Sampling for View Synthesis: From Local Light Field Fusion to Neural Radiance Fields and Beyond

Explore the evolution of view synthesis techniques in computer graphics and vision through this 50-minute conference talk by Ravi Ramamoorthi at #ICBS2024. Delve into the challenges of capturing and rendering novel views of complex real-world scenes, with applications in AR, VR, and 3D photography. Examine the impact of deep learning on image-based rendering and discover the Local Light Field Fusion algorithm, which enables practical view synthesis from irregularly sampled views. Learn about the extension of plenoptic sampling theory and its implications for efficient view sampling. Investigate recent developments in scene representations, including neural radiance fields, and consider the ongoing importance of sparse view synthesis. Reflect on the potential for prescriptive sampling guidelines in modern image-based rendering algorithms as you gain insights into this cutting-edge field.