Initial Facial Contour Approximation via Ellipse Fitting From Skin Color
When applying Chan-Ves (C-V) model to segment a face from an image, the evolving result is always influenced by the initial position, especially in a real scenarios with complex background. An improved initial face position approximation method is proposed based on ellipse fitting from skin color. We first use a skin color model to compute the candidate of the face region, and then fit an ellipse according to the set of face region by lest squares method. Tests conducted on two face dataset have demonstrated that the accuracy of the detected initial face region at ratio above 95％, which lays a good foundation for face segmentation by the C-V model. Experimental results show that, with our estimated contour, C-V model could converge quickly and precisely.
skin color face contour C-V model level set
Xinxin Li Xun Gong Ran Xia
Jincheng College of Sichuan University Chengdu, China School of lnformation Science and Technology Southwest Jiaotong University Chengdu, China Institute of Computer Science and Technology Chongqing University of Posts and Telecommunications Ch
国际会议
成都
英文
1511-1515
2010-12-17（万方平台首次上网日期，不代表论文的发表时间）