Additional Outputs: Luminance Output, Z-depth,mat.l/obj. id

[lycium]

ah i see; yeah i was sort of looking at it from a programmer's point of view ("the data is right there").

well, if there's a spec for how the colour-coding should be done (i suppose after tone mapping to account for the potentially huge dynamic range) it should be quite easy to add to voilet. unfortunately i don't have time right now, as i'm just finishing off the last (and heaviest) of my exams for this year; the last one is on the 19th.

thanks for the maxwell link! i didn't dl it yet because of the lack of integrated studio thingy and the registration, but i did check out fryrender (because of that small drama in the other thread). i did see the different channel options in that renderer but didn't check it out because, until this forum thread, i didn't think about what complications this might introduce for a renderer.

regarding sample scenes, man you have no idea how much i'm starving for that; i'm also writing a little mlt-based renderer, started just before exams, and eventually i'll need some reasonable data to feed it. i was thinking of following the indigo format, however... most of the examples that come with indigo don't work!! it would be SO good for the community if:

1. someone would make an example scene package, where they all work
2. mlt is on by default!

regarding point 2, a few times now i've seen people posting first renders here which have the telltale "lone fireflies" that seem to indicate a lack of mlt. i know nick renders with mlt off usually (!), but i think artists working with transparent surfaces (e.g. that recent pens scene) it should be on by default just in case i think.

well, now i've just spammed like crazy and am rambling :S hopefully i'll find the time to work on violet if there is a concrete feature request, and please don't be shy to forward cool scenes for me to render with indigo / maxwell / fry until exams are over :D that would be really awesome.

regards,


thomas

ps. i just checked the maxwell page for demo downloads, and it said "Se hai già un account registrato", which i think is italian for "register an account you lazy bum" yes? hmm maybe i should.

[bkircher]

Depending on the background of every user, different things might be important, and might be possible to be changed by the user, or not.

Lyc, that luminance-output would be positively cool from you to realize!
Apart from that, before getting upset, please read the posts:
I was talking of sampling "once" during "preprocessing". That would be similar to Mental Ray, especially if an ID channel puts out only one colour without antialiasing. I don't now if indigo uses something like a preprocessing stage before entering regular rendering.

Finally, as far as "lazy", "stupid" and "complaining" are concerned, please calm down and get a little less judgmental.

[psor]

@lyc

... hmm, ... I can tell you, you don't have to register. Go to download, upper right,
then klick on Maxwell Render Demo Version ... and you're there. Or just klick here!

And for the scenes, just give us some infos about the file format that you
would like to have, ... OBJ, 3DS, Indigo XML ... what ever. Just let us know.

And thanx for the offer to tune in Violet! So ... I wish you success with your
exams - good luck and see ya around!



take care
psor

[Kram1032]

Wohoo! Finally someone, who's gonna work on Violet, again

[WytRaven]

um...did he actually say he was going to work on it? All i see is I'm working on yet another MLT renderer and I don't have time to work on Violet...

[Kram1032]

He said, he doesn't have time, due to exams and the last exam will be 19th

[lycium]

Or just klick here!

thank you!

And for the scenes, just give us some infos about the file format that you would like to have, ... OBJ, 3DS, Indigo XML ... what ever. Just let us know.

i was a bit unclear in my post, actually talking about two issues; the first is that indigo needs a really nice example scene package, since most of the included ones don't work and there's no real nice interior or something like that.

the other is that eventually i'd really like to try render some nice 3d models in my new renderer, and good scenes are always difficult to find! so much so, that there's a special section on the ompf ray tracing forum just for that: http://ompf.org/forum/viewforum.php?f=7 (contribute there to be an instant hero! )

And thanx for the offer to tune in Violet! So ... I wish you success with your exams - good luck and see ya around!

thanks

btw, i had a simple idea for how it can be done without depending on tonemapping: just have some correspondences between luminance and rgb colour (user supplied), and then it interpolates. the other option is the obvious one, to just have some palette of colours which map directly (linearly) to the luminance, but since the luminance is usually logarithmically distributed (or needs tonemapping in general) it'll plug into some tonemapping code.

btw wytraven, my renderer has barely even been started and is completely unusable; even if that weren't the case i wouldn't be advertising it on here. oh and your robot project looks awesome

[dougal2]

a luminance output from violet would be very simple.

*has worked on violet*

edit: *looks at code* I used luminance values for my original histogram implementation.
BTW, luminance is nothing more than a weighted average of the RGB values:

Code: Select all

float Histogram::RGBLuminance(const Colour3& pval)
{
    //convert RGB to Luminance - CCIR 601
    return  (0.299*pval.r + 0.587*pval.g + 0.114*pval.b);
    
}

[lycium]

a luminance output from violet would be very simple.

that's what i said, but it seems what is desired isn't the luminance itself but a false-colour image of the luminance.

BTW, luminance is nothing more than a weighted average of the RGB values

that's the easiest way to get the low dynamic range to false colour mapping; however there was talk of using it as a lighting analysis tool, so directly using the computed luminance (Y channel from XYZ) without any post processing (i.e. it's exactly the radiance along the camera ray or irradiance at the image plane, in units wm^-2s^-1 or wm^-2) would be preferred for this.

so i guess, there can be two modes:

1. just re-palette the rgb-blended luminance after tonemapping etc. (2 second job in photoshop/gimp btw)
2. allow the user to set up (luminance,false-colour) pairings and interpolate among those.

[OnoSendai]

Luminance output (in false colour) seems like a reasonable idea, and fits quite well with the general direction of Indigo.
I think it could be slotted in as another tonemapping option.
Z-depth and mat-ids are a bit more problematic, because due to DOF blur and antialiasing, you can have multiple depths and mat-ids per pixel.
Still, I guess i could just do something as a pre-pass as suggested.

[bkircher]

Great !

As to ambiguities on a pixel level (or below with supersampling),
that would not be that critical, i think. The option of a different supersampling setting for the prepass compared to the "real thing" would be great though.

... and thanks for looking into that matter

[CTZn]

Fine, now we'll have a scanline algorithm in Indigo

Luminance in real shades of gray is of some interest as well, for manual tuning. Thanks.