Figure 2
The two components of the graphical model comprise (a) a noise model, and (b) a Markov Random Field (MRF). The graph contains a directed component (noisy sensor) and an undirected component which represents the MRF. The figure shows cliques in the graphical model for an example RV Qu,v. The noise model specifies that the observed RV Ru,v is independent of all other RVs except its parent Qu,v. The MRF states that the marginal likelihood of Qu,v is independent of all other RVs in the graphical model except Qi,v and Qu,k. Each interaction RV in Q participates in a noise model clique, and multiple cliques in the MRF.