交互式分割模型与算法流程.doc

中图法分类号：TP391.41 文献标识码：A 文章编号：140312(2014) 论文引用格式： 结合相似与空间约束的图像分割 张峥嵘1，詹天明2，韦志辉3 1.南京理工大学理学院，南京, 210094；2. 江苏大学计算机科学与通信工程学院，镇江, 212013；3.南京理工大学计算机科学与工程学院，南京, 210094) 摘 要：图像中的目标一般含有很多子类，仅仅利用某个子类的特征无法完整分割出目标区域。针对这一问题，提出一种相似与空间约束的图像交互式分割方法。首先，通过手工标记的样本组成各个目标的字典，通过相似度量搜寻测试样本与各个目标的字典中最相似的再结合图像的空间约束，构建图像分割模型最后利用连续最大流算法求解，实现图像分割。关键词：图像分割；相似性搜索；空间约束；连续最大流算法 Image segmentation integrating the similarity fitting and spatial constraint Zhang Zheng-rong1, Zhan Tianming2, Wei Zhi-hui3 1.School of Science, Nanjing University of Science and Technology, Nanjing, 210094;2. School of Computer Science and Communication Engineering, University of Jiangsu, 212023; 3.School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, 210094) Abstract: Objective: The image cannot be segmented very well by using one feature in one subclass due to the objects in the image region contain lots of different subclasses. The sparse representation classifier (SRC), a combined result of machine learning and compressed sensing, is able to solve such problems by constructing a dictionary which contains all of the features of all subclasses. However, the performance of SRC depends on the over-complete dictionary, that means it needs lots of training samples to obtain a good segmentation result. On the other hand, SRC method cannot obtain good segmentation results when we only have less training samples. Aiming to solve this problem, an interactive segmentation method based on similarity search and spatial constrain is proposed in this paper. Method: Firstly, each object dictionary which contains different features from corresponding subclasses is designed separately by manual labeling. The whole dictionary is constructed by arranging these object dictionaries obtained by manual labeling in sequence. Then, the most similar feature of the test sample is searched in each dictionary. The fitting term is built by using the distance between the testing sample and the mo