Robust tracking and recovery of human body parameters is a vital input to applications such as motion capture, virtual conferencing, surveillance and innovative interfaces supporting gestures. Past approaches for tracking human bodies are either based on infrared sensors, magnetic markers or computer vision. Such a tracking system should be fast enough for real-time and less sensitive to noise. In addition, if the application is placed in a public space such as a cinema hall, there could be additional difficulties. Public spaces are inherently unconstrained environments where clothing, lighting, background, occlusion and reflectance can vary and so represent a significant challenge to tracking techniques. In this research, we develop a novel algorithmic technique that can be used to estimate 3-D (3 dimensional) joint positions for a human body in a public space. This approach is based on markerless vision-based tracking. Our results show that the estimates of the body parameters obtained are sufficiently robust to the changing environment of a public space.