Automatic logo detection and extraction using singular value decomposition

Logo detection and extraction from a document image has found its importance in applications like logo based document image retrieval and logo based classification of documents. In this paper we propose a novel method for automatic logo detection and extraction from document image using mathematical tool Singular Value Decomposition (SVD). The method proposed is tested for publicly available realistic document image database Tobacco-800. We also compared the results of proposed method with current state of the art. Average logo detection rate achieved with proposed method is 89.52%.