LMZMPM: Local Modified Zernike Moment Per-Unit Mass for Robust Human Face Recognition

In this work, we proposed a novel method, called Local Modified Zernike Moment per unit Mass (LMZMPM), for face recognition, which is invariant to illumination, scaling, noise, in-plane rotation, and translation, along with other orthogonal inherent properties of the Moments (ZMs). The LMZMPM computed each pixel in neighborhood size $3\times 3$ , then considers complex tuple that contains both phase magnitude coefficients as extracted features. As it components feature, has more information about image thus preserves edge structural information. We also propose hybrid similarity measure, combining Jaccard Similarity L1 distance, applied feature set classification. feasibility technique on varying illumination been evaluated CMU-PIE extended Yale B databases an average Rank-1 Recognition (R1R) accuracy 99.8% 98.66% respectively. To assess reliability method variations evaluate AR database obtain R1R higher than recent state-of-the-art methods. shows very high recognition rate Heterogeneous Face well, 100% CUFS, 98.80% CASIA-HFB.