Image and signal restoration using pairwise Markov trees

This work deals with the statistical restoration of a hidden signal using pairwise Markov trees (PMT). PMT have been introduced recently in the case of a discrete hidden signal. We first show that PMT can perform better than the classical hidden Markov trees (HMT) when applied to unsupervised image segmentation. We next consider a PMT in a linear Gaussian model with continuous hidden data, and we give formulas of an original extension of the classical Kalman filter.