An Improved Method for Face Recognition with Incremental Approach in Illumination Invariant Conditions

Abstract

In this paper we propose an enhanced method with an acceptable level of accuracy for face recognition with an incremental approach in invariant conditions like illumination, pose, expressions and occlusions. The proposed method hold the class-separation criterion for maximizing the input samples as well as the asymmetrical characteristics for training data distributions. This enhanced approach helps the learning model to get adjusted the weak features inline with enhanced or boosted feature classifier for online samples. This enhanced model also helps in calculating feature loses during the training process of offline samples. For representing the illumination invariant face features local binary pattern (LBP) are extracted from the input samples and IFLDA is used for representation and classification. This modified algorithm with incremental approach gives the acceptable results by detecting and recognizing the faces in extreme illuminations varying conditions.