Multi View Photometric Stereo

Multi-view photometric stereo (MVPS) aims to reconstruct a 3D object's shape and surface properties from multiple images taken under varying lighting conditions. Recent research heavily utilizes neural networks, particularly those based on neural radiance fields (NeRFs) and implicit surface representations, to jointly estimate geometry, material properties (like BRDFs), and lighting, often incorporating techniques like per-pixel intensity rendering and shadow modeling for improved accuracy. This approach surpasses traditional methods by achieving higher accuracy and robustness, especially in challenging scenarios with sparse views or complex materials, leading to advancements in 3D modeling, computer vision, and related fields.