Parallel Edge Localization and Contour Integration - Dynamic Receptive Fields

The paper describes an edge localization and contour integration technique that is parallel and modular in terms of the communication and computation structures. The technique can also detect corners and junctions accurately. Our technique is biologically motivated, in particular the biological behavior known as dynamic receptive fields or shortterm plasticity. Extensive experiments are conducted on both synthetic and natural data and demonstrate the effectiveness of the technique. keywords: grouping and segmentation, biologically inspired computational model, image features ? Research partially supported by University of South Carolina Parallel Edge Localization and Contour Integration Dynamic Receptive Fields