顔輪郭曲線座標を用いた二次元変形関数

概要

論文の詳細を見る
For characterization or discrimination of a 2D-man-face image, it is usually examined a face-outline and features, such as eyes, nose, lip, and etc. A standard face-outline is easily obtained by averaging the outline curve function defined in a properly normalized domain. But it is difficult to obtain a standard feature, it has to represent and average displacement of features, moreover, it may be needed to average of their time variations. We represent the correspondence of one image to the other displacement of features as a 2D-warping function using the face-outline curve coordinate system. This coordinate system is spanned by analogues of face-outline curves and arguments around the origin. In this method, the 2D-face image is automatically normalized by the face-curve, and mapped on to the disc. Adaptation of this coordinate system not only saves calculation but also makes easy to determine the warping function. We describe successive procedures and examine an ability of this method.
社団法人電気学会の論文
2004-04-01