Object recognition requires the selection of related image features for consideration by subsequent visual processing stages. This grant will support theoretical and experimental studies of image feature selection and linking in human and machine vision. Mathematical, psychophysical, and computational studies will be carried on in parallel the first year, with selection algorithms implemented in Lisp code that can be shared with other institutions. Further development, including implementation and testing on a Connection Machine, is planned for a second-year continuation. A third year of effort will integrate research results with those of synergistic projects funded by DARPA and ONR.
