Face Features Based Biometric Watermarking of Digital Image Using Singular Value Decomposition for Fingerprinting

In this paper, a semi-blind biometric watermarking scheme is proposed for fingerprinting application. Watermark is derived from face image using Principal Component Analysis. These face features are then embedded in host image using block-based watermarking scheme, which uses Singular Value Decomposition transform. This watermarking scheme works by initially dividing the original image into non-overlapping blocks, applying the SVD transform to each of them and subsequently embedding a watermark into the singular vectors. Each watermark value is embedded by modifying angles formed by the right singular vectors of each block with respect to some arbitrary plane. During embedding process the orthogonal property of the right singular vectors matrix is preserved. After extracting the features from watermarked host image it is correlated with face database features to get the approximately correct image. The robustness of this watermarking technique is tested by applying various attacks.