MOTION SEGMENTATION IN RGB IMAGE SEQUENCE BASED ON STOCHASTIC MODELING

A method of motion segmentation in RGB image sequences is presented in details. The method is based on moving object modeling bq a six-varia te Gaussian distribution and a hidden Markov random field (MRF) framew ork. It is an extended and improved version of our previous work. Base d on mathematical principles the energy expression of MRF is modified. Moreover, an initialization procedure for the first frame of the sequ ence is introduced. Both modifications result in new interesting featu res. The first involves a rather simple parameter estimation which has to be performed before the use of the method. Now, the values of Maxi mum Likelihood (ML) estimators of the parameters can be used without a ny user's modifications. The last allows one to avoid finding manually the localization mask of moving object in the first frame. Experiment al results showing the usefulness of the method are also included.