Author Topic: Tonemapping - Plz Halp  (Read 115859 times)

2020-05-08, 17:50:05
Reply #300

Jpjapers

  • Active Users
  • **
  • Posts: 1644
    • View Profile

Gonna test it by recreating a similar scene. But I observe this spectral effect on white walls caused by any edge that deflects the light, no matter what color the material has... can be a white window frame, etc.
So I assume it is in fact the spectral effect you most strongly see with prisms.

The shadows appear coloured because the bright yellow light from the sun is blocked from the shadow area, leaving only indirect light and blue light and vice versa.

2020-05-08, 17:54:17
Reply #301

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile

Gonna test it by recreating a similar scene. But I observe this spectral effect on white walls caused by any edge that deflects the light, no matter what color the material has... can be a white window frame, etc.
So I assume it is in fact the spectral effect you most strongly see with prisms.

The shadows appear coloured because the bright yellow light from the sun is blocked from the shadow area, leaving only indirect light and blue light and vice versa.

Well, maybe that's the reason, However, this effect is very strong in reality and I never see it in my renders. Only due to compression artifacts in JPGs. :)))))))

But: you clearly see that the shadows are actually multiple shadows overlapping, with the spectral colors at the edges.
I remember we learned about this effect in physics in the 9th grade. :)

2020-05-08, 18:00:46
Reply #302

Jpjapers

  • Active Users
  • **
  • Posts: 1644
    • View Profile
But: you clearly see that the shadows are actually multiple shadows overlapping, with the spectral colors at the edges.
I remember we learned about this effect in physics in the 9th grade. :)

Yeah because the colours are only in the areas where the indirect and direct lighting bounces dont cancel each other out.

2020-05-08, 19:53:49
Reply #303

PROH

  • Active Users
  • **
  • Posts: 1219
    • View Profile
@ Jpjapers - Thanks :)

2020-05-08, 21:34:37
Reply #304

cjwidd

  • Active Users
  • **
  • Posts: 1077
    • View Profile
    • Artstation
Hard to draw a definitive conclusion taking into account that you are not a proficient Fstorm user and this may depend on the shader setup [...]

I'm sort of on the fence about whether this is important for this comparison; if the intention were to induce Corona to emulate Fstorm in its entirety, than this experiment is not useful. The point here was to compare a simple stimulus in both renderers as close as possible to default. Of course if we leave Corona at literal default, the comparison is not meaningful at all because Corona does not tonemap by default, i.e. straw man comparison.

Regardless, I'm sure there is more that could be done to the shader in Corona to get better visual parity with Fstorm.

That being said, there is something unusual happening in the grey box scenes (third image in each set) because the grey floor / walls are MUCH brighter in the Corona scene by comparison - same 186,186,186 grey material that appears in the first render of each group.
« Last Edit: 2020-05-08, 21:53:19 by cjwidd »

2020-05-08, 21:52:24
Reply #305

cjwidd

  • Active Users
  • **
  • Posts: 1077
    • View Profile
    • Artstation
I can barely see any difference.
Fstorm and corona next to each other. Image divided in half. I see a little more blur in the fstorm rendering.

I guess I'm being trolled aren't you the one that allegedly removed Fstorm watermarks from some earlier tests?

Just to be clear, comparing the second shot for both Fstorm and Corona Renderer does not produce an identical result on my end.

The grey cylinder was used to calibrate the Exposure (to some extent). I found that -.75 EV brought the Corona Renderer version in-line with the Fstorm render. The parquet material was applied to the cylinder and rendered unadjusted from the grey cylinder scene.

« Last Edit: 2020-05-08, 23:46:02 by cjwidd »

2020-05-08, 22:10:58
Reply #306

James Vella

  • Active Users
  • **
  • Posts: 542
    • View Profile
I guess I'm being trolled - aren't you the one that allegedly removed Fstorm watermarks from some earlier tests?

lol!! I think you are being trolled :D Made me giggle though

2020-05-08, 22:13:01
Reply #307

cjwidd

  • Active Users
  • **
  • Posts: 1077
    • View Profile
    • Artstation
Wow, fake news even on the Corona forums - great...

@lolec If you are editing watermarks out in PS and reposting here, then I would invite you to stop posting to the thread. I'm not sure why that type of behavior would be considered warranted or appropriate for the present discussion. If you can't be honest or transparent in your dealings on the forum, then please consider sharing your thoughts elsewhere.

2020-05-08, 22:17:56
Reply #308

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
Wow, fake news even on the Corona forums - great...

@lolec If you are editing watermarks out in PS and reposting here, then I would invite you to stop posting to the thread. I'm not sure why that type of behavior would be considered warranted or appropriate for the present discussion. If you can't be honest or transparent in your dealings on the forum, then please consider sharing your thoughts elsewhere.

I came here to learn about how we can get more realistic renders. Dudes, don't quarrel please.

2020-05-08, 23:13:16
Reply #309

lolec

  • Active Users
  • **
  • Posts: 179
    • View Profile
Wow, fake news even on the Corona forums - great...

@lolec If you are editing watermarks out in PS and reposting here, then I would invite you to stop posting to the thread. I'm not sure why that type of behavior would be considered warranted or appropriate for the present discussion. If you can't be honest or transparent in your dealings on the forum, then please consider sharing your thoughts elsewhere.

1. Unless I broke a forum rule inherited from 19-century prudes, I will respectfully decline your invitation and continue to post.

2. There are many theories floating around, mine was that people had a slight bias to quickly call Fstorm renders better. I TESTED that theory and at least two people (Who, btw took it waaaaay better than you) partially confirmed that sometimes that can happen. NO this does not mean there is no difference between Fstorm and Corona, but maybe we are now open to the possibility of human bias, maybe just a tiny percent. I think that is a way of moving the conversation forward and is moving us all towards a more objective analysis.

3. Please stop using the words Honest and Transparent referring to this. It was a silly experiment done tongue in cheek, absolutely NO ONE was hurt, other than your butt, apparently.

4. I DO see a difference in the images you posted... they are clearly different. I'm not sure why LUPAZ posted an image that looks modified to reduce the difference, maybe he is in fact trolling you, and maybe that made you angry. But you see... LUPAZ is not the same as LOLEC. For someone who claims to see minute differences, you missed a big one there. Here is an image that might help.


I don't need to tell you which one I think is better :)

Again, I don't think faking the watermark hurt anyone and we are all adults who can laugh about that kind of stuff, we want the same thing, don't fight. I hope to keep this conversation going as I'm learning from this trhead and I hope others are too.

2020-05-08, 23:46:24
Reply #310

cjwidd

  • Active Users
  • **
  • Posts: 1077
    • View Profile
    • Artstation
@lolec thanks for clarifying, the post has been edited for accuracy

2020-05-09, 02:46:46
Reply #311

lolec

  • Active Users
  • **
  • Posts: 179
    • View Profile
What I actually wanted to post since months: look at natural light behavior in the photo taken in my living room. Just like a glass prism, even trivial objects do break light in its spectral colors.

I was wondering many times how rendered images would look if Corona calculated this effect...

That is a great example of how you can provide more realism with any engine! And how attention to detail plays a big role some times.

Even though it might look like a prism or dispersion, it is actually a much simpler phenomenon. This video might be interesting to you:

So, what is happening is that the light SOURCE has all sort of colors, intensities and directions. I would need to see the outside of that window to understand what's happening, but because of the orange colors in the light, my guess is blue sky and there is a brick wall nearby? Maybe.

Anyway, HDRs do an OK job at emulating exterior lighting, but to get the full effect you need to add nearby object that can interact with the light in a natural way.

Here is an image I made using many suns with different colors and intensities. Of course this is not natural! but I was trying to match your picture and show how that effect is the result of light source and not light dispersion.




2020-05-09, 16:27:34
Reply #312

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile

Just to be clear, comparing the second shot for both Fstorm and Corona Renderer does not produce an identical result on my end.




I just adjusted the exposure in photoshop.

2020-05-09, 18:03:20
Reply #313

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile
One more test I wanted to make.
This test isn't directly related to tone mapping, but I think that tone mapping is affected by it.


For this test I'm using path tracing for secondary bounces. The images have no saturation. I did try to match contrast and exposure slightly (without going crazy about it). No bloom, no DOF.
I notice a couple of things that I think are helping Fstorm almost unwantedly:

1- The energy of bounced light is much less in Fstorm. It makes it look more 3D-like, but it separates colors from surface to surface and adds deeper clarity to corners and small cracks or spaces.

I really wish there was a setting in corona to lower this energy that worked better than Max Sample Intensity. A while ago I started this thread about this issue: https://forum.corona-renderer.com/index.php?topic=23129.msg164407#msg164407

I also think this is the reason why nobody in this thread is talking about exterior renderings: because secondary bounces don't play such an important role in exteriors. If the issue with corona was exclusively tone mapping, why nobody is talking about exterior renderings?
Note: I don't think this is necessarily an issue. Corona looks great, but it seems that the trend in photos and renderings is this one: one where every surface is very separate from each other clearly.





2- Corona applies blur to all textures by default. Fstorm doesn't even have blur in their material AFAIK.
Why? Why blur by default?
The only reason I don't remove it is because Corona's bump mapping doesn't work without blur.




3- And one more thing. This is a question:
Does anyone know what this does exactly?



2020-05-09, 18:07:01
Reply #314

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
What I actually wanted to post since months: look at natural light behavior in the photo taken in my living room. Just like a glass prism, even trivial objects do break light in its spectral colors.

I was wondering many times how rendered images would look if Corona calculated this effect...

That is a great example of how you can provide more realism with any engine! And how attention to detail plays a big role some times.

Even though it might look like a prism or dispersion, it is actually a much simpler phenomenon. This video might be interesting to you:

So, what is happening is that the light SOURCE has all sort of colors, intensities and directions. I would need to see the outside of that window to understand what's happening, but because of the orange colors in the light, my guess is blue sky and there is a brick wall nearby? Maybe.

Anyway, HDRs do an OK job at emulating exterior lighting, but to get the full effect you need to add nearby object that can interact with the light in a natural way.

Here is an image I made using many suns with different colors and intensities. Of course this is not natural! but I was trying to match your picture and show how that effect is the result of light source and not light dispersion.





Quite cool how you recreated the effect. Does it not bring a nice level of realism, compared to boring grey shadows?

Your guess is correct, outside is a building with a wooden outer shell that reflects a lot of warm / orange light when it is indirectly lit by the sun.
However... from HDRs, which quite often have large areas with different colors, I do not consciously remember to have seen that effect.
I´ll keep an eye on it. :)