Handwritten Signature Verification Dhawan, Ashish Ganesan, Aditi R. ECE 533 Project – Fall 2005
Introduction • Need for signature verification: • Signature: very common metric. • Types of verification: • Online - captures dynamic data. • Offline - uses features from the image. • Tough pattern recognition problem. • Types of forgeries: • Casual. • Skilled.
Pre-processing • Noise Removal: • Gaussian Noise. • Use of Average filter. • Inversion of Image. • Conversion of Image to Binary: • Use of Automatic Global thresholding.
Averaged and Inverted Image Original Image Thresholded Image
Geometric Features Extraction • Slant Angle: • Signature is assumed to rest on an imaginary line known as the Baseline. • The angle of inclination of the baseline to the horizontal is called the Slant Angle. • Center of Gravity. Original Image Baseline Rotated Image
Features Extraction • Aspect ratio: • Ratio of width to height of the signature. • Normalized Area: • Ratio of the area occupied by signature pixels to the area of the bounding box. Bounding box of the signature
Features Extraction • Slope of the line joining the Centers of Gravity of the two halves of signature image. Right Half Left Half
Verification and Results • Extracted features from Test-Images are used in deriving the mean values and standard deviations, which are used for final verification. • The Euclidian distance in the feature space measures the proximity of a query signature image to the genuine signature image of the claimed person. • If this distance is below a certain threshold then the query signature is verified to be that of the claimed person otherwise it is detected as a forged one.
Conclusion and Future Work • Conclusion: • The system is robust and can detect random, simple and semi-skilled forgeries. • A larger database can reduce false acceptances as well as false rejections. • Future Work: • Collection of larger database. • Addition of extra features. • Number of edge points: Edge point is a point that has only one 8-neighbor. • Number of cross points. Cross point is a point that has at least three 8-neighbors.