Face Localization In Color Images Using Dynamic Time Warping And Integral Projections

The face localization and the facial features extraction have received a great attention in the context of the human-computer interaction. In this paper a method for face localization is described. This method locate human faces in color scenes with some variations in the illumination and pose. The first task is segment the image in areas with skin color, the biggest area is selected, then the integral projections are extracted at this area; dynamic time warping is used to align these integral projections with some templates. The templates are created from an average face which was extracted of a set of images that contain just a face. If the alignment is good the area is declared a true face.