中文摘要：

头发在人体表观中具有重要作用，然而，因为缺少有效的形状模型，头发分割仍然是一个非常具有挑战性的问题．本文提出了一种基于部件的模型，它对头发形状以及环境变化更加鲁棒．该模型将局部与全局信息相结合以描述头发的形状．局部模型通过一系列算法构建，包括全局形状词表生成，词表分类器学习以及参数优化；而全局模型刻画不同的发型，采用支持向量机（Support vector machine，SVM）来学习，它为所有潜在的发型配置部件并确定势函数．在消费者图片上的实验证明了本文算法在头发形状多变和复杂环境等条件下的准确性与有效性．

英文摘要：

Hair plays an important role in human appearance. However, hair segmentation is still a challenging problem partially due to the lack of an effective model to handle its arbitrary shape variations. In this paper, we present a part- based model, which is robust to hair shape and environment variations. The model combines local and global information to describe the hair shape. The local model is learned by a series of algorithms, including global shape word vocabulary construction, shape word classifier learning and parameter optimization, while the global model which depicts different hair styles is learned using support vector machine （SVM） to configure parts and define potentials for all underlying hair shapes. Experiments performed on a set of consumer images show our algorithm＇s capability and robustness to handle hair shape variations and complex environments.