Biometric hash: high-confidence face recognition

Discriminative Projection Selection Based Face Image Hashing

“…This technique is employed in Ngo et al's method [1], [2]. The raw face image hash vector f i, j elements are binarized with respect to a pre-determined fixed threshold as follows:…”

“…Ngo et al [1], [2] uses FLD as a feature extraction method which is applied before the random projection step in the algorithm. Their FLD transform is not user specific and aims to discriminate face images belonging to different users in the database.…”

“…We employ two different quantization methods: (1) Binary quantization (BQ) [1], [2] and (2) The proposed Gaussian mixture model (GMM) based quantization method. In our simulations, we employ these quantization methods separately to show the performance of the system.…”

“…In most of these applications, biometric templates are stored in databases and/or smart cards, thus raising questions such as data security and privacy. Various biometric hashing methods, which mostly depend on random projections, are proposed to protect the biometrics data [1]- [5] in the literature. See Fig.…”

“…Ngo et al [1], [2] employ feature extraction methods (i.e. Principle Component Analysis (PCA), Fisher Linear Discriminant (FLD), Wavelet transformation, Wavelet Transform with PCA, Wavelet Transform with Fourier Mellin Transform etc.)…”