均一性和空间紧致性。尽管已经出现了许多改进的方法,如采用进行空间约束的聚类方法和使用其它保持空间紧致性的方法进行结果修正等,不过不同空间划分之间的协调、新的参数复杂性和算法复杂性反而使得聚类分析技术失去其简单有效的优势。给出一种新的医学图像分割算法,通过结合K均值方法和各向异性滤波技术,保持图像空间紧致性并解决过分割和图像噪声问题,同时弱对象也能够被提取出来。对比实验以及应用表明,该算法具有良好的分割结果和性能。
Many traditional image segmentation algorithms require input parameters which are hard to determine but have a significant influence on the segmentation result.Clustering techniques based methods are simple,efficient and insensitive to parameters but they focus on the feature space and can't keep the balance of region homogeneity and spatial compactness,Though a lot of effort has been devoted such as spatial guided clustering analysis and posteriors refinement with spatial constrained method,the weights of different feature space and complexity introduced by using other techniques are still waiting to be resolved.Here we propose another medical image segmentation method which combines the K-means method and curvature anisotropic diffusion filter to take care of the balance as well as over-segmentation and noise problem.And weak object will also be easier to be extracted.The experiments show encouraging results in applicability and reliability of the proposed algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第23期226-228,248,共4页
Computer Engineering and Applications
关键词
图像分割
聚类分析
各向异性滤波
弱对象分割
image segmentation
cluster analysis
anisotropic diffusion
weak object segmentation