a signature is unique to everyone as finger prints this is because of the way it is signed, there different points where you have to check for similarities, they are the following

1. Strokes: you and i have a unique type of stroking technique if i draw a line i would go from bold line to thin line gradually which would look like a spear if its done in reality, where as you would have a continuous bold line.

2. lettering: for example "r" has different ways of writing, according to cognitive psychology when a human uses to way of writing a letter, he/she would think it as a artistic object rather than a letter, so there you would find mismatching if it is done again on a new paper.

3. calligraphy: i think about 15 to 20% of the worlds signature would at let have one calligraphic letter, you should also check on that.

4. computer vision: in computer vision using imaging processing technique you can use edge detection as a way to find the difference between two images, this could also help in your case.

hope this helps.

In temrs of image recorgnition I suggest you take a look on the reference below,

Turk M Pendatl A, 1991. Eigenfaces for recognition. J Cognitive Neiroscience 3(1):71-86

Transformation seems to be interesting. I have no

experience with handwriting but with OCR of

geometric figures that resemble characters.

Find the pixels of a single object (in your case,

perhaps, the whole signature) with floodfill or

similar. Now choose some hundred pairs of

points on the figure. Determine the distances.

Should work with the squares as well. Put the

distance into relationship to a maximum.

Create a histogram from these measurements.

Comparing objects is reduced to comparing

histograms.

Regards,

Joachim

I found it very useful to analyze the drawing pattern (i.e. we captured the movement of the pen instead of the final image). We found writing speed was more informative the shape for authenticity detection (if you imitate, you write slower).