|
|
|
|
|
|
by jmiskovic
567 days ago
|
|
|
It is a common pattern in rendering. You achieve your desired results, but it is nowhere near realtime. You try to lower the sampling resolution which helps a lot with speed, but you end up with ugly visual artifacts. Introducing the blue noise in the right place helps with attenuating these sampling artifacts. Only usually you would sample the premade "blue noise" texture rather than using the shader to compute it. This article does a good job of explaining how it helps in raymarching: https://blog.maximeheckel.com/posts/real-time-cloudscapes-wi... |
|
|
IMO the linked article by Casey Muratori is actually kind of weak: a lot of waffling, doesn't actually show the results of using this point set in the original problem, just using the sampling pattern isn't enough to eliminate aliasing, and there are more important properties for point sequences to have (discrepancy), etc. It's somewhat unpopular these days to link dense texts instead of "softer" articles, but IMO one should just go straight to PBRT for this topic, there's an entire chapter with excellent presentation: https://pbr-book.org/4ed/Sampling_and_Reconstruction
Final note: the Don Mitchell mentioned in this article is one of the authors on the go-to reconstruction filter used in a lot of software, Mitchell-Netravali (the latter recently passed away): https://en.wikipedia.org/wiki/Mitchell%E2%80%93Netravali_fil...