Abstract: Most of the information related to the morphology ofgalaxies relies on the presence, shape and texture ofpatterns such as bars, rings, and spiral arms.Automatic classification of astronomical imagesrequires therefore that objects represented in theimage are indexed by the recognition of suchmorphological structures. We describe a methodology tosegment well-known images of galaxies in theirmorphological features in an automated fashion. Thedetection, description and classification of thesefeatures is achieved through the use of attributedrewriting systems that allow the identification ofstructures according to rules formalizing theinterpretation strategy followed by astronomers in thevisual classification of galaxies. A priori knowledgeabout the physical characteristics of the objects inthe image, as well as the image acquisition processcontrols the interpretation activity of the automatedsystem allowing the design of flexible algorithms thatcan be tailored to a wide range of images differing inresolution, noise characteristics, and objectorientation. Some results obtained applying the systemto CCD images of spiral galaxies are presented.!14
展开▼
