M.A. Demidov, N.G. Volchenkov, A.K. Troitsky, A.A. Khramov

Face detection problem consists of two independent parts - to search supposed positions of face locations and confirm face presence into these positions. It is necessary to choose such face elements, which will be useful to solve this task. There are different methods, which use complex filters, shape analysis, color components etc. Eyes, forehead, nose and cheeks were chosen as elements to detect supposed face location in image. The approach based on specially formed local extremes in image was used to detect such facial elements. Local extremes are formed in the lightness areas, in the areas with the most high-frequency lightness changing into these areas and at the borders of these areas. Then local extremes corresponded to the same facial elements are grouped into global extremes. Considering image scale, real geometrical face dimensions, mutual facial element locations, face positions are detected on this basis.