基于圆形区域定位的肖像类唐卡图像头饰分割算法  

Headdress Segmentation Algorithm for Thangka Portrait Image Based on Circular Region Location

作　　者：费翔[1,2] 刘传胜[2] 苗放[1] 

机构地区：[1]成都理工大学信息科学与技术学院,四川成都610000 [2]中国科学院遥感与数字地球研究所数字地球重点实验室,北京100094

出　　处：《电脑与电信》2013年第9期27-31,44,共6页Computer & Telecommunication

基　　金：国家科技支撑计划项目:虚拟旅游与文化资源协同系统研发与应用;项目编号:2012BAH 33F00

摘　　要：针对肖像类唐卡图像头光区域呈圆形的特点,将圆检测算法应用到肖像类唐卡图像头光区域的定位中,并在此基础之上提出了一种自动分割头饰的算法。该算法首先利用圆检测算法定位出头光所在的大概位置;然后利用头光颜色的空间分布特征、头饰外轮廓以及边缘检测结果实现头饰分割。实验结果表明,该算法在肖像类唐卡图像头饰的分割中具有头饰定位准确的特点,并且主尊比例和头饰大小的变化不会影响分割效果。Aiming at the characteristic of circular back-light of headdress in Thangka portraits,circle detection algorithm is applied to back-light region location of headdress in this article,and a new headdress segmentation algorithm is proposed basing on the region location method. The algorithm firstly uses the circle detection algorithm to locate the general location of back-light of headdress;then it uses the color spatial distribution of back-light of headdress,outer contour of headdress and the result of edge detection to segment headdress. Experiments show that the algorithm in this article is characterized by accurate location of headdress,and the proportion of main statue and the size of headdress change do not affect the segmentation effect.

关 键 词：头饰检测 图像分割 图像识别 圆检测 唐卡图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]