Constraint reasoning for region-based image labelling
Carsten Saathoff · 2006
In this paper the authors present an approach to exploit explicit spatial knowledge for the labelling of image regions. The authors employ a two-step process, consisting of an initial labelling carried out by a segment classification algorithm and a second step refining the initial labelling based on knowledge about the spatial arrangements of concepts. For the second step the authors transform the initially labelled image into a constraint satisfaction problem and employ standard constraint reasoning techniques to ensure the spatial consistency of the labels. The authors provide an evaluation and some hints on future research directions.