Abstract:
Reflectance Transformation Imaging (RTI) is a widely used method for obtaining detailed per-pixel reflectance data by photographing an object under varying lighting conditions. The gathered information can then be utilized to re-light the subject interactively and reveal surface details that would be impossible to see with a single photo. In this thesis, we propose a data-driven approach based on an Implicit Neural Representation (INR) of the light transport function to interactively relight the scene in a photorealistic way. Comparisons with existing state-of-the-art methods demonstrate the feasibility of the approach and suggest further investigation of INRs for RTI applications.
