Interior Lighting, Noise and DiffRadius Researching

[condar]

i found this great "Quick tip" from another page...

...Basically, you have four result-levels for photons:

* Type 1. Dots or "blood cell" circles: Means you are looking up too few of them
* Type 2. Well covered but "cloudy"/"splotchy" (not smooth): Looking up a tad too few
* Type 3. Smooth and detailed: "just right"
* Type 4. Over-smooth and totally smeared : Looking up too many / too large radius

Basically, the photon search stops when either the radius is reached or the count is reached. So if you set it to 100 photons, 10 cm, it will look within for the 100 nearest photons within 10 cm.

* If the 100 nearest are within 1 centimeter, those 100 will be used, none further than 1 cm will be used
* If the 100 nearest are within 10 cm, they will all be used.
* If the 100 nearest are within 1m, only those within 10cm will be used.

From this you can trivially understand that if you hit the "number limit" this acts as an adaptively variable radius... the more photons, the better the detail, the less the photons, the more radius is looked up in.

However, if the count is too high, the radius is always hit first. So you will always "smear" the result over the radius. This is oversmooth (type 4 above).

The trick is to get to type 3 "just right", but the good thing is that when combined with FG, the accuracy need is much less. The cloudy (type 2) solution works just fine with FG. A spotty/circly (type 1) solution wont work, a over-smoothed (type 4) solution "works" but can put light where it's not supposed to be.

So with FG the "margin of error" is large.

My workflow is generally:

1. Turn off FG
2. Look at the photons-only result
3. Is it dotty? Increase radius, or photon count, until it stops being "dotty"
4. Is it super-smeared, decrease radius.
5. When neither "dotty" nor "super smeared" (generally "cloudy" or "nice"), be happy
6. Turn FG back on.
7. Rejoice.

http://www.mymentalray.com/wiki/index.p ... on_mapping

[Yaroslav_L]

Now I'll try to get nice picture without light on the celling...
And one more time about colours: I don't think about the colours - just pick it!
For example white celling is little blue near the window. So render/lighting correct it automatically...
Of course, I understand that there are no "clear" colours. But when I prepare the drawings and specifications to the client/contractor I just choose as white (as red) as it possible from Oikos, Ticcurila or another palette...
So the colour is not a problem! But rendering settings for interior without much of light - what is the question.
And we must to see not only to the walls/celling corners, but to all the scene!
Glossy reflection on tiles for example!
So, let's continue!

[Yaroslav_L]

Can anybody render this scene in some physically correct render - lux or indigo?

[Yaroslav_L]

Thank You Samo, for the settings, but!
Your Photon Map looks good, but final picture isn't so nice and clear...
This is what I get with your settings:


And this one I get with only changing DiffRad to 0.75:


I don't want to start "settings war", but just want to understand what is CORRECT settings for the !REAL! interior scene - not for "cornell box".
Thank You.

[Samo]

Yaroslav

Changing the sun light into a narrow spot light so the full flux of photons gets into the room, that's the basis of my photon map improvement, not some magical combination of settings. I use the spot lamp as a photons cannon, so the photon map patches into the kitchen gets more photons to average from, and there is a better distribution of hits.

Photon map settings depends on the scene size, the lighting setup and the optimisation of the flux of photons, there isn't a mathematical or unique approach, every scene is different. You have to enable Show map button as I've been trying to explain these last days, and 'read' its results.

If fact if you enable Show map in clay mode in the scene above, you'll see it has nothing to do with the map shown here:

[Yaroslav_L]

AHA!!!
The goal was in SpotLight!
Sorry and thank U one more time, Samo!
I'll try it!

[xelptic]

Again some nice tips from Samo..
After reading some articles about PM I was sure that we can visualy test if PMap is good, but I don't have any testrenderings or screenshots to compare with. Now I see that my previous attempts wasn't good enough. Photons must blend together more than I thought was ok.

Also we can see that the "rule": DiffRadius should be twice smaller than wall thickness isn't a rule actually. Radius can be bigger or smaller than wall - proper settings of Photon Emitters (Lights) , Search and Photon count are parameters which also have a great impact on final rendering. Diff Radius isn't MAIN parameter; Radius, Cont and Search are just same important

I agree with Yaroslav that for testing PM its better to have complex scene than simple Cornell Box or Box_with_Holes - as we can see in this tutorial http://www.yafaray.org/documentation/tu ... /photonmap. So Thanks goes for you Yaroslav, for the Blend file

I think we are all "smarter" now, and learn something about PM : )

[Arkinauta]

I think we are all "smarter" now, and learn something about PM : )

I totally agree.
Right now, this thread is the better (and with more info, comparations, etc) place to learn about "correct" PM configuration in Yafaray.

Many thanks to Yaroslav and all the others for making this post so interesting!

[AMFATY]

Guys,you are making a good progress of yafaray in this thread.samo's render is the best result you got.first,when i saw it i thought it was rendered in an unbiased engine like lux, a clean render with some AAsamplings it will be wonderful,.

yes, yafaray needs that types of threads where a one provides his complex scene and the experts deal with it in the thread .this help the users to get the best results like realism and quality of the industry standard engines like vray,and this is good for the developers team,also.

yes,yafaray best workflow still seems unclear for most of the users like PM.

Guys,keep this type of threads.

thanks,

[theguy101]

Hey Samo, do you think its possible to upload the file with your last render settings. I've been tryin to achieve what you did in the last render, but i'm not quite getting it right. The contrast and brithness in yours seems to be perfect, the realism really jumps out.
And btw gracias por el tutorial en espanol!!

thnks in advance

[condar]

I think that the original settings have 2 problems

1.- Too much light
2.- Small window with no glass

Increasing the Sunlamp power (power=4) produces a nice color and great reflections but the exterior and the window frame look overexposed, and glossy reflections generate more noise. no real. So I tried adding glass to the window, more balanced light and pure Glossy Materials, no Glossy as Diffuse. After a few tries I got this. The first image is the render result without any tone mapping saved as PNG. The second image is the same render saved as EXR and tonemapped with Qtpfsgui (Reinhard 02). Not the same reflections as the other examples but remember I added glass to the window, it's an indoor scene illuminated with some Sunlight like and the total light is more balanced.

[Samo]

'As diffuse' enabled in the glossy floor tiles is producing those ugly artifacts, Yaroslav_L and others

[Yaroslav_L]

So, what for this "as diffuse"?

[Samo]

http://www.yafaray.org/documentation/us ... #asdiffuse

[Yaroslav_L]

Samo I know it already!
And this is exactly about my tile mat settings:
glossy=0.2
Exponent=300 - this is low or high? Between 1 and 5000 this is low! So I'm right!

"As Diffuse
When As diffuse is enabled in Photon mapping, it means that the photon map will be used to calculate the glossy reflection, instead of raytracing it, which results in faster render times. In practice, this method is recommended only for glossy surfaces with a low Exponent, since the precision of glossy reflections calculated with photon mapping is always lower. Besides, As diffuse option enabled will be likely to produce flickering of the glossy reflection in animations."