Real-time Background Subtraction for Video Avatar

Background subtraction is widely used method ofdetecting moving objects in static background in computervision fields. In this paper, we present real-time backgroundsubtraction method for video avatar system or videoconferencing without blue screen. We observed the noise ofdigital camera and made a range matrix for backgroundmasking. Noise reductions were applied to both backgroundand mask using median filter blurring and open operation. Ourexperiment was performed using notebook PC and built-inwebcam. The results showed our approach is feasible toreal-time indoor environments.