This paper develops a mathematical model of visual perception related to the encoding of spatial information. Understanding how the brain represents spatial properties of objects in physical space remains a central and challenging problem in psychophysics. We propose that neural mechanisms, specifically the Fisher information encoded in neural population activity, serve a role analogous to the energy-momentum tensor in physics, thereby generating a space-dependent metric tensor. This leads to a curved visual space characterized by a curvature tensor. Using this framework of non-Euclidean geometry, we analyse well-known phenomena in visual optics through concepts such as the Fisher-Rao metric, and psychometric distance.