A short YouTube presentation describing a cone-based ray tracing algorithm adapted to render photorealistic hair and furry objects with very large strand counts (100,000+). The video explains cone tracing vs. traditional ray tracing, the challenges of hitting thin hair strands with thick cones, proposed data structures and cone–fiber intersection techniques, adaptive cone sizing, and reported speedups of about 4–6× (many renders under five minutes, some under one minute). It also contains typical channel CTAs.
Chapters
Dear fellow scholars, this is two minute papers with this is a cone based ray tracing algorithm for rendering photorealistic images of hair and furry objects where the number of hair strands is typically over 100,000. Okay, wait, what do these terms mean exactly? Ray tracing means a bona fide light simulation program where we follow the path of many millions of light rays between the light sources and our camera. This usually means that light reflections and refractions, lens blur and defocus are taken into consideration. This feature is also referred to as depth of field or dof. In short, as you can see in the video, a fully ray traced system like this for hair and fur leads to absolutely beautiful images that you can see throughout this footage. So what about the cone based part? Earlier we had an episode about voxel cone tracing which is an absolutely amazing technique to perform ray tracing in real time. It works by replacing these infinitely thin light rays with thicker cone shaped rays, which reduces the execution time of the algorithm significantly at the cost of mostly a minor, sometimes even imperceptible degradation in image quality. Since the hair strands that we are trying to hit with the rays are extremely thin and cone tracing makes the rays thicker, extending this concept to rendering fur without non trivial extensions is going to be a fruitless endeavor. The paper contains techniques to overcome this issue and an efficient data structure is proposed to store and find the individual hair strands and a way to intersect these cones with the fibers. The algorithm is also able to adapt the cone sizes to the scene we have at hand. The previous techniques typically took at least 20 to 30 minutes to render one image, and with this efficient solution, we'll be greeted by a photorealistic image at least four to six times quicker in less than five minutes. While some examples were completed in less than a minute, I cannot get tired of seeing these tiny photorealistic furry animals in Pixar movies, and I am super happy to see there is likely going to be much, much more of these. By the way, if you're subscribed to the channel, please click the little bell next to the subscription icon to make sure you never miss an episode. Also, you can follow us on Twitter for updates. Thanks for watching and for your generous support. I'll see you next time. Sam.
