Convex layers: A new tool for recognition of projectively deformed point sets

Abstract

A new more efficient algorithm for recognition of projectively deformed point sets is presented. It supposes the type of projective transformation, where the orientation of a triangle is preserved. First, convex layers of two compared sets are computed and left-tangents are found. If their structures are different, then an advanced algorithm is used. It creates the new structure of one set according to the other set. If the numbers of the points in the convex layers differ from each other, then the suitable points are moved to other layers, and if the left-tangents do not link the corresponding points, then their structure is corrected. The suitable combination of optimal structure correction criteria was investigated by a numerical experiment, the angle criteria were found as the best. Finally, five-point cross ratios of sequences of the points from both sets are computed and compared.