This report studies a statistical skin-color model and its adaptation. By quantitative analysis and goodness-of-fit test, we reveal that (1) skin-color differences among people can be reduced by intensity normalization, and (2) under a certain lighting condition, a skin-color distribution can be characterized by a multivariate normal distribution in the normalized color space. We then propose an adaptive model to characterize human skin-color distributions for locating human faces under different lighting conditions. The parameters of the model are adapted by a linear combination of the known parameters. The maximum likelihood criterion has been used to obtain the optimal estimation of the coefficients. The model has been successfully applied to a real-time face tracker and other applications.

29 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu