Motion Detection Using Wavelet Analysis and Hierarchical Markov Models

This paper deals with the motion detection problem. This issue is of key importance in many application fields. To solve this problem, we compute the dominant motion in the sequence using a wavelet analysis and robust techniques. So, we obtain an estimation of the dominant motion on several image resolutions. This method permits to define a hierarchical Markov model in a natural way. Thanks to this modelization, we overcome two problems: the solution sensibility in relation to the initial condition with a Markov random field, and the temporal aliasing. Moreover, we obtain a semi-iterative algorithm faster than using the multi-scale techniques. Thus, we introduce a fast and robust algorithm in order to compute the motion detection in an image sequence. This method is validated on real image sequences.