This paper introduces a novel method for generating new views from arbitrary camera positions without depth information or feature matching. The key idea is to interpret input images as 2D slices of a 4D light field, which characterizes the flow of light through unobstructed space in a static scene with fixed illumination. The authors describe a sampled representation for light fields that allows for efficient creation and display of inward and outward-looking views. They have created light fields from large arrays of rendered and digitized images, using a video camera mounted on a computer-controlled gantry. The paper also addresses the challenges of antialiasing during creation and resampling during slice extraction, and presents a compression system that can compress light fields by more than 100:1 with minimal loss of fidelity. The authors discuss the advantages and limitations of their method, including the need for high sampling density, the restriction to regions free of occluders, and the requirement for fixed illumination. They compare their approach with depth-based and correspondence-based view interpolation methods, highlighting the benefits of their light field representation in terms of simplicity and compression efficiency.This paper introduces a novel method for generating new views from arbitrary camera positions without depth information or feature matching. The key idea is to interpret input images as 2D slices of a 4D light field, which characterizes the flow of light through unobstructed space in a static scene with fixed illumination. The authors describe a sampled representation for light fields that allows for efficient creation and display of inward and outward-looking views. They have created light fields from large arrays of rendered and digitized images, using a video camera mounted on a computer-controlled gantry. The paper also addresses the challenges of antialiasing during creation and resampling during slice extraction, and presents a compression system that can compress light fields by more than 100:1 with minimal loss of fidelity. The authors discuss the advantages and limitations of their method, including the need for high sampling density, the restriction to regions free of occluders, and the requirement for fixed illumination. They compare their approach with depth-based and correspondence-based view interpolation methods, highlighting the benefits of their light field representation in terms of simplicity and compression efficiency.