FLBP: Fechner local binary pattern for face representation

Most images are ultimately observed and interpreted by humans, so the ideal image descriptor should take into account the effects of human psychology and psychophysics. In this paper, we develop a novel feature descriptor named Fechner local binary pattern (FLBP) based on the well-known psychological law, Fechner’s law. FLBP describes images using mental perception, which is a logarithmic function of the stimulus change, allowing for a more detailed and hierarchical representation of the represented image. In addition, considering the structural features of the face, we adjusted the size of the blocks so that it can be large enough to include the complete face organ(s), since these face organs, like the eyes, nose, and mouth contain the most discriminative features. The addition of changeable size blocks effectively reduces the effects of noise and illumination. Experiments on four face image databases demonstrate the effectiveness of the proposed FLBP method.