Inferring geometry and material properties from Mueller matrices with machine learning

TL;DR

This paper investigates whether Mueller matrices contain enough information to infer surface geometry and material properties simultaneously using machine learning, demonstrating successful predictions and reconstructions from optical data.

Contribution

The study shows that machine learning models can accurately predict surface normals and material types solely from Mueller matrices, even when the material is unknown.

Findings

Surface normals can be predicted from Mueller matrices.

Material types can be identified using Mueller matrices.

Diagonal elements are crucial for material characterization.

Abstract

Mueller matrices (MMs) encode information on geometry and material properties, but recovering both simultaneously is an ill-posed problem. We explore whether MMs contain sufficient information to infer surface geometry and material properties with machine learning. We use a dataset of spheres of various isotropic materials, with MMs captured over the full angular domain at five visible wavelengths (450-650 nm). We train machine learning models to predict material properties and surface normals using only these MMs as input. We demonstrate that, even when the material type is unknown, surface normals can be predicted and object geometry reconstructed. Moreover, MMs allow models to identify material types correctly. Further analyses show that diagonal elements are key for material characterization, and off-diagonal elements are decisive for normal estimation.