Offline signature verification is considered to be a difficult problem due to variations found in the signature, even from the same person. Although several approaches have been proposed, this problem is still open.
Mathematical morphology offers a wide range of tools for binary image characterization, including tortuosities, directions, curvatures, thicknesses and projections, as well as grey level and colour characterization methods.
In 2009, the CMM has submitted an offline signature verification system to the competition organized within the ICDAR conference, and has been ranked first with a 9% equal error rate. It should be noted that this method can take into account one or several reference signatures, as it is the case in forensic casework.
Some examples of the images provided in this competition are shown below. The first two signatures are genuine whereas the third one is a forgery.
If you are interested in collaboration, or want further information, please contact Etienne Decencière.