A Three-Dimensional Inverse Rendering Method Based on Point Light Source Position Estimation

Inverse rendering, a technique for reconstructingthree-dimensional (3D) geometric structures and surface materials from two-dimensional (2D) images, is of great significance in fields such as robotic manipulation, virtual reality, and augmented reality. However, existing inverse rendering methods based on neural implicit representations are primarily suitable for scenarios with fixed light sources and struggle with dynamic point lighting. This paper introduces a novel 3D inverse rendering method that estimates the positions of point light sources, facilitating the 3D reconstruction of geometrical structures and the determination of light source locations from a collection of images captured from various perspectives, even under dynamic illumination conditions from an unknown point light source.