A variety of problems in low- and high-level vision are studied. The low-level vision (i.e. image processing) problems being addressed are edge detection, stereo correlation, contour grouping, image segmentation, and figure-ground discrimination. Various computational approaches such as genetic algorithms, simulated annealing, neural networks, and parallel and distributed processing are being investigated in the context of these low-level vision problems. In high-level vision, the current research is focused on the identification and localization of objects in range and intensity images from prestored CAD models. Efficient recognition and localization algorithms based on graph theory such as subgraph isomorphism and hypergraph monomorphism are being investigated. Issues related to efficient retrieval from large object model databases are also being addressed. In particular, hierarchical index and hash structures well suited for object models represented as attributed relational hypergraphs are being investigated. The research in low- and high-level vision is being applied to several application areas such as automated industrial inspection, geographic information systems and multi-media systems.
