Single-image reflectance and transmittance estimation from any flatbed scanner

Abstract

We propose a method to estimate the reflectance and transmittance of a flat physical sample from a single scan acquired with any standard flatbed scanner. Our method is the first to estimate these properties from a single uncalibrated scan, without requiring any additional hardware or multiple captures. We achieve this by learning a latent space of material properties from a large dataset of synthetic scans, and then optimizing the material parameters to match the input scan. Our method is able to disentangle the diffuse and specular reflectance, as well as the transmittance and roughness of the sample. We demonstrate the effectiveness of our approach on a wide variety of materials, including fabrics, wood, and stone, and show that it outperforms state-of-the-art methods for material estimation from a single image