Rapid background subtraction from video sequences

This paper presents a technique to detect the moving objects quickly and accurately by using a background subtraction algorithm called RaBS (Rapid Background Subtraction). The aim of RaBS is to distinguish between moving objects (referred as foreground) from the static one (referred as background) rapidly and efficiently. RaBS constructs for each pixel, a pixel model which is a collection of background samples taken in the past at the same location or in the neighborhood. The pixel model is constructed using the background pixels alone excluding the foreground pixels, if any. The current pixel is classified as foreground or background by comparing it with the corresponding pixel model. Before declaring a pixel as foreground it is passed through a shadow detection mechanism. If it is shadow then the current pixel will be classified as background but this pixel will not be used for updating the pixel model. The pixel model is then updated by using a random policy. Finally, the value of background pixel is propagated into the pixel model of a neighboring pixel. The update algorithm considers the pixels in the neighborhood also thereby exploiting the spatial dependencies. Spatial dependency makes the algorithm robust to camera oscillations. The algorithm relies exclusively on integer computations which makes it suitable for embedded applications.