摘要
在经典的融合空间信息的模糊聚类图像分割方法中,图像像素的空间信息大,都采用正方形的邻域窗来获取。为了更好地分割出图像中的边界及细节信息,对不同形状邻域空间信息的模糊聚类图像分割进行了研究。在该方法中,首先采用圆形、三角形和菱形邻域窗获得图像像素的空间信息,然后分别将这三种空间信息引入到融合空间信息的模糊聚类图像分割中。Berkeley图像上的分割实验表明分别采用圆形、三角形和菱形邻域窗获得图像像素空间信息的模糊聚类图像分割方法在分割性能上要优于融合正方形邻域窗空间信息的方法。
In the classical fuzzy clustering image segmentation algorithms based on the spatial information, most spatial information is obtained by the square neighborhood windows. In order to better segment the boundary and details of the image, a fuzzy clustering image segmentation algorithm based on the spatial information with different shape neighborhood is proposed in this paper. Firstly, the spatial information is acquired by the circle, triangle and diamond neighborhood win-dows. Secondly, these three kinds of spatial information are introduced into fuzzy clustering image segmentation with spa-tial information. The segmentation experiments on Berkeley images show that the performance of fuzzy clustering image segmentation with circle, triangle and diamond neighborhood spatial information separately is better than the method with square neighborhood spatial information.
出处
《计算机工程与应用》
CSCD
北大核心
2015年第10期12-15,35,共5页
Computer Engineering and Applications
基金
国家自然科学基金(No.61102095
No.61202153)
陕西省自然科学基础研究计划资助项目(No.2012JQ8045
No.2014JQ8336)
陕西省科学技术研究发展计划资助项目(No.2014KJXX-72)
中央高校基本科研业务费专项资金资助(No.GK201503063)
关键词
图像分割
模糊聚类
空间信息
image segmentation
fuzzy clustering
spatial information
