Weber binary pattern and Weber ternary pattern for illumination-robust face recognition

Local ternary pattern (LTP) is a noise-robust version of local binary pattern (LBP). They are both encoding for the differences between the intensity of the center pixel and its neighborhoods. In this paper, based on Webers law we propose two new local descriptors, named Weber binary pattern (WBP) and Weber ternary pattern (WTP), which utilize binary and ternary encoding separately for the evaluation of the relative local gray-scale difference. While sharing the merits of computational simplicity and noise tolerance embedded in LBP and LTP, WBP and WTP are more illumination-robust and can be regarded as adaptive versions of LBP and LTP. Experimental results in Extended Yale-B face database demonstrate the effectiveness of the proposed WBP and WTP in illumination-robust face recognition.