Matting Algorithm, Based On A Static Background Differences For Synchronized Camera System

In a multi-PC/camera synchronized video-cameras system, separating the moving foreground object from the static background is the first step of real-time modeling of the user's virtual human body. Background subtraction technique has been widely used in many vision systems as a preprocessing step for object detecting and tracking. However, due to the high complexity of the environment's lighting sources in a real scene, as well as the uncertainty of static background and dynamic foreground in different scenes, how to implement the effective and stable matting and tracking is always what the researchers pursue.This paper thoroughly concludes all the used background subtraction algorithms and provides five major steps. These steps include related concepts, methods, rationales and the analyses and explanations from the perspectives of color spaces as well as human visions. Based on that, this paper implements two on-line background subtraction algorithms and their improvements from a static background scene. Moreover, both of them could detect the moving objects from a static background scene effectively and efficiently. In the mean time, the shadow and highlight areas could be detected and separated from moving objects. The first algorithm and its improvement are good at simple static background. They use a simple background model, sets of user defined or calculated thresholds and some kind of color attributes to implement a quick background subtraction algorithm. The second algorithm and its improvement are good at complex static background. They first develop a computed color model to separate the color difference to brightness and chromaticity components. Then through off-line training a series of background images taken for a period of time, a statistical background model and the thresholds are obtained.Finally, a pixel classification is proposed to perform the background subtraction on-line. The former method demands a relatively short time as a whole process. It is suitable to the indoor environment with clear background. While the latter one could adapt to the environment better such as slightly changing illumination. So it could be used in either indoor or outdoor environment. At last, based on the implementation of these two kinds of background subtraction methods, this paper implements real-time display platform. The user could choose a suitable method dynamically according to characteristics of scene and their demands.