Document image retrieval using signature as query

Abstract

In this paper, we presents a new approach to document image retrieval based on signature. The database contains document images with English text combined with headlines, ruling lines, logo, trade mark and signature. In searching a repository of business documents, task of interest is that of using a query signature image to retrieve from a database. The signature retrieval task involves a two step process of extracting all the signatures from the document and performing a match on these signatures. We have proposed to use the rotated complex wavelet filters (RCWF) and dual tree complex wavelet transform (DT-CWT) together to derive signature feature extraction. In retrieval phase, Canberra distance measure is used. Further, relevance with query point movement (QPM) is used to improve the retrieval performance. Experimental results shows that document image retrieval rate of proposed method is satisfactory.