Visual Hull

Visual hull is a 3D reconstruction technique that estimates the volume occupied by an object based on its silhouettes from multiple viewpoints. Current research focuses on improving the accuracy and efficiency of visual hull methods, particularly by integrating them with deep learning architectures like Neural Radiance Fields (NeRFs) and employing techniques such as voxel carving and Gaussian splatting to refine 3D models. These advancements are enhancing applications in areas such as high-resolution volumetric capture of humans and objects, improving the speed and quality of 3D reconstruction from sparse or event-based data. The resulting improvements in accuracy and efficiency are significant for various fields, including computer vision, 3D modeling, and virtual/augmented reality.