M. Kelly, M.D. Levine An approach for simultaneously extracting object information from images is presented. Annular operators are used to identify symmetric relationships existing between sets of edge elements. Operators are applied at multiple scales to edge data which have been extracted at multiple scales from a gray level image. Symmetric structures make it possible to identify where objects are positioned within an image, and are used as a basis for constructing coarse descriptors for the objects found in the scene. Annular symmetry operators represent a novel approach for interpreting image data. In contrast with previous approaches, using these operators makes it possible to segment image regions corresponding to objects in the scene. The result of this segmentation also produces a convenient means for describing the structure of objects in terms of the symmetric relationships between edge features. Whereas previous methods of regional analysis have been overly sensitive to small scale contour variations or gaps, our approach allows for the delineation of regional information based upon its spatial scale. It is also possible to compute symmetries when contours are fragmented. To our knowledge, no other approach has attempted to treat regional scale in such a direct way. By comparison, our method provides significant advantages that allow computer-based visual perception to be performed in practical settings. One important application for this work is to provide visual feedback for autonomous mobile robots. It is likely that in such applications, visual feedback would represent the principal means used to acquire knowledge about the environment. This visual knowledge should include two important components relating to objects in the robot's environment. The first is an estimate of the locations of the physical objects. The second is a means for representing these objects so that the robot is able to recognize and/or interact with them. Using annular symmetry operators, it is possible to address both of these issues simultaneously so that useful information pertaining to the location and structure of the objects can be computed. Currently we are working on a real-time implementation of this method so that it may be integrated into an autonomous robotic system.