A deformable template method for eye tracking on full face images is presented. The strengths of the method are that it is fast and retains accuracy independently of the resolution. We compare the me\$\backslash\$-thod with a state of the art active contour approach, showing that the heuristic method is more accurate.
|Title of host publication
|Den 14. Danske Konference i Mønstergenkendelse og Billedanalyse
|Published - 2005