A Fast Face Detection Scheme Utilizing HSV-based Skin Color Model with K-L Transform

摘要

Fast face detection is very essential for many applications such as video surveillance and human computer interface. This study presents a fast face detection method using the HSV-based skin color model under non-constrained scene conditions. First, a skin color model that combines the HSV color space with K-L transform-based skin color filter is designed. Then, skin regions in the input image are detected and face candidates based on the spatial arrangement of these skin patches are generated. Finally, the mouths in all of the candidate human face regions are verified by a mouth detector. Experimental results show that the proposed scheme is not only fast and efficient, but also robust to head rotation.