Triplet markov trees for image segmentation

Jean-Baptiste Courbot (1, 2) , Emmanuel Monfrini (3, 4, 5) , Vincent Mazet (2) , Christophe Collet (2)

1 MOKAPLAN - Méthodes numériques pour le problème de Monge-Kantorovich et Applications en sciences sociales
2 ICube - Laboratoire des sciences de l'ingénieur, de l'informatique et de l'imagerie
3 TIPIC-SAMOVAR - Traitement de l'Information Pour Images et Communications
4 TSP - INF - Département Informatique
5 CNRS - Centre National de la Recherche Scientifique

This paper introduces a triplet Markov tree model designed to minimize the block effect that may be encountered while segmenting image using Hidden Markov Tree (HMT) model-ing. We present the model specificities, the Bayesian Maximum Posterior Mode segmentation, and a parameter estimation strategy in the unsupervised context. Results on synthetic images show that the method greatly improves over HMT-based segmentation, and that the model is competitive with a hidden Markov field-based segmentation.

Mots clés: Unsupervised segmentation; Triplet Markov Tree; Image Segmentation

Domaines: Traitement du signal et de l'image [eess.SP]; Applications [stat.AP]; Méthodologie [stat.ME]

Jean-Baptiste Courbot, Emmanuel Monfrini, Vincent Mazet, Christophe Collet. Triplet markov trees for image segmentation. SSP 2018: IEEE Workshop on Statistical Signal Processing, Jun 2018, Fribourg-en-Brisgau, Germany. pp.233-237, ⟨10.1109/SSP.2018.8450841⟩. ⟨hal-01815562⟩