Stroke extraction is a necessary part of the majority of semantic based off-line signature
verification systems. This paper discusses some stroke extraction variants which can
be efficiently used in such environments. First the different aspects and problems
of signature verification are discussed in conjunction with off-line analysis methods.
It is shown, that on-line analysis methods perform usually better than off-line methods
because they can make use of the temporal information (and thereby get a better perception
of the semantics of the signature). To improve the accuracy of off-line signature
verification methods the extraction of semantic information is necessary. Three different
approaches are introduced to reconstruct the original strokes of a signature. One
purely based on simple image processing algorithms, one with some more intelligent
processing and one with a pen model. The methods are examined and compared with regard
to their benefits and drawbacks on further signature processing.
