Robust face recognition for illumination removal using DT-CWT and EMD

In many various applications facial images are dramatically changed especially by lighting variations, so that facial appearance changes caused serious performance degradation in face recognition. In this paper we describe a method to address illumination removal for face recognition using Empirical Mode Decomposition (EMD) to decompose subimages of Dual-Tree Complex Wavelet Transform (DT-CWT) into their intrinsic mode function that correspond to the dominant illumination factors. The DT-CWT subimages we reconstruct provide good directional selectivity in six different fixed orientations at different scales without these illumination distortion components. We then perform verification experiments using algorithms such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and improved Orthogonal Neighborhood Preserving Projections (IONPP) to demonstrate the effectiveness of EMD as an illumination compensation method. Results are reported on the CMU PIE database.