A Fast Face Detection Method Based On Skin Color And Adaboost

This work aims at designing a face detection system with high speed of processing and a reduced number of false positives compared to the well known Haar feature-based face detector developed by Viola and Jones. In this method, HS channels of HSV color space are logically combined for skin-color region extraction, then a blob-contour-property based filtering is used to indicate the optimum ROI (regions of interest) on the original image and finally Adaboost is given the face candidates ROI for face presence confirmation if any. Our experiment has shown that the processing time is reduced around ten times compared to Adaboost used alone in which the whole image needs to be scanned.