An efficient illumination normalization method for face recognition

摘要: In this paper, an efficient representation method insensitive to varying illumination is proposed for human face recognition. Theoretical analysis based on the model and shows that effects of lighting a image can be modeled by sequence multiplicative additive noises. Instead computing these noises, which very difficult real applications, we aim reduce or even remove their effect. our method, local normalization technique applied image, effectively efficiently eliminate effect uneven illuminations while keeping statistical properties processed same as in corresponding under normal condition. After processing, will have similar pixel values Then, images are used The algorithm has been evaluated Yale database, AR PIE YaleB database combined using different recognition methods such PCA, ICA Gabor wavelets. Consistent promising results were obtained, show greatly improve results.