Skin-color modeling and adaptation

Abstract

Abstract. This paper studies a statistical skin-color model and its adaptation. It is revealed that 1 human skin colors cluster in a small region in a color space; 2 the variance of a skin color cluster can be reduced by i n tensity normalization, and 3 under a certain lighting condition, a skin-color distribution can be characterized by a m ultivariate normal distribution in the normalized color space. We then propose an adaptive model to characterize human skin-color distributions for tracking human faces under di erent lighting conditions. The parameters of the model are adapted based on the maximum likelihood criterion. The model has been successfully applied to a real-time face tracker and other applications.

Skin color-based video segmentation under time-varying illumination

“…Statistical approaches can be subdivided further into: parametric approaches [4,5,7,8,12,22,24,25] and non-parametric approaches [1,14,20,31].…”

“…A number of different color spaces have been used; however, normalized RGB [7,12,20,31] and HSV [8,14,22,24] are the most common color spaces used. It has been shown that in addition to being tolerant to minor variations in the illuminant, these color spaces also tend to produce minimum overlap between skin-color and background-color distributions [23].…”

“…Locating and tracking patches of skin-colored pixels through an image sequence is a tool used in many face recognition and gesture tracking systems [1,4,5,7,8,11,12,13]. Skin-color segmentation, particularly useful for its orientation and size invariance, is usually used for localization in early stages of these higher-level systems.…”

Video Segmentation for Markerless Motion Capture in Unconstrained Environments

“…While invariant to most movements, colours are highly susceptible to lighting conditions; shadows, sensor errors, light colour temperature and directionality of the light source, all contribute in the way colours are represented [32]. Yang et ah [33] have observed that despite environmental conditions, skin colours have a tendency to cluster within the RGB colour space. Specifically the clustering effect can be modeled using a Gaussian distribution (see Figure 2.2).…”

“…This colour property is said to follow Gaussian statistics [33]. By constructing an appropriate statistical model and applying a threshold to each pixel's membership to this model, identification of skin patches can be made.…”