Wang Wei-bin, Tian Jun-wei, Huang Yong-xuan, Liu Da-lie
Department of Plastic Surgery, Zhujiang Hospital, Southern Medical University, Guangzhou 510282, China.
Nan Fang Yi Ke Da Xue Xue Bao. 2006 May;26(5):667-9.
To find a new method for edge detection and modeling in frontal facial contour image.
Searching circle-based edge detection algorithm was developed on the basis of Sobel edge detector. Apriori knowledge of the facial contour and searching limitations as the minimum curvature radius, concave-convex property, and maximum edge disconnected distance were used to detect the edge of frontal facial contour. The frontal facial contour model was established with least squares curve fitting methods, and the relationship between the model rank and model precision was analyzed.
The edge detected by the new method was consistent with the actual edge of the facial contour and the irrelevant edge was well eliminated. Variation of the 2,4,10 time models from the actual image were compared, which identified the 10 time model as the best one.
The effectiveness and practicability of this new method for edge detection and modeling has been tested, which provides a theoretic basis for designing facial contouring image system.
寻找一种用于正面面部轮廓图像边缘检测与建模的新方法。
在Sobel边缘检测器的基础上开发基于搜索圆的边缘检测算法。利用面部轮廓的先验知识以及搜索限制条件，如最小曲率半径、凹凸特性和最大边缘断开距离，来检测正面面部轮廓的边缘。采用最小二乘曲线拟合方法建立正面面部轮廓模型，并分析模型阶数与模型精度之间的关系。
新方法检测出的边缘与面部轮廓的实际边缘一致，无关边缘得到了很好的消除。比较了实际图像的2次、4次、10次模型的变化情况，确定10次模型为最佳模型。
该边缘检测与建模新方法的有效性和实用性得到了验证，为面部轮廓图像系统的设计提供了理论依据。
