Quality Based Illumination Compensation for Face Recognition

Abstract

Design of a robust face recognition system is greatly affected due to varying lighting conditions, poses and expressions. The accuracy of the system can be increased by normalizing and compensating the illumination variations in the pre-processing step. This paper presents a wavelet based illumination compensation method after verifying the quality of the face images. Here the image quality is measured in terms of luminance with reference to the well-lit face images. We measure the quality of the image globally as well as locally before normalization and compensation. Experimental results on the Extended Yale-B face database show that the proposed method improves the face recognition performance for the face images with large illumination variations. It is simple and can be efficiently implemented in face recognition system when the face images are badly affected by varying illumination conditions.