Stroke matching for off-line signature verification based on bounding rectangles

One of the key tasks in computer vision is the identification and matching of coherent shapes in different images. In signature verification these shapes are represented by the strokes of the signature. Strokes may fall apart, close up, change or simply disappear among different signatures of the same signer. This paper proposes an algorithm to tackle the stroke matching problem using an area based method for finding and linking corresponding strokes and dynamic time warping to perform intra stroke matching. Using the above method the two dimensional image comparison problem can be reduced to a one dimensional case. The method was tested on the corpus of the 2004 Signature Verification Competition; experimental results are presented and evaluated. A generalized form of the algorithm could well be applied on other areas of computer vision.