Abstract – In this paper, a variational message passing framework is proposed for Markov random fields. Analogous to the traditional belief propagation algorithm, variational mes...
A quadratic programming formulation for multiclass image segmentation is investigated. It is proved that, in the convex case, the non-negativity constraint on the recent reported ...
In this paper we propose a novel technique to image completion that addresses image renaissance through a graph-based matching process. To this end, a number of candidate seeds wi...
Markov random fields are designed to represent structured dependencies among large collections of random variables, and are well-suited to capture the structure of real-world sign...
Tanya Roosta, Martin J. Wainwright, Shankar S. Sas...
Advanced surveillance systems for behavior recognition in outdoor traffic scenes depend strongly on the particular configuration of the scenario. Scene-independent trajectory analy...