Real-time face detection based on adaptive downsizing and skin filter for HD camera

The paper presents a novel face detection scheme that enables real-time processing with low false-alarm rate in HD videos. The proposed method includes an adaptive image resizing based on the prior knowledge of detectable face size in minimum and a fast removal of non-facial region based on skin filter. There are three contributions of the paper; the first is fast processing by using an adaptive down-sizing scheme based on the minimum face size of texture-based face detector, the second is low false-alarm rate achieved by the use of skin filter to exclude non-facial regions, and the last is easy adaptation to existing face detection schemes in such that the proposed method outputs only a part of entire image region that may possibly include true faces.