##Offline Signature Recognition and Verification

This is a Python-OpenCV implementation of a system that does offline signature recognition and verification.

###Requirements

• Python 2.7
• OpenCV 2.2
• python interface for OpenCV
• numpy

###The paper

For a similar system you can look at the following paper:
"Off-line arabic signature recognition and verification", M.A. Ismail, Samia Gad - Pattern Recognition (2000) Elsevier

###Signature data

I had no hope getting access to the data the Mr. Ismail used in his paper. So I started to search on the internet for some data and I found some puplically available on http://sigcomp09.arsforensica.org/NISDCC-preview/. There is a collection of 60 authentic signatures form 12 peoples, And 151 forged version for each persons signiture.

All those data are located on data directory.

####Naming convension

Each person has a 3 digit ID like 001, 002, etc. Each signiture regardless if it is authentic or forged is done 5 times, and to separate each signature a 3 digit id is appended to the end.

So the file 001001_000.png means: It's the 1st signature, which is signed by person 001 (first 001), which belongs to person 001 (second 001). So we know that this signature is authentic. And there is 4 more versions of it named 001001_001.png to 001001_004.png.

An example of a forged signature would be 021001_000.png. Which means person 021 tried to forge person 001's signature.

All forged examples are moved to forged directory in data\forged for simplicity.

###Code

All the codes can be found in src directory. These codes are meant to be a proof of concept, and are not ready for production.

####State of the code

All the things that I said above about what this project does, has not happended yet.

As of march 2012 I'm trying to implement the recognition part of the project.