Automatic object detection based on adaptive background subtraction using symmetric alpha stable distribution

Automatic detection of objects is critical to video tracking systems. One of the simplest techniques for detection is background subtraction (BS). BS refers to the process of segmenting moving regions from image sequences. The BS process involves building a model of the background and extracting regions of the foreground (moving objects). In this paper, we propose an extended cluster BS (CBS) technique based on symmetric alpha stable (SαS) distributions. The developed method functions at cluster-level as against the traditional pixel-level BS methods. An iterative self-adaptive mechanism is presented that allows automated learning of the distribution of the model parameters. The results of the CBS SαS algorithm on real video sequences show improvement compared with a CBS using a Gaussian mixture model.