A Fully Automatic Approach to Facial Feature Tracking Based on Image Registration

This paper presents a real time, fully automatic facial feature detection and tracking approach. The head pose and facial action is tracked by a modified Candide 3D wireframe model based on an improved image registration technique. An effective model shape and position initialization method is also proposed. Experimental results demonstrate that our system is accurate, robust and fast enough for common applications, even when there are great pose and expression variations.