anubis at January 18th, 2005 06:39 — #1
Description the image shows a 20min rendering of my path tracer (rest in peace). you can see a diffuse sphere, a glass sphere with fake fresnel (notice the caustics though) and a purely specular sphere. all the light in the room is diffusely reflected.
allthough i was quite satisfied with the path tracer you can still notice some banding effects in the shadows due to bad recycling of the stratified light samples.
i used stratified samples for the path tracing as well, which actually gave a nice effect during image generation because the scene was sort of "scanned" from left to right by the pixels. the anti aliasing is done via jittered sampling of the rays.
an earlier version of the path tracer had support for traceing CSG surfaces rather efficently but i have no really interesting pictures of that on my machine (allthough i still have an article on the topic lying around, which i haven't released so far but is completely finished).
i abandoned the path tracer though in favor of photon mapping based on the kd tree construction technique proposed in the "Faster Photon Mapping using the Voxel Volume Heuristic" paper on [www.openrt.de](www.openrt.de). you can expect to see nice pictures rendered by the photon mapper soon
dk2 at January 18th, 2005 11:47 — #2
awsome work! I bet there's room for optimizations to reduce that 20min rendering time.
I'd be interested in reading your paper.
anubis at January 18th, 2005 12:58 — #3
well... for one it's path tracing and that is only getting so fast. once i have the photon mapping finished i expect things to be much faster.
bladder at January 18th, 2005 22:55 — #4
sweet picture. I should really start getting into this "one hour rendering" business. I wonder how long before this stuff becomes real time.
anubis at January 19th, 2005 07:33 — #5
with photon mapping the map generation "only" takes a few seconds and the rendering can be done at least "interactive". i'm pretty sure that dual core processors will bring the change here