Abstract: A new probabilistic background-foreground model based on Hidden Markov Models (HHMs) is presented. The hidden states of the model enable discrimination between Foreground and Background. This method is composed of two phases. First, an ICE (Iterative Conditional Estimation) algorithm is introduced to learn the unknown HMM parameters. In the second stage, each pixel is classified with an MPM (Maximum Posterior Marginal) classification algorithm. The potential and efficiency of the method have been proven through simulations under Matlab.
A.R. Debilou and S. Aouragh , 2006. An HMM-Based Model for Moving Object Detection. Asian Journal of Information Technology, 5: 1034-1038.