3D Scene Surface Reconstruction from Radiance Fields

Reconstructing complex 3D scenes is a fundamental challenge in computer vision, computer graphics, and robotics, with applications in virtual reality, robotic navigation, and simulation. While radiance field-based representations enable high-fidelity novel view synthesis, extracting accurate surface geometry remains a major challenge. This thesis studies the problem of 3D scene surface reconstruction. By improving geometry quality and accelerating reconstruction, our work advances 3D scene modeling, making high-quality surface extraction from radiance fields more practical and scalable for real-world applications.