Simultaneous Motion Detection and Background Reconstruction with a Mixed-State Conditional Markov Random Field

TOMÁS CRIVELLI; BRUNO CERNUSCHI FRÍAS; PATRICK BOUTHEMY; JIAN-FENG YAO

Marsella, Francia, Octubre 2008

Congreso; 10th European Conference on Computer Vision (ECCV'08), Springer-Verlag LNCS Volume 5302, Pages 113-126.; 2008

INRIA Rhone-Alpes, Francia

We consider the problem of motion detection by backgroundsubtraction. An accurate estimation of the background is only possible ifwe locate the moving objects; meanwhile, a correct motion detection isachieved if we have a good available background model. This work proposesa new direction in the way such problems are considered. The mainidea is to formulate this class of problem as a joint decision-estimationunique step. The goal is to exploit the way two processes interact, evenif they are of a dissimilar nature (symbolic-continuous), by means of arecently introduced framework called mixed-state Markov random fields.In this paper, we will describe the theory behind such a novel statisticalframework, that subsequently will allows us to formulate the specificjoint problem of motion detection and background reconstruction. Experimentson real sequences and comparisons with existing methods willgive a significant support to our approach. Further implications for videosequence inpainting will be also discussed.