On the Construction of a Color Model for Real-time Visual Moving Objects Segmentation

A color model, called RGB-Ellipse, is developed in this paper. The proposed color model takes advantages of CIE-L*a*b*94 and CMC(l:c) in determining color differences. Furthermore, the transformation between RGB and RGB-Ellipse is linear. As a result, we are able to manipulate the noise tolerance processing as well as the computation efficiency in dealing with color differences. By using the above features, we develop a real-time moving objects segmentation scheme. The developed segmentation scheme consists of two main steps: (1) region seed determination and region growing and (2) region-based change detection and background update. The results from a visual surveillance system are given to highlight the value of the proposed color model and moving objects segmentation scheme.