Cycles: support Mie Scattering with particle size smaller than 5um #130234

Weizhen Huang · 2024-11-13T18:01:35+01:00

Weizhen Huang commented

2024-11-13 18:01:35 +01:00

Previous implemenation of 5 < d < 50 was taken from the main paper,
fitting for smaller sizes are found in the supplemental. They are less
forward-scattering.

Previous implemenation of 5 < d < 50 was taken from the main paper, fitting for smaller sizes are found in the supplemental. They are less forward-scattering.

Weizhen Huang added the

Module

Render & Cycles

label 2024-11-13 18:01:35 +01:00

Weizhen Huang added 1 commit 2024-11-13 18:01:44 +01:00

Cycles: support Mie Scattering particle size 0 < d < 5 e3c6062132

Previous implemenation of 5 < d < 50 was taken from the main paper,
fitting for smaller sizes are found in the supplemental. They are less
forward-scattering.

Weizhen Huang changed title from ~~Cycles: support Mie Scattering particle size 0 < d < 5~~ to Cycles: support Mie Scattering with particle size 0 < d < 5um

2024-11-13 18:02:04 +01:00

Weizhen Huang changed title from ~~Cycles: support Mie Scattering with particle size 0 < d < 5um~~ to Cycles: support Mie Scattering with particle size smaller than 5um

2024-11-13 18:02:27 +01:00

Weizhen Huang commented

2024-11-13 18:13:20 +01:00

Plot for the phase function (in log scale):

Plot for the fitted parameters:

Note tha `\alpha` has a different scale than the other parameters.

I am quite surprised by the discontinuities, especially at d = 0.1 and 1.5, and how they managed to fit a function so complicated, but the rendering seems fine.

## Plot for the phase function (in log scale): ![mie_scatter_phase.svg](/attachments/5e6b8bbc-5050-4ac1-a65a-bd0bf94d32bb) ## Plot for the fitted parameters: ### Note tha \(\alpha\) has a different scale than the other parameters. ![mie_scatter_parameters.svg](/attachments/cea69d15-5b53-4ff3-8111-e5d54668e527) I am quite surprised by the discontinuities, especially at d = 0.1 and 1.5, and how they managed to fit a function so complicated, but the rendering seems fine.

mie_scatter_parameters.svg

40 KiB

mie_scatter_phase.svg

54 KiB

Weizhen Huang requested review from Sergey Sharybin 2024-11-13 18:13:41 +01:00

Weizhen Huang requested review from Lukas Stockner 2024-11-13 18:13:46 +01:00

Lukas Stockner requested changes 2024-11-13 18:28:28 +01:00

Lukas Stockner left a comment

If I remember correctly, this also needs to be done in the OSL version.
Otherwise LGTM, I skipped these initially but I guess we might as well add them.

If I remember correctly, this also needs to be done in the OSL version. Otherwise LGTM, I skipped these initially but I guess we might as well add them.

Weizhen Huang commented

2024-11-13 18:29:11 +01:00

@LukasStockner Ah, I keep forgetting we have this OSL thing

Sergey Sharybin commented

2024-11-13 18:41:22 +01:00

It should be possible to use volume_util.h from node_scatter_volume.osl and avoid duplication?

It should be possible to use `volume_util.h` from `node_scatter_volume.osl` and avoid duplication?

Alaska referenced this pull request from blender/blender-test-data

2024-11-25 13:01:55 +01:00

Cycles: Add tests for new volumetric scatter phase functions #25

Weizhen Huang force-pushed mie-0-5 from e3c6062132 to ad3816a13a

2024-12-10 12:10:31 +01:00

Compare

Weizhen Huang added 2 commits 2024-12-10 12:11:22 +01:00

Add Mie Scattering for OSL 72bfbed733

Precompute log(d) for some SVM paths fb2219136d

Weizhen Huang requested review from Lukas Stockner 2024-12-10 12:12:52 +01:00

Sergey Sharybin commented

2024-12-10 16:38:57 +01:00

I remember we've discussed possibility of avoiding duplication by adding another header where we can have shareable code between OSL and SVM (as the current volume_util.h can not be included from OSL) . Were there some complications with that?

I remember we've discussed possibility of avoiding duplication by adding another header where we can have shareable code between OSL and SVM (as the current `volume_util.h` can not be included from OSL) . Were there some complications with that?

Sergey Sharybin approved these changes 2024-12-10 16:40:06 +01:00

Sergey Sharybin left a comment

Had some quick catch-up IRL with Weizhen.

The pointers and fast_cos are annoying for sharing. So its more practical to move forward with this PR and look into deduplication later.

Had some quick catch-up IRL with Weizhen. The pointers and fast_cos are annoying for sharing. So its more practical to move forward with this PR and look into deduplication later.

Weizhen Huang merged commit c99b7e66b2 into main

2024-12-13 15:51:14 +01:00

Weizhen Huang referenced this issue from a commit

2024-12-13 15:51:15 +01:00

Cycles: support Mie Scattering with particle size smaller than 5um

Weizhen Huang deleted branch mie-0-5

2024-12-13 15:51:19 +01:00

Alaska commented

2024-12-13 15:54:52 +01:00

@weizhen Before you go on holiday.

You said you'd like extra render tests for the new particle sizes after this was commited.

Were there specific particle sizes you wanted tested? Or should I create a test for each section shown in your graph here: #130234 (comment)

@weizhen Before you go on holiday. You said you'd like extra render tests for the new particle sizes after this was commited. Were there specific particle sizes you wanted tested? Or should I create a test for each section shown in your graph here: https://projects.blender.org/blender/blender/pulls/130234#issuecomment-1339084

Weizhen Huang commented

2024-12-13 15:59:15 +01:00

@Alaska we could do one for each range, d <= 0.1f, d < 1.5f and d < 5.0f, otherwise I don't think the exact value matters.

I picked a few for the drawing because I was wondering how they look like. I could create a repo and push my python script there, if it's of interest.

EDIT: I put the script here https://projects.blender.org/weizhen/plot/src/branch/main/volume_phase_function.py

@Alaska we could do one for each range, `d <= 0.1f`, `d < 1.5f` and `d < 5.0f`, otherwise I don't think the exact value matters. I picked a few for the drawing because I was wondering how they look like. I could create a repo and push my python script there, if it's of interest. EDIT: I put the script here https://projects.blender.org/weizhen/plot/src/branch/main/volume_phase_function.py

👍 1

Sign in to join this conversation.

No reviewers

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Cycles: support Mie Scattering with particle size smaller than 5um #130234

Plot for the phase function (in log scale):

Plot for the fitted parameters:

Note tha \alpha has a different scale than the other parameters.

Note tha `\alpha` has a different scale than the other parameters.