Face illumination processing using nonlinear dynamic range adjustment and gradientfaces
Zhijun Yang, Xiangfei Nie, He Xue, Wenyi Xiong · 2017
In this paper, a novel illumination processing approach based on nonlinear dynamic range adjustment and gradientfaces is discussed.Firstly, the grayscale of face image is adjusted by nonlinear dynamic range adjustment using hyperbolic sine function in logarithm domain.After finishing the adjustment, gradientfaces is used to enhance the high frequency component of face image and extract distinguishing facial feature.Secondly, the data dimensionality is reduced by principal component analysis (PCA), the nearest neighbourhood-based classifier based on cosine distance is adopted for face classification.The experiment results on Yale-B frontal face database demonstrates that the average face recognition rate of our technique can reach to 99.74%.Therefore, it is insensitive to different lighting resources.