Periocular region-based age-invariant face recognition using local binary pattern

Abstract

The performance of the biometric face schemes suffers severely due to the variation in the subject’s aging. Designing the face recognition systems which are invariant to the aging process is challenging as the age patterns are different for the different individuals and also limited databases are available. The aging-based face recognition is still an open challenge for researchers as none of the existing methods are on par with human ability in recognizing the similarity across two faces. In the proposed paper, the age-invariant features of the face are derived using the local descriptor, local binary pattern (LBP). Preprocessing techniques like enhancement and denoising are applied to the images to enhance the accuracy of the designed system. Chi-square distance is used as a classifier to find the matching score between two feature vectors of the probe and gallery images on four unique, challenging datasets. Publicly available face datasets such as FG-Net, FRGC, FERET, and Georgia Tech are used for the experimentation, and the results prove that the proposed system is robust to the changes in age and outperforms most of the existing systems.