摘要
传统的K-Means算法在图像分割中只与特征向量有关,从而忽略了像素间的空间位置关系,因而分割模型是不完整的.本文利用Markov随机场描述图像像素间的邻域关系,引入拒绝度的概念到聚类目标函数中的同时,提出了初始类别及初始中心点的确定方法,提出了较为完备的基于Markov随机场图像分割算法.并通过实验验证该分割方法在效果及效率上的有效性.
Traditional K-means arithmetic model is half-baked because of the segmentation of the images is only concernedabout feature vectors but ignore the spatial relation between two of the pixels. This paper use Markov random field to describe theneighborhood system of the image pixels. Then the concept of ＂refusal＂ is introduced in the K-means arithmetic. Meanwhile, themethod of initial cluster number and center is also decided. Therefore the more integrity model of image segment arithmetic based on Markov random field is formed. The arithmetic is validated by experiment later.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第12期2700-2704,共5页
Acta Electronica Sinica
