The computational theory of visual shape developed by Kimia, Tannenbaum and Zucker presumes the existence of global, bounding contours, and is based on a hierarchical ``shock'' structure which corresponds to natural intuitions about parts, bends, and protrusions. However, early vision often only reveals fragments of these bounding curves. These fragments must somehow be grouped together, and any such grouping process will likely reflect object structure. In this project we are attempting to apply the KTZ theory to curve fragments, with particular emphasis on determining how the local properties of the shock structure can constrain the process of ``gluing'' together curve parts into closed curves.
J. August, S.W. Zucker
A novel method for blur estimation in which the scale of estimation is adapted to the local image structure allows depth segmentation in complex images where conventional Fourier-based techniques would fail.