Face Recognition Through Enhanced Algorithms and Techniques

Abstract

The research work presented in this thesis is mainly concerned with the upcoming newlinechallenge in face biometric i.e. recognizing faces altered due to plastic surgery. Recently, the newlineability of varied algorithm expresses intricacy in recognizing face invariant to plastic surgery newlinesince it has the characteristic of texture deviations of the skin. Even though plastic surgery is newlinelooked to be an inspiring concern in the domain of face recognition, the theme has to be newlinerestudied regarding theoretical and investigational perspectives. Further, this recognition newlineprocedure is extremely exploited for the persistence of authentication and security via analysis of newlineimage and computerization. newlineThe basic mode of face recognition comprises two stages include verification phase and newlineidentification phase. The earlier is used for matching two faces however the later is utilized for newlinematching the database with hundreds or thousands of face images and the demand face image. newlineMoreover, as per the researches and scientists, the plastic surgery face recognition of humans is a newlinechallenging trait, and hence this research work contributes three plastic surgery face recognition newlinemodels. newlineThe first model proposes plastic surgery face recognition approach with Volume- Scale newlineInvariant Feature Transform (V-SIFT) features and Support Vector Machine (SVM). The model newlineincludes various phases like pre-processing, feature extraction and classification phase. Here, VSIFT newlinedescriptor is for accurate feature extraction, and SVM classifier is for classification newlinepurpose. SVM is planned to adopt as the recognition system since it has certain unique newlinecharacteristics over other systems like misclassify. Further, the outliers in face image are newlineidentified with Random Sample Consensus (RANSAC). In SIFT feature, two main parameters newlinesuch as radius and Enlarge factor (EF) are there. The SIFT feature performance is analysed by newlinevarying these two main parameters like radius and EF. However, the model suffers from less newlineaccuracy in recognition faces. newlineThe research work proposes the nex