Akçay, H. GökhanAksoy, Selim2016-02-082016-02-082013http://hdl.handle.net/11693/27995Date of Conference: 24-26 April 2013Given a query region, our aim is to discover and retrieve regions with similar spatial arrangement and characteristics in other areas of the same large image or in other images. A Markov random field is constructed by representing regions as variables and connecting the vertices that are spatially close by edges. Then, a maximum entropy distribution is assumed over the query region process and retrieval of the similar region processes on the target image is achieved according to their probability. Experiments using WorldView-2 images show that statistical modelling of compound structures enable high-level and large-scale retrieval applications. © 2013 IEEE.TurkishImage retrievalMarkov random fieldSpatial arrangementsCompound structuresMarkov Random FieldsMaximum entropy distributionRemote sensing image retrievalRetrieval applicationsSpatial arrangementsSpatial data modelStatistical modellingMarkov processesProbability distributionsSignal processingImage retrievalConference Paper10.1109/SIU.2013.6531469