Line segment distribution of sketches for Persian signature recognition

Abstract

A novel fast method for line segment extraction based on chain code representation of thinned sketches (or edge maps) is presented and exploited for Persian signature recognition. The method has a parallel nature and can be employed on parallel machines. It breaks the macro chains into several micro chains after applying shifting, smoothing and differentiating. The micro chains are then approximated by straight line segments. Length and position distributions of the extracted line segments are used to make a compact feature vector for Iranian cursive signature. The feature vector is invariant under affine transforms and can be used effectively in paperless office projects. Experimental results show fast response and accurate recognition/retrieval rate.