State of the Art in Ray Tracing Animated Scenes

Abstract

Ray trac­ing has long been a method of choice for off-line ren­der­ing, but tra­di­tion­ally was too slow for in­ter­ac­tive use. With faster hard­ware and al­go­rith­mic im­prove­ments this has re­cently changed, and real-time ray trac­ing is fi­nally within reach. How­ever, real-time ca­pa­bil­ity also opens up new prob­lems that do not ex­ist in an off-line en­vi­ron­ment. In par­tic­u­lar real-time ray trac­ing of­fers the op­por­tu­nity to in­ter­ac­tively ray trace mov­ing/animated scene con­tent. This presents a chal­lenge to the data struc­tures that have been de­vel­oped for ray trac­ing over the past few decades. Spa­tial data struc­tures cru­cial for fast ray trac­ing must be re­built or up­dated as the scene changes, and this can be­come a bot­tle­neck for the speed of ray trac­ing.

This bot­tle­neck has re­cent­ly re­ceived much at­ten­tion by re­searchers that has re­sulted in a mul­ti­tude of dif­fer­ent al­go­rithms, data struc­tures and strate­gies for han­dling an­i­mated scenes. The ef­fec­tive­ness of tech­niques for ray trac­ing dy­namic scenes vary dra­mat­i­cally de­pend­ing on de­tails such as scene com­plex­ity, model struc­ture, type of mo­tion and the co­herency of the rays. Con­se­quently, there is so far no ap­proach that is best in all cases, and de­ter­min­ing the best tech­nique for a par­tic­u­lar prob­lem can be a chal­lenge. In this State of the Art Re­port (STAR), we aim to sur­vey the dif­fer­ent ap­proaches to ray trac­ing an­i­mated scenes, dis­cussing their strengths and weak­nesses, and their re­la­tion­ship to other ap­proaches. The over­all goal is to help the reader choose the best ap­proach de­pend­ing on the sit­u­a­tion, and to ex­pose promis­ing ar­eas where there is po­ten­tial for al­go­rith­mic im­prove­ments.