An efficient illumination normalization method in a transformed domain

Abstract

This paper proposes a novel illumination normalization approach with low computation complexity for face recognition. In this proposed method, a block-wise WalshHadamard transform (WHT) is employed in the logarithm domain. An appropriate number of low-frequency WHT coefficients are zeroed to compensate for illumination variations. Experiments on different databases demonstrate that the proposed method obtains results comparable to those of conventional Discrete Cosine Transform method but with a higher efficiency. It also achieves better performances for cases with larger illumination variations. Furthermore, both analytical proof and experimental results demonstrate that principal component analysis (PCA) can be directly implemented in the WHT domain.

Visual illumination compensation for face images using light mapping matrix

“…Perez and Castillo [16] propose a similar method which applies Genetic Algorithms to search appropriate weights to rescale the low-frequency DCT coefficients as well. Furthermore, Nie et al [17] and Zhichao and Joo [18] propose discarding coefficients, respectively, in discrete wavelet transform (DWT) and block-wise Walsh-Hadamard transform (WHT) instead of DCT to eliminate the negative effects of illumination variations. Preda and Vizireanu [19] use the wavelet coefficients in luminance component for watermark embedding.…”