#### # Information Retrieval and Interpretation Sciences Amember of ETRO(Electronics and In.pdfVIP

A Mo del-Based Approach to the Automatic Extraction of Linear Features from Airb orne Images, using Mathematical Morphology and MRF theory A. Katartzis, H. Sahli Technical Rep ort: IRIS-TR-0062, VUB-ETR O department, 2000. IRIS: Information Retrieval and Interpretation Sciences A memb er of: ETRO Electronics and Information pro cessing IMEC Interuniversity Micro Electronics Centre Address: Antonis Katartzis VUB ETR O , Pleinlaan 2, 1050 Brussels, Belgium Tel.: ++32-2-629 2858 Fax.: ++32-2-629 2883 Electronic mail: akatarzi@etro.vub.ac.b e 0 A Mo del-Based Approach to the Automatic Extraction of Linear Features from Airb orne Images, using Mathematical Morphology and MRF Theory. A. Katartzis, H. Sahli 1 Intro duction The identication of linear features by means of digital image analysis is a generic task in the eld of remote sensing. These linear features can b e either roads, bridges, vegetation alignments or dierent geological formations. Their detection can b e used in several applications such as image coregistration, cartographic applications and geomorphologic studies. Several approaches have b een prop osed in the literature, most of them dealing with the problem of road extraction from either synthetic ap erture radar SAR images or optic visible range images. Usually they are based on two criteria: a local criterion involving the use of lo cal op erators and a glob al criterion incorp orating additional knowledge ab out the structure of the ob jects to b e detected. The metho ds based on lo cal criteria evaluate lo cal prop erties on the image by using either an edge or line detector [1] [2] [3] [4] or morphological op erators [5]. The p erformance of these metho ds can b e greatly increased by using techniques that intro duce some global constraints in the image analysis pro cess. These techniques lead to an optimal