Efficient face-based feature recognition

Abstract

Previous work of the authorsshowed how multiply connected protrusions and depressions could be identified through the detection of bounding face-cycles. This method, as previously described, had two weaknesses. Fust, it assumed that the 2~D axis of the object was given by the human user. Second, identification of the face-cycles, using standard algorithms from graph theory, carried a high combinatorial overhead. This paper reportsfurtherwork which overcomes both of these problems, and illustrates a new cycle finding search strategy which has been implemented using the ACIS modeler.