Dynamic skin detection in color images for sign language recognition

Abstract

Skin detection is the first step of processing in many approaches to face and gesture recognition. This paper presents research aimed at detecting skin in digital images for Polish Sign Language recognition. There are many methods for detecting human skin, including parametric skin models defined in various color spaces and statistical approaches which require appropriate training. The presented method is based on statistical model updated dynamically for every image in which human faces can be detected. The detection is performed in luminance channel based on geometric properties of human faces. The experiments proved that effectiveness of this approach is higher than application of general skin detection models. © 2008 Springer-Verlag.