Real-time Background Subtraction for Video Avatar

Abstract

Background subtraction is widely used method of detecting moving objects in static background in computer vision fields. In this paper, we present real-time background subtraction method for video avatar system or video conferencing without blue screen. We observed the noise of digital camera and made a range matrix for background masking. Noise reductions were applied to both background and mask using median filter blurring and open operation. Our experiment was performed using notebook PC and built-in webcam. The results showed our approach is feasible to real-time indoor environments.