We present DepthLight, a novel method to represent estimated spatial lighting for photorealistic Visual Effects (VFX). Previous approaches either relied on environment maps that fail to take into account localized lighting effects or used simplified 3D representations of lights that do not fully capture the complexity of the lighting process.

DepthLight addresses these limitations by using a single LDR image with a limited field of view (LFOV) to build an emissive texture mesh, producing a simple and lightweight 3D representation for photorealistic object relighting. Our approach includes a two-step HDR environment map estimation process. First, an LDR panorama is generated using a photorealistic generative model. Then, an LDR to HDR network reconstructs the HDR panorama. The photorealistic LDR panorama enables depth estimation, allowing for a mesh representation which enables spatial lighting information. This flexible pipeline can be easily integrated into different VFX production workflows.

In our experiments, DepthLight shows that virtual objects are seamlessly integrated into real scenes with an accurate simuation of the lighting. We compared our results to the ground truth lighting using Unreal Engine 5, as well as to regular approaches that uses pure HDRI lighting techniques. Finally, we have additionally validated our approach using image quality assessment metrics.