Title: Markov object processes: From area-interaction to linear networks
Authors: Marie-Colette Van Lieshout - CWI/UT (Netherlands) [presenting]
Abstract: An overview of Markov object processes and their applications in image analysis will be given starting from the pioneering work in the 1980s on segmentation by means of Markov random fields. Influence zone based spatio-temporal point processes, deformable template models and sequential object processes will be presented. Such models are useful for higher level image analysis tasks including tracking and recognition. Turning to the intermediate level, we will discuss polygonal Markov field models and show that, on the one hand, discrete versions of such mosaics are dual to Markov random fields, and, in the other direction, that certain polygonal Markov field models can be seen as a non-overlapping marked point process. Finally, we will briefly indicate how to construct point processes on linear networks.