Off-Line Handwritten Signature Retrieval

Abstract

This paper presents a new method for off-line handwritten signature retrieval using rotated complex wavelet filters (RCWF) and dual tree complex wavelet transform (DT-CWT) jointly for extracting details in twelve different directions of the signature image. This paper addresses issue in the context of a database of handwritten signature images and describes a system for similarity retrieval. The performance of the system has been tested with a signature image database of 192 signatures. From experimental results, we achieved a recall of 73.95% when considering the top 12 results. The results indicate that the proposed system is able to retrieve signatures with high accuracy even when a part of a signature is missing.