In this paper, we present a framework of body-part segmentation for a real human wearing multiple-colored suit, which has different colors for individual body parts. The segmentation framework would be very useful for estimating postures of a real human. Our work firstly runs a robust background modeling and detection method called codebook so that the human area is extracted from the background. Then, we apply the mean shift segmentation as well as edge detection to the extracted human area. Experimental results show that the proposed method achieves better performances than conventional segmentation methods.
展开▼
