Feature-based human face detection

We propose face detection based on adopted skin color model, recently published by the author, in the absence of pripori information about number of faces, their locations and their sizes. The skin model is used to detect skin pixels. Then a merging procedure based on variance analysis is applied to segment the skin regions. Geometrical analysis and information analysis classify "face" or "not face" of the skin regions in the images. Geometrical analysis is based on that all the faces has approximately elliptical shape while the information analysis is based on the amount of skin region in the detected elliptical shape. This classification technique has the same performance for different sizes of faces also different orientation and inverted faces. We compare the performance of the proposed face detection with experimental results of work on cluster-based approach. The obtained results are comparable with the reported results. At the end the conclusion and future work are given.