Face Biometric-Based Document Image Retrieval Using SVD Features

Abstract

Nowadays, a lot of documents such as passport, identity card, voter id, certificates contain photograph of a person. These documents are maintained on the network and used in various applications. This paper presents a novel method for the retrieval of documents using face biometrics. We use trace of singular matrix to construct face biometric features in the proposed method. K-nearest neighbor approach with correlation distance is used for similarity measure and to retrieve document images from the database. Proposed method is tested on the synthetic database of 810 document images created by borrowing face images from face94 database [1]. Results are compared with discrete wavelet transform features (DWT), which is counterpart of singular value decomposition (SVD). Proposed features in combination with correlation similarity measure provided mean average precision (MAP) of 75.73% in our experiments.