Performance enhancement of local vector pattern with generalized distance local binary pattern for face recognition

In this paper a generalized distance based local binary pattern (GDLBP) is proposed which achieves better recognition accuracy over LBP and significantly increases the recognition accuracy when combined with LVP. The proposed GDLBP is combined with local vector pattern (LVP) to achieve better recognition rates. Proposed GDLBP explores the spatial relationships of the neighboring pixels at different radii and inter radii distances. Proposed descriptors GDLBP and GDLBP-LVP have been analyzed and compared with the existing LBP and LVP over most challenging benchmark facial image databases LFW, and color FERET. Analysis of the descriptors shows that the proposed GDLBP and GDLBP-LVP achieves noticeably better results as compared to LBP and LVP.