5 ms·

The Color of Noise (2014)

80 points by ekzhang 2y ago

lovegrenoble 2y ago
[dead]
jmiskovic 2y ago
The same algorithm is described more clearly in Bridson's paper from 2007: https://www.cs.ubc.ca/~rbridson/docs/bridson-siggraph07-poissondisk.pdf https://www.cs.ubc.ca/~rbridson/docs/bridson-siggraph07-pois...
goodmachine 2y ago
"In fact, people have actually made spatial pattern generators that allow you to input the frequency profile that you want, and get the corresponding point pattern out. It’s really quite neat, and I highly recommend reading this paper so you can see some other possible noise parameters, like anisotropy." I had to hunt this 'custom colour' noise paper out since that link was dead. It is quite neat. Here it is on ACM in case anyone else is interested: Point Sampling with General Noise Spectrum (2012) https://dl.acm.org/doi/10.1145/2185520.2185572 https://dl.acm.org/doi/10.1145/2185520.2185572
- captaincrowbar 2y ago
  Non-paywalled version on one of the authors' site: https://www.liyiwei.org/papers/noise-sig12/ https://www.liyiwei.org/papers/noise-sig12/
teo_zero 2y ago
Something sounds odd. If points are very close, they have high spacial frequency. If I reject points that are too close, I'm excluding high-frequency points and keeping only low-frequency ones. A distribution that favor lower frequencies and filters out higher ones is red, not blue. Am I getting it wrong?
- wtallis 2y ago
  I think the color terms and frequency may make sense in reference to a comparison against uniformly-spaced points such as a rectangular grid. If you perturb uniformly-spaced points with high-frequency/short-wavelength adjustments, you won't get the large-scale clumping that results from also including low-frequency/long-wavelength distortions.
  - bearsnowstorm 2y ago
    I had this question and needed this answer - thank you
bschwindHN 2y ago
I was looking through the source code for Bevy's Screen-Space-Ambient-Occlusion (SSAO) implementation and noticed they are using blue noise as well: https://github.com/bevyengine/bevy/blob/56d559102858d4ce8a5b5e3bccd7053dd687a197/crates/bevy_pbr/src/ssao/mod.rs#L787 https://github.com/bevyengine/bevy/blob/56d559102858d4ce8a5b... That link takes you to this shadertoy: https://www.shadertoy.com/view/3tB3z3 https://www.shadertoy.com/view/3tB3z3
- jmiskovic 2y 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-with-volumetric-raymarching/ https://blog.maximeheckel.com/posts/real-time-cloudscapes-wi...
  - pixelpoet 2y ago
    And for an IMO much better explanation how it can be used more or less everywhere, this presentation by Playdead games on how it's used in INSIDE: https://www.youtube.com/watch?v=RdN06E6Xn9E https://www.youtube.com/watch?v=RdN06E6Xn9E 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 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_filters https://en.wikipedia.org/wiki/Mitchell%E2%80%93Netravali_fil...
danwills 2y ago
I would be very curious to hear if anyone knows how the color names are assigned?.. Is there any link to frequencies/spectra of light or is it kinda arbitrary?
- 082349872349872 2y ago
  the former: https://news.ycombinator.com/item?id=42291527 https://news.ycombinator.com/item?id=42291527 https://news.ycombinator.com/item?id=42292668 https://news.ycombinator.com/item?id=42292668
  - gsf_emergency 2y ago
    And here I was thinking it was because noise space is non-riemannian

80 points by ekzhang 2y ago

[dead]

The same algorithm is described more clearly in Bridson's paper from 2007: https://www.cs.ubc.ca/~rbridson/docs/bridson-siggraph07-poissondisk.pdf https://www.cs.ubc.ca/~rbridson/docs/bridson-siggraph07-pois...

goodmachine 2y ago

"In fact, people have actually made spatial pattern generators that allow you to input the frequency profile that you want, and get the corresponding point pattern out. It’s really quite neat, and I highly recommend reading this paper so you can see some other possible noise parameters, like anisotropy." I had to hunt this 'custom colour' noise paper out since that link was dead. It is quite neat. Here it is on ACM in case anyone else is interested: Point Sampling with General Noise Spectrum (2012) https://dl.acm.org/doi/10.1145/2185520.2185572 https://dl.acm.org/doi/10.1145/2185520.2185572

captaincrowbar 2y ago
Non-paywalled version on one of the authors' site: https://www.liyiwei.org/papers/noise-sig12/ https://www.liyiwei.org/papers/noise-sig12/

teo_zero 2y ago

Something sounds odd. If points are very close, they have high spacial frequency. If I reject points that are too close, I'm excluding high-frequency points and keeping only low-frequency ones. A distribution that favor lower frequencies and filters out higher ones is red, not blue. Am I getting it wrong?

wtallis 2y ago
I think the color terms and frequency may make sense in reference to a comparison against uniformly-spaced points such as a rectangular grid. If you perturb uniformly-spaced points with high-frequency/short-wavelength adjustments, you won't get the large-scale clumping that results from also including low-frequency/long-wavelength distortions.
- bearsnowstorm 2y ago
  I had this question and needed this answer - thank you

bschwindHN 2y ago

I was looking through the source code for Bevy's Screen-Space-Ambient-Occlusion (SSAO) implementation and noticed they are using blue noise as well: https://github.com/bevyengine/bevy/blob/56d559102858d4ce8a5b5e3bccd7053dd687a197/crates/bevy_pbr/src/ssao/mod.rs#L787 https://github.com/bevyengine/bevy/blob/56d559102858d4ce8a5b... That link takes you to this shadertoy: https://www.shadertoy.com/view/3tB3z3 https://www.shadertoy.com/view/3tB3z3

jmiskovic 2y 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-with-volumetric-raymarching/ https://blog.maximeheckel.com/posts/real-time-cloudscapes-wi...
- pixelpoet 2y ago
  And for an IMO much better explanation how it can be used more or less everywhere, this presentation by Playdead games on how it's used in INSIDE: https://www.youtube.com/watch?v=RdN06E6Xn9E https://www.youtube.com/watch?v=RdN06E6Xn9E 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 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_filters https://en.wikipedia.org/wiki/Mitchell%E2%80%93Netravali_fil...

danwills 2y ago

I would be very curious to hear if anyone knows how the color names are assigned?.. Is there any link to frequencies/spectra of light or is it kinda arbitrary?

082349872349872 2y ago
the former: https://news.ycombinator.com/item?id=42291527 https://news.ycombinator.com/item?id=42291527 https://news.ycombinator.com/item?id=42292668 https://news.ycombinator.com/item?id=42292668
- gsf_emergency 2y ago
  And here I was thinking it was because noise space is non-riemannian