Markov random field approach to region extraction using Tabu Search

This paper describes a region extraction algorithm based on the concept of Markov random fields. Markov random fields (MRFs) are characterized by using a Gibbs Distribution which equates back to the MRF. A heuristically developed energy functional is presented and used with the MRF in an efficient and accurate manner. Since the MRF used in this work is defined using the polar coordinate system, a very large search space exists for radial lengths and sites. To aid in pursuing these radial sites, a combinatorial optimization technique known as Tabu Search is exploited. Also provided is an extensive empirical study on aerial imagery and parts detection, in addition to a final discussion and description of future work.