ABSTRACT
Analysis and understanding of video sequences is an active research field. Many applications in this research area (video surveillance [18], optical motion capture [2], multimedia application [16]) need in the first step to detect the moving objects in the scene. So, the basic operation needed is the separation of the moving objects, called the foreground, from the static information, called the background. The process mainly used is background subtraction [12, 14, 24]. The simplest way to model the background is to acquire a background image that does not include any moving object. In some environments, the background is not available and can always be changed un-
for Video
der critical situations like illumination changes, or objects being introduced or removed from the scene. So, the background representation model must be more robust and adaptive. The different background representation models can be classified into the following categories:
• Basic background modeling: In this case, the background is modeled using an average [11], a median [42], or a histogram analysis over time [64]. Once the model is computed, pixels of the current image are classified as foreground by thresholding the difference between the background image and the current frame as follows:
d (It(x, y), Bt−1(x, y) ) > T. (5.1)
Otherwise, pixels are classified as background. T is a constant threshold; It(x, y), Bt−1(x, y) are respectively the current image at time t; and the background image at time t− 1. d(., .) is a distance measure that is usually the absolute difference between the current and the background images.
