Towards High-level, Intuitive Descriptors of Material Appearance

Abstract

Material appearance perception depends both on the physical interaction between light and material, and on how our visual system processes the information reaching our eyes. Currently, there is a disconnect between the physical properties underlying material appearance models used in simulations, and perceptual properties that humans rely on when interpreting visual depictions of material appearance. Our goal is to bridge this gap, creating high-level, intuitive descriptors of visual appearance that are linked to the underlying physical properties of the material. This can in turn benefit final applications such as material appearance editing, acquisition, gamut mapping, or compression. Here, we review two approaches proposing representations of appearance that are better aligned with human perception: one based on the use of intuitive attributes, and another exploring the use of natural language descriptions of material appearance. All our data and models are publicly available.