Reconstruction of Pen Strokes in Off-line Signature Verification

OData support
Dr. Kővári Bence András
Department of Automation and Applied Informatics

Handwritten signature has been one of the oldest biometric identification methods. It is still very widespread in everyday life, used on different checks, bills and contracts. But forgeries result in a great deal of financial damage, so proper understanding and effective automation of signature-based identification is rewarding. In the past few decades much research has been done on the subject.

Identification can be done by using online or offline methods. In case of online signature identification a digitizing tablet is used, and it provides lots of information about the signature: the exact position of the pen, the acceleration and pressure, both as functions of time. With this information the authentication or verification may be performed accurately, however the need for a special hardware is a huge drawback

Offline identification takes only a scanned picture of the signature, only this can be used for comparison and forgery detection. Benefits are that any signature could be analyzed, no special hardware is needed. The main disadvantage is the lower accuracy. In practical life, this method has greater importance.

The purpose of the Signature Verifier System developed at Department of Automation and Applied Informatics of Budapest University of Technology and Economics is to develop algorithms for satisfactory accuracy business purposes. In the system, the automated signature verifying consists of many elementary steps. This way, the different modules can be evaluated and optimized objectively and independently.

One of the critical steps of the system is reconstructing the strokes of the thinned signatures, namely the vectorization of the signature. This means the identification of the lines, which are drawn with one movement without raising the pen. A good vectorization algorithm helps much in signature verification.

My work was to develop such a vectorization algorithm for the Signature Verifier System. In my thesis, I explain in details the developing process, the operation of the algorithm, the achievable results and different possibilities for further enhancement.


Please sign in to download the files of this thesis.