Stroke Extraction and Stroke Sequence Estimation for Off-Line Signature Verification

B. Kovari, G. Kiss, and H. Charaf (Hungary)


signature verification, curve fitting, stroke recovery


This paper describes a method to extract stroke information from static two dimensional images. Our approach avoids the usage – and thereby the side effects – of morphological thinning and introduces a heuristic approach optimized for human handwriting. The algorithms structure makes it possible to isolate and independently implement key elements, such as ink detection or the location of junction points. A local approach is introduced to estimate the stroke sequence and thereby determine the relation of different strokes connected to junctions. Some issues of the algorithm are highlighted and – as far as possible – also eliminated. Finally an experimental validation of the algorithm is presented and evaluated. The recovered strokes can be used to recover temporal information from the images, thereby allowing existing off-line handwriting recognition and signature verification methods to be traced back to their simpler on-line equivalents.

Important Links:

Go Back