Face Recognition Based on Adaptive Soft Histogram Local Binary Patterns

摘要

In this paper we propose the adaptive soft histogram local binary pattern (ASLBP) for face recognition. ASLBP is an extension of the soft histogram local binary pattern (SLBP). Different from the local binary pattern (LBP) and its variants, ASLBP is based on adap-tively learning the soft margin of decision boundaries with the aim to improve recognition accuracy. Experiments on the CMU-PIE database show that ASLBP outperforms LBP and SLBP. Although ASLBP is designed to increase the performance of SLBP, the proposed learning process can be generalized to other LBP variants.