Circular artifacts

Use this forum for general user support and related questions.
Post Reply
yiusay
Posts: 4
Joined: Mon Jul 13, 2020 9:11 am

Circular artifacts

Post by yiusay » Sat Aug 01, 2020 11:44 am

I am testing a simple scene to generate caustics under ripples on water surface. Then, I am observing some weird circular pattern in the rendered image. Can you let me know whether this is expected result? If not, can you let me know how to remove these circular artifacts?

The scene consists of a plane (corresponding to water surface) and a floor, and a big plane behind the camera. If the big plane is removed from the scene, the circular artifacts are gone.

The circular artifacts appear if the big plane is there.
https://drive.google.com/file/d/1d0QdEI ... sp=sharing

No artifacts appear if the big plane is removed.
https://drive.google.com/file/d/1moVef ... p=sharing

This is the corresponding blender file. (The big plane is names as "background-image")
https://drive.google.com/file/d/1tVzHh ... p=sharing

Thank you
Yusuke

User avatar
B.Y.O.B.
Developer
Posts: 3676
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Circular artifacts

Post by B.Y.O.B. » Sat Aug 01, 2020 12:10 pm

I guess you are using the caustics cache and the caustic lookup radius is simply too big?

yiusay
Posts: 4
Joined: Mon Jul 13, 2020 9:11 am

Re: Circular artifacts

Post by yiusay » Sat Aug 01, 2020 4:39 pm

Thanks for quick reply!

Yes, I am using caustic cache. The lookup radius was 75mm, which seemed to be too big.

I reduced it to 5mm and got this result. Do you think this result looks reasonable? (The number of samples is 120)
https://drive.google.com/file/d/1vsadjP ... sp=sharing

I actually do not understand the algorithm of how caustic cache and lookup radius work. If any, can you point to some document or academic paper discussing those things?

User avatar
B.Y.O.B.
Developer
Posts: 3676
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Circular artifacts

Post by B.Y.O.B. » Sat Aug 01, 2020 7:06 pm

yiusay wrote:
Sat Aug 01, 2020 4:39 pm
I reduced it to 5mm and got this result. Do you think this result looks reasonable? (The number of samples is 120)
It looks like you need more photons. If you use periodic update, simply render for longer time, otherwise try to increase the "Max. Size" parameter.
yiusay wrote:
Sat Aug 01, 2020 4:39 pm
I actually do not understand the algorithm of how caustic cache and lookup radius work. If any, can you point to some document or academic paper discussing those things?
Any explanation on photon mapping should help you out, there are probably plenty of them, the technique is quite old.
This might also help: https://wiki.luxcorerender.org/PhotonGI ... e_Settings

The lookup radius is basically the size of the "splotch" a photon creates. It defines the blurriness of the result - the higher the radius, the more blurry, the lower the radius, the more noisy the result (and the more photons you will need to clear it up).
So it's a good idea to find a balance between blur and noise, unfortunately this has to be done through trial-and-error and might be a bit time consuming.

Post Reply