Real-time Detection of Between-the-Eyes with a Circle Frequency Filter
Shinjiro Kawato, Nobuji Tetsutani · 2002
In this paper, we propose a method to extract the center point between the eyes or Between-the-Eyesin real time, which can be a good clue in detecting other face feature points. In this research, we put emphasis on real-time processing. First, a skin-color region is extracted as a face candidate. Next, a circle-frequency filter, which we proposed before, is applied to this region to extract candidate points for Between-the-Eyes and also face rotation angles in the image plane at the candidate points. Each de-rotated local pattern at a candidate is compared with a template pattern of Between-the-Eyes,which is extracted from an image database of 400 faces, and a true candidate is selected. We tested the algorithm with the face database, and the false detection rate was 1.5%. We implemented the system on a PC with a Pentium III 866MHz CPU. It ran at 27 frames/secondwithout any special hardware.