CHOOSING THE BEST COLOR SPACE AND PROJECTION FOR SKIN DETECTION

RAMOS, H; FRERY, ALEJANDRO CESAR; LUCINI, MARIA MAGDALENA

Encuentro; IX Encontro de Modelagem Computacional; 2006

This paper presents a quantitative comparison of the performance of skin detection techniques using a diversity of input data. These data stem from several color spaces and different dimensionality reduction techniques. Among the dimensionality reduction procedures, we present here the use of a robust version of the principal components transformation recently proposed in the literature. We use a database consisting of four hundreed and fifty images containing faces in a wide variety of illumination conditions. The comparison is made by means of classification accuracy using a mixture of Gaussian distributions as the decision rule.