Face Recognition under Varying Illumination Based on Local Binary Patterns with Circle Threshold

Abstract

To solve the challenging problem of face recognition under varying illumination conditions, we propose in this paper a novel LBP operator which we refer to as Local Binary Patterns with Circle Threshold (CT-LBP) operator. The CT-LBP operator can keep more discriminating information than the original LBP operator without losing the simplicity and effectivity of the original LBP operator. Extensive PCA-based experiments conducted on the Yale Face Database B demonstrate that the proposed method has been of general recognition rate of 98.13% on Yale Face Database B against extreme ighting conditions and outperforms the ones based on the original LBP operator.