Spatiotemporal Multiresolution Associated to MRF Modelling for Motion Detection
Abstract
A Markov Random Field (MRF) based motion detection algorithm in a spatiotemporal pyramid framework is presented. The originality of this approach consists applying multiresolution both in space and time in the same way. Hence, in integrating spatial and temporal information on larger neighborhoods, it is possible to deal with some difficult cases such as uniform moving areas or subpixel motion. Furthermore, the computational cost and the delay induced by such a method remain compatible with real-time applications.