Face tracking in color video sequences
Peter Gejguš, Martin Šperka · 2003
This paper presents a new extension of the KLT (Kanade--Lucas--Tomasi) face tracker, which is robust against loss of tracked points. Stochastic skin-color model is used for face segmentation in the frames contained in a video sequence. Ellipse-fitting is used for selection of the face candidates because of the elliptical face shape. The proposed method achieved real-time performance. The results seem very promising.