Title: Signature verification is the process of determining the legitimacy of a handwritten signature' It i
1Offline Signature Verification based on Shape and
Density Feature
Presented by Tan Yng-Zing 40218409
Supervised by Prof. Kurt Kubik
Introduction Signature verification is the
process of determining the legitimacy of a
handwritten signature. It is a form of personal
biometric which can be used to identify an
individual.
- Methodology
- Steps in this methodology
- Data acquisition
- Genuine and forged signatures were collected and
further digitized into 115x60 array of 256 grey
levels. - Pre-processing
- Signatures are subjected to pre-processing to
remove any unwanted noise. The original image is
binarized, thinned and also converted to high
density images. - Feature Extraction
- Eight primary feature sets are categorized, and
the set that gives the best system reliability is
used. Some of the features extracted are
signature slants, baseline, width, area and high
density factor. - Distance measure
- Before comparison of signatures could be done, a
distance is computed from the differences in the
features of the signatures - Verification Process
- A threshold (VTH) based on the distance of the
cluster computed by the above equation is used. - If dist lt VTH, the sample is judged to be genuine
- If dist gt VTH, the sample is judged to be a
forgery
Objective To implement and improve upon an
existing offline signature verification system
which will not only accept genuine sample, but
also reject all falsification suggested.
Threshold
Conclusion In this thesis, 400 samples were
tested. The result is a system that can verify
most genuine samples and detect almost all kinds
of forgery. As shown in the graph above, both
classes genuine and forged samples, are
partitioned by the threshold value distinctively.
Background At times, whether a transaction could
go through or not depends largely on the
signature itself. Therefore, verification of the
signature is of great importance. There is also a
need to detect unskilled and skilled forgeries
with greater reliability.