While graphic rendering algorithms have made many advances in the last twenty years, many important physical effects are still being ignored. The current ray-tracing and radiosity methods for evaluating surface light emission make gross compromises that produce images that may be aesthetically pleasing in isolation but orders of magnitude incorrect when compared to reality.
PHOTON is a new approach that converges to a perfect solution to Kajiya's Rendering Equation. It sends photons from light sources, not rays from a viewpoint, and then computes physical interactions with a dynamic histogram that refines where a gradient appears statistically likely to exist. The forward particle-transport method of PHOTON make it easy to add fluorescence and polarization effects to the physics. Early experiments indicate that these effects make a large difference in perceived light and color intensities. From SIGGRAPH Proceedings, it appears that neither effect has ever been taken into account in rendering algorithms before.
PHOTON is only now possible because of the capacious memories of current workstations and servers. We use a Silicon Graphics Power Onyx with 1 GB of main memory, a cluster of 16 Indy workstations (64 MB each), and a 44 processor (32 MB each) Intel Paragon supercomputer for the generation of the histogram used to render a scene from any viewpoint. As can be seen from the above graphic, a recognizable image is produced quickly and steadily improves as the calculation continues. The rendering details of the example graphic show the progress made towards the final scene.