Illumination and Posture Invariant Face Recognition Using Radon and Nonsubsampled Contourlet Transform

In order to eliminate the effects of illumination and posture on face recognition, a novel algorithm based on illumination and posture invariant using Radon and nonsubsampled contourlet transform (NSCT) is proposed. This algorithm first performs logarithm and Radon transform on face image and constructs rotation, shift, scale topological invariant (posture invariants) . Then NSCT is used for analysis with multi scales and multi directions. Finally the adaptive normal shrink filter is applied to high frequency sub-bands for illumination and posture invariant extraction. Experimental results on ORL and Yale face databases show that the proposed algorithm can eliminate the effects of changes in illumination and posture on face recognition.