Multiple Histogram Technique for Robust Skin Color Based Segmentation

With the spreading of the ubiquitous computing; a growing demand for vision based recognition and communication. The first step in the latter demand is the segmentation process. Pixel based segmentation techniques have been widely used in the field of pattern recognition. The actual information that can be extracted from pixel based techniques is the color. In this work statistical technique based non-parametric skin distribution method has been proposed for modeling skin color pixel by modeling histogram based skin color approach on several color models and then unifies all the models into single model to produce a superior model. The idea of the proposed work is inspired form Hasan [1] but in their work they applied parametric technique for modeling skin color distribution using Gaussian Mixture model (GMM). Three metric are used to evaluate system performance. The proposed system achieved 99.08 classification rate, and the output results proved that the proposed system outperforms on other systems that applied single histogram base skin color segmentation on each color model separately.