Hand tracking in time-varying illumination

Tracking hand using a Webcam in daily environment always suffers from limited dynamic range and changing light conditions. This work presents a novel approach to generate steady hand segmentation from the videos captured by a Webcam in the time-varying illumination. Our approach consists of two parts: automatic gain control (AGC) during capture and motion of skin distribution estimating. A Markov model is exploited to estimate and predict the skin color distribution and camera parameters. We show that examples of segmented hand in a variety of lighting conditions. This method is used in our augmented reality map navigation system for bare hand control. The experiment shows this process can run in real time and the error rate is acceptable.