Statistical Models for Skin Detection

Abstract

We consider a sequence of three models for skin detection built from a large collection of labelled images. Each model is a maximum entropy model with respect to constraints concerning marginal distributions. Our models are nested. The first model is well known from practitioners. Pixels are considered as independent. The second model is a Hidden Markov Model. It includes constraints that force smoothness of the solution. The third model is a first order model. The full color gradient is included. Parameter estimation as well as optimization cannot be tackled without approximations. We use thoroughly Bethe tree approximation of the pixel lattice. Within it , parameter estimation is eradicated and the belief propagation algorithm permits to obtain exact and fast solution for skin probability at pixel locations. We then assess the performance on the Compaq database.