结合领域知识和多特征表示的唐卡破损区域分割算法-中南大学学报.PDFVIP

第 47 卷第 7 期 中南大学学报( 自然科学版) Vol.47 No.7 2016 年 7 月 Journal of Central South University (Science and Technology) July 2016 DOI: 10.11817/j.issn.1672-7207.2016.07.021 结合领域知识和多特征表示的唐卡破损区域分割算法 1, 2 1 2 胡文瑾 ，王维兰 ，刘仲民 (1. 西北民族大学 数学与计算机科学学院，甘肃 兰州，730030； 2. 兰州理工大学 电气工程与信息工程学院，甘肃 兰州，730050) 摘要：针对佛像类唐卡中出现的一致性破损区域的分割进行研究。首先对头光区域投影，利用一维函数对称性检 测方法得到图像的对称轴，提出基于对称轴的分块分割方法得到初始分割结果；然后利用 Gabor 变换提取纹理特 征，结合 Lab 空间颜色特征，构造多尺度多特征集合，最后采用K 最近邻分类算法(K−Nearest Neighbor, KNN)得 到一致性破损区域的模板。研究结果表明：该方法对于具有对称性的佛像类破损唐卡图像中出现的一致性破损区 域的分割效果良好。 关键词：唐卡；一致性破损区域；图像分割；镜像对称；多特征表示 中图分类号：TP391 文献标志码：A 文章编号：1672−7207(2016)07−2326−08 Damaged region segmentation of Thangka based on domain knowledge and multi-feature 1, 2 1 2 HU Wenjin , WANG Weilan , LIU Zhongmin (1. School of Math and Computer Science, Northwest University for Nationalities, Lanzhou 730030, China; 2. School of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, China) Abstract: A method was proposed to segment the consistent broken area of Buddha Thangka. Firstly, the head light area was projected and the symmetry axis was obtained by one-dimensional function symmetry detection method, and then the initial segmentation result was received based on the symmetric axis block segmentation. Secondly, the texture feature was extracted by Gabor transform, and the multi-scale features set was constructed combined with the Lab color space feature. The consistent broken a