TEXTURED IMAGE SEGMENTATIONBASEDON SPATIALDEPENDENCE USINGA MARKOV RANDOM FIELDMODEL

Imagesegmentation isaprimary step inmanycomputer vision tasks. Although manysegmentation methods havebeenproposed inthe last decades, there isnogeneric methodthat canbeapplied ina great variety ofimages. Thisworkpresents anewimagesegmentation method using texture features extracted bywavelet transforms combined withspatial dependence modeled byaMarkovrandom field (MRF).Themethod initially produces acoarse segmentation, whichisrefined through arelaxation method based onanewenergyfunction. A setoftextured images isusedtodemonstrate the effectiveness oftheproposed method.