A background subtraction for a vision-based user interface

In this paper, we propose a robust and efficient background subtraction algorithm for a vision-based user interface. We separate a user of interest as precisely as possible from the acquired images in order to convey the user's intension properly into the system through the vision-based user interface. Although the background subtraction techniques have been adopted in many vision-based interfaces to extract or track moving objects of interest in the images, they still suffer from the changes of lighting, such as shadows and highlighting. The proposed method removes effectively such interferences of lighting changes by exploiting pixel-wise statistical characteristics and threshold values in the well-known two color spaces (RGB and normalized RGB). According to experimental results, the proposed algorithm can be applied to various applications requiring real-time segmentation from the image sequences on the fly.