Development of novel models for face recognition using orthogonal polynomial transforms and neural networks

Abstract

This thesis proposes models for automatic face recognition using newlinemomentbased orthogonal polynomial transforms and artificial neural newlinenetworks The major issues in automatic face recognition system are face newlinelocalization feature extraction and classification To strengthen the newlineperformance of automatic face recognition a feature invariant face newlinelocalization procedure is developed to separate the face from the background newlineIt uses the elliptical shape of the face to localize the face region newlineThree types of face recognition models are proposed in the work newlineThe first approach employs Generalized PseudoZernike Polynomial GPZP newlineand Tchebichef Polynomial TP for feature extraction to satisfy the minimal newlinerequirement of a face recognition system namely to overcome the curse of newlinedimensionality invariant to size and tilt of the face and invariant to natural newlinelighting condition newline newline