Light fields and visibility papers concentrated mostly on shadows , light source acquisition, and the fast determination of visibility for real-time culling and rendering.
Ray Space Factorization for From-Region Visibility
The technique described in this paper concentrated on determining which parts of the scene are visible from a region (as opposed to determining the same from a point). The advantage of this technique being that the PVS (Potentially Visible Set) could be cached between rendering steps and thus, in an interactive environment, there are fewer calculations required per frame.
By rotating a vertical plane around the view cell, a test is made of each polygon at a particular incidence angle to determine occlusion of the polygon by polygons in front of it. These solutions are then used to mark polygons in the scene as either potentially visible or completely invisible. As the view cell moves, additional calculations can be made and added incrementally to the set (while others are taken away by re-testing far away points) and thus the system can be efficient for moving objects.
Structured Importance Sampling of Environment Maps
Switching from strict visibility to lighting, this technique samples an environment map based on the "importance" of the data in the map and creates representative point light sources that can reasonably simulate environment mapping.
The key to doing this effectively is the selection of portions of the image map that are most likely to contribute to the specular reflections in the scene. As such, the environment map is analyzed and broken down into regions based on size and brightness that reflect (no pun intended) their importance to the lighting in the scene, with areas of greater importance being represented by more lights.
In the end, the technique can be precomputed once for a given environment map and most environment maps can be reasonably represented by between 100 and 300 point lights.
Relighting with 4D Incident Light Fields
This paper deals with the revisualization of real objects with new lights. Modifications in lighting can include Position, Intensity, and Color. Although the technique is mostly academic right now (owing to the large number of controlled camera samples necessary to make the technique function), it is very effective at simulating a variety of lighting conditions after images have been made, thus reducing the amount of time necessary to create "interesting" lighting effects for real-world objects, since a polygonal model isn't necessary.
However, the technique currently requires a large (32 positions x 7 heights) number of projector moves at 36 samples per position, which is over 8,000 reference images.