Quantitative Analysis of Kernel Principal Components and Kernel Fishers Based Face Recognition Algorithms Using Hybrid Gaborlets

Face recognition is a genre of biometric method used to discern the face of a person from a set of databases. Procuring a low-dimensional feature with aggrandized discriminatory power is of pre-eminent importance to face recognition. In this paper a face recognition analysis is suggested based on Hybrid Gaborlet and Kernel Fisher Analysis. Flustered SVD (Fsvd) focuses at deriving an illumination invariant image. DWT disintegrates the image into wavelet sub-bands while Hybrid Gaborlet extricates the facial features proficiently. KPCA helps in dimensionality reduction, KFA performs non-linear mapping and induces Fisher analysis. In the classification phase, the Nearest Neighbor method (KNN) and various distance classifiers are exploited to calculate the distances between prototype vectors and the corresponding stored vectors. The analysis results in excellent recognition precision using Dr. Libor Spacek and Caltech segmented databases with disparity in pose, illumination and facial expressions.