An approach to Illumination and Expression Invariant Multiple Classifier Face Recognition
Hidangmayum SaxenaDevi, Dalton Meitei Thounaojam, Romesh Laishram · International Journal of Computer Applications · 2014
Face recognition is still a challenging task due to the different problems such as pose, illumination, expression and occlusions.Due to these problems, it needs an extensive research.Various face recognition techniques exists but they suffer from one or the other limitations.One technique could not provide robust solution.Therefore, a combination of different face classification techniques can lead to robust solution.Here, the individual output of three different classifiers PCA, KPCA and Fisher face are normalized and combined using SUM rule, with their individual normalized matching score as the input feature vector.The system is tested on AT&T Face Database in which our proposed system is robust against illumination and expression as well as increasing the recognition performance.