欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 图像技术 • 上一篇    下一篇

基于FCM和图割的交互式图像分割方法

  

  • 出版日期:2010-04-30 发布日期:2015-08-11

Interactive Image Segmentation Method Based on Fuzzy C-Means and Graph Cuts

  • Online:2010-04-30 Published:2015-08-11

PDF (PC)

280

被引次数

12

摘要： 为了提高在前景和背景颜色相似情况下图像的分割效果，提出了一种基于模糊C均值聚类（FCM）和图割的交互式图像分割方法。首先，利用分水岭算法对图像进行预处理，将图像分成多个小区域，用区域代替像素点进行分析。然后，采用模糊C均值算法对用户标记的前景区域和背景区域分别进行聚类分析，挖掘用户交互所提供的隐藏信息。用未标记区域的颜色分量到前景区域及背景区域类心的最小距离表示相似能量，用未标记区域与其相邻区域的相关性表示先验能量。最后，利用最大流/最小割算法求能量函数的全局最优解。与其他方法相比，该文方法具有较好的分割性能，能从前景背景相似的图像中较精确地提取感兴趣的物体，且用户操作简单。

关键词: 计算机应用, 图像分割, 模糊C均值（FCM）, 交互式图像分割

Abstract: In order to improve the segmentation performance when the foreground and the background are similar, an interactive image segmentation method based on fuzzy C-means and graph cuts algorithm is proposed. Firstly, the image is pre-segmented into a large number of small partitions using watershed algorithm which will be used instead of image pixels. Then, cluster centers of the foreground and background are calculated by the fuzzy C-means method. Finally, the likelihood energy is determined by the minimum distances from the unlabelled node to the foreground and background clusters, and the prior energy is determined by the relationship between the unlabelled node and its neighbor areas. The maximum-flow/minimum-cut algorithm is used to solve the global optimization problem. Compared with the results of other method, the experimental results show that our method can obtain a better segmentation performance with easy user interaction even though the foreground and background are similar.

Key words: computer application, image segmentation, fuzzy C-means, interactive image segmentation