Skin Color Saliency Map Model

Abstract

In this paper, we proposed a skin color saliency map model to localize driver's head and hand area, in order to implement driver's activity monitoring model. For localizing driver's head and hand areas, we consider skin color saliency map, which is maximizing skin color region and minimizing background intensity, based on center surround difference and normalization algorithms with skin color filtered normalized color features. After constructing the skin color salience map, the driver's head and hand regions are selected by local maximum energy searching with generalized histogram projection method and four-directional region growing based entropy maximization algorithm to find suitable size of the driver's head and hand areas. Experimental results show that propose model can successfully localize driver's hand and head region.