Face Detection Using the Improved Feature Tracker

In traditional feature tracker method (Lucas-Kanademethod), feature points have to found which doesn't have this feature in current scene by occlusion. In this paper developed a novel method for correct feature tracker. In proposed feature tracker, we used the estimation value at each pyramidal level. When the face was occlusion, we deleted these feature points, which is useless feature. This method is robustness to face rotation. It has two main modules, candidate face region detecting and tracking. For detect a face region in diversely conditioned image, we utilized skin color detector, which defines explicitly the boundaries of skin cluster in some color space. Then, tracking the region, we used the Harris corner detector and the improved feature tracker which has robustness for face rotation. In experimental result, we estimated the performance of face tracking algorithm and show its robustness in orientation.