[mtripoli]

I'm following it (not!)... but then again, I spend alot of my time saying

"...you want fries with that?...



MT

[tom]

if the fries come with an unbiased sauce, no matter...

[Micha]

... This noise error will converge by its own to the solution of the render equation with no user intervention (no parameters) and the benefits are that there is no need for user settings. The smoothing error does not converge by its own and it needs the user to see and try again until he gets with something that he likes. ....

Hi, could it be possible, that Maxwell smooth the noise at lower SL and during the rendering the noise go and the smoothing too? A synthesis of the fast classical biased way and the new slow perfect maxwell way.
Biased rendering is not the problem for, if the rendering will be better and better during the great "cooking" method.

[iker]

Maybee maxwell and vray should combine forces? oooo

...and Pepsi and Coca-Cola too!

[MetinSeven_com]

i think both renders are great the artist, is what counts not the tool , but everyone fit and likes one tool to work....

I agree with maestro Zuliban. Like a number of other guys over here I use both Maxwell and V-Ray.

I love Maxwell because of its stunning photorealism and straightforward "less is more" controls.

I love V-Ray because of the way it allows you to customize the result to your liking and because of its seamless 3ds Max integration.

Etcetera etcetera. Every renderer has its advantages and disadvantages, but I think it's the artist's refreshing visual concept and a convincing scene execution that counts in the first place. I've seen good and bad results rendered with both V-Ray and Maxwell. No matter how great a renderer is, it's the artist that has to convince and impress the viewer.

Just my $0.02.

Cheerio,

Metin

[Thomas An.]

Hi, could it be possible, that Maxwell smooth the noise at lower SL and during the rendering the noise go and the smoothing too? A synthesis of the fast classical biased way and the new slow perfect maxwell way.
Biased rendering is not the problem for, if the rendering will be better and better during the great "cooking" method.

As much as I am a Maxwell purist, Micha may have a point to think here.

-->What if Maxwell started biased and ended unbiased in time ?

(on the other hand, if Oscar has a say in it then maybe the Maxwell engine can be optimized so much and be so fast that any biased methods will become obsolete).

[MetinSeven_com]

Personally I don't think NL should change their unbiased approach. Concessions towards biasedness would undoubtedly affect the unparallelled realism of the rendered results.

I am quite satisfied with the current unbiased approach and don't mind filtering some last noise using a post-processing tool. I guess NL should focus on:

• more rendering acceleration without affecting the unbiased rendering result,
• a powerful material editor,
• the new and improved sunlight/moonlight system,
• necessary functions like the possibility to control which object is visible in the alpha channel and matte integration,
• more seamless integration into each piece of 3D software.

Cheers,

Metin

[MetinSeven_com]

Hi Flipside,

As far as I understand, "unbiased" means that the renderer makes no concessions regarding approximating a result as realistic as possible. In other words: there are no options to turn off reflections in reflective surfaces, no options to turn off DOF, no options to suppress the falloff of light strength, no blurring methods to suppress noise, no choice of different anti-aliasing methods, etcetera. Maxwell aims to mimic the physical properties of the real world as close as possible, without options to make the result biased.

Cheers,

Metin

[KRZ]

maxwell should stay unbiased till the end. if you want additional noise reduction try the new photoshop..it has an improved filter to do that

[j_petrucci]

flipside, biased doesn't mean "noisy" and unbiased doesn't mean "noiseless"; in math terms, is like the process to numerically evaluate an integral: you start approximating its area with big rectangles and the results is less than the correct one if you take the inscribed ones, greater if you take the circumscribed ones.

having the opportunity to take infinite infinitesimal rectangles, you could tell the exact area value summing all of them, but you can't go forever, so the you always have an approximating error, i.e. the exact value minus your current value you are iteratively searching; well this error is the noise you still find in the image Maxwell is rendering, thus it samples finer and finer each iteration (less sparse illuminated points) and the points not yet calculated near the calculated ones gives you that noise impression.

a biased render doesn't converge to the exact solution, it directly make prior approximations on the lightning model and interpolate the data your settings permit it to reach, it doesn't go further than that.

take for example the ray number setting of standard radiosity renderers: you can tell how many light rays rebounds from the surfaces. this is clearly a biased approximation! surfaces hit by light reflect infinite rays in every directions, so "ordinary" renderers has to take only some of them and interpolate the results.

Maxwell instead takes into consideration all of them and begin to calculate light points evenly distributed in every directions... the more time you give it, the more points it can compute and less "noise" you can see on the surfaces.

and "Maxwell defeat the rules of raytracing", well... IMHO both agree and disagree: I agree regarding the way Global Illumination is calculated (totally physically accurate), disagree about the word "defeat" because the rules light rays follow are always the same, the only thing that changes is how they are considered.

that's all folks!

[MetinSeven_com]

Thanks for this extensive elucidation, Mr. Petrucci.

Flipside, the noisiness of an early rendering stage doesn't mean that the result isn't already unbiased. That Maxwell is unbiased means that the result isn't affected by settings that bend reality. Maxwell works towards a result as realistic as possible (eliminating noise), while it is being unbiased.

Cheers,

Metin

[MetinSeven_com]

I still think you are confusing "unfinished" with "unbiased", Flipside. See my previous post.

Cheers,

Metin

[MetinSeven_com]

By the way Flipside, do you actually use Maxwell or are you a V-Ray only user?

[trick]

I can't understand why it should be a problem for some to make algorythms that "bias" the "unbiased" output to get noiseless results faster. If you insist on unbiased output you simply switch things off and wait a few hours/days more. Even the eye transmits biased images to the brain (did you ever walk into a building on a sunny day), so why would you want physical accuracy you don't need. There even are a people that only see what they want to see. If background processes stay unbiased, I'm really OK with some customisable "biasing" filters

I don't like it when I have to open PS to denoise my renders.
I also don't like it when I have to go to another renderer just because I need some invisible lights casting light or shadows or invisible objects cast some shadows or make reflections...

[MetinSeven_com]

If the biasing would be optional then I wouldn't mind. I just don't want blurring algorithms and other biased methods to infect the current impressive Maxwell rendering results. The unbiased approach is one of Maxwell's biggest features after all.