Illumination compensation for face recognition using adaptive singular value decomposition in the wavelet domain

Abstract

Lighting variations are a challenge in face recognition. To overcome this problem, this paper proposes a novel illumination compensation method called adaptive singular value decomposition in the 2D discrete wavelet domain (ASVDW) to enhance face images. First, an efficient brightness detector based on the blue pixel values of the red green blue (RGB) color channels is used to classify the color face image into dark, normal, or bright before applying the corresponding Gaussian template. The RGB color channels of the face image are then transformed to the 2D discrete wavelet domain. The frequency subband coefficients of the three color channels are automatically adjusted by multiplying the singular value matrices of these frequency subband coefficient matrices with their corresponding compensation weight coefficients. An efficient image denoising model is then applied, and a 2D inverse discrete wavelet transform is applied to obtain the ASVDW-compensated color face images without the lighting effect. In addition, a region-based ASVDW method (RASVDW), which entails the application of the ASVDW algorithm in four regions of an image, is introduced to reduce the computing time. Experimental results validate the efficiency of the proposed methods.