A Robust Method of Facial Feature Tracking for Moving Images

Robustness of feature tracking is really important to recognize facial expressions and movements of human in the natural state. This paper presents a technique for a robust facial feature tracking for the human facial videos. Our technique first detects particular feature points by template matching, and then tracks the points in moving images by optical flow estimation. At the same time, the technique determines the reliabilities of the detected feature points. If the technique determines a feature point is not reliable, it executes template matching in the neighbor domain or estimates the position of the feature point according to the level of the reliability. By those steps, when the difficulty to detect the feature point is dissolved, we can detect and track feature points again appropriately and quickly. The technique can be applied to understanding of user's delicate feelings, conditions, or psychology.