Image normalization method for face identification under difficult lighting conditions

Abstract

Difficulties related to poor illumination conditions are one of the main reasons for which many face identification algorithms fail in real-life situations. The paper presents a new method for image normalization, based on simple techniques such as binarization and histogram equalization, that effectively removes the shadows and provides illumination invariants that significantly improve the accuracy of face identification process. During experiments the proposed method outperformed the state of the art approach based on anisotropic smoothing.