Recovering Shape Deformation by an Extended Circular Image Representation

Abstract

Interpretation of deforming shapes involves the qualitative description of deformation. A new representation of a closed 2D shape is introduced as a cyclic sequence of the extended circular images of the convex and concave segments of its contour. A measure of similarity between two segments that accounts for both shape and orientation is defined that depends on the correlation coefficient of their extended circular images. Dynamic programming based on this similarity measure is used to establish correspondences between segments of the two contours. Finally, a qualitative description of the deformation between the two shapes is computed by accounting for unassociated segments via curvature processes. The result is shown of the application of the deformation recovery scheme to dynamic cloud silhouette analysis using hand-traced input from real satellite images.