Detail-Enhance Face Illumination Normalization Based on LDCT-Wavelet

摘要

As a kind of self-carry and unique biological characteristic, face has been widely used in authentication field. Although in recent years face recognition has achieved rapid progress, there are still some technical fortresses difficult to overcome. Illumination is one of the major challenges, a large number of researches have showed that the accuracy of face recognition is highly dependent on illumination variations. In this paper a novel DCT-Wavelet approach is proposed, through rescaling the low-frequency coefficients based on entropy optimized by rand-shake and wavelet de-noising by maximizing detail energy with strategy adaptive differential evolution (SaDE), image can normalize illumination and enhance detail at the same time. Experiments showed this approach has outstanding advantage over other LDCT methods and the error rate of face recognition can be decreased approximate 3%.