Skip to Main Content
In this communication, two color-based methods for detecting faces in various scenes are compared and improved in terms of detection speed and/or accuracy. The first method according to Y. Raja et al. (1998) consists in automatically calculating a face color model by using the Expectation Maximization (EM) algorithm in the Hue Saturation (HS) subspace. Pixels with the color of the face are used by the EM algorithm to find the location of the face(s). The second method consists in using the histogram backprojection method for calculating the face color probability for every pixel in the image according to J.B. Martinkauppi and M.N. Soriano (2001), J.B. Martinkauppi et al. (2001) and M.N. Soriano et al. (2000), the location of the face(s) being then determined from this information.