A MULTI-HISTOGRAM CLUSTERING APPROACH TOWARD MARKOV RANDOM FIELD FOR FOREGROUND SEGMENTATION.

This paper presents a Bayesian approach for foreground segmentation in monocular image sequences. To overcome the limitations of background modeling in dealing with pixel-wise processing, spatial coherence and temporal persistency are formulated with background model under a maximum a posteriori probability (MAP)-Markov random field statistical (MRF) framework. Fuzzy clustering factor was introduced into the prior energy of MRFs for the new implementation scheme, where contextual constraints can be adaptively adjusted in terms of feature cues. Experimental results for several image sequences are provided to demonstrate the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]