Simple Face-detection Algorithm Based on Minimum Facial Features

This article presents a algorithm for rapid and accurate face-detection. The algorithm detects human faces by the geometric correlations between locations of faces and hairs. Ranges of skin color are used to figure out possible face regions so as to initially localize the faces; furthermore, probable hair blocks in an image are determined by means of hair color spectrums. Grouped skin and hair blocks decide candidate face areas in light of the geometric relation. The accuracy of the single-face detection is higher than 92% with the frame-rate of 10 fps, if tested with a simple background and sufficient light source. The algorithm can be transferred from a PC to embedded devices, such as a DSP platform. This system-level implementing ability brings great potential to customized and reusable applications as well as miniature systems.