Re: [Algorithms] Dynamic Gamma / Tone Mapping

[Emanuele S. <in...@fo...>]

I agree with the tools. In fact we have custom ones...and use Lightwave =
as it's native HDR. (sky generation, volumetric particle baking, to name =
a couple)
I also agree on the limited dynamic range needed. The original =
St.Peter's basilica probe has got a maximum luminance value of ~1100.
Not sure what you mean with "color contrast" ...but if you can generate =
the whole data in HDR and keep it in realtime, you shouldn't experience =
such a loss.

What VALVE did on their last "not-quite-HDR show" could be a simple =
gamma add operation.
Once you have the overall frame's luminance, as said by someone else =
previously, you could proceed like this:

- RTT the framebuffer
- Pass it to a gamma operator (pixel shader) just once
- Add the result on a fullscreen quad N times

The gamma operator is obviously the key to the desired effect.
You might want to have a function that enhances bright values and =
darkens midtones when you get a high overall frame luminance, or maybe =
the other way around, to never get over-exposed nor pitch-black =
lighting....but the latter is against the current common "desired look", =
it seems. :)

Unless they've dropped lightmapping, I doubt they've used real HDR data =
in that demo.

As you can see from our website, we instead manage HDR lightmaps, =
together with dynamic lighting.




Emanuele Salvucci
Maya|Lightwave
Game Technical Artist
Lscript developer

www.forwardgames.com
ema...@fo...

"#3. He causes things to look different so it would appear time has =
passed."

-------------------------------------------------------------------------=
-------




----- Original Message -----=20
From: "Wolfgang Engel" <wengel@RockstarSanDiego.com>
To: <gda...@li...>
Sent: Thursday, December 01, 2005 5:23 PM
Subject: RE: [Algorithms] Dynamic Gamma / Tone Mapping


The idea is to choose 256 luminance values out of the 64k that are =
possible by using 16-bit textures. Choosing the 256 values can happen =
over this whole range and under different criteria. You have a =
middle-grey value that helps you to emphasize a specific range of =
luminance values and with the light adaption result, you move the 256 =
values you choose up and down through the 64k luminance value range.
So the main idea is to choose the best suitable luminance range that =
fits to the scene. The practical problems are
- you do not really have 64k luminance values anywhere. The reason for =
this is that we do not have the tools and teams do not have necessary =
the experience on how to do this. So you end up with a much lower range =
of values.
- as already mentioned: color contrast gets lost
- what I can see in some 360 games now: the light adaption process is =
not very precise it stutters sometimes
=20
I think the most difficult part about HDR/tone-mapping is to establish =
an art pipeline that can produce predictable results. The problem starts =
just by using Photoshop. Try to open up a HDR texture with CS2 and save =
it ... it took some time until this really worked. At the same time we =
tried the latest beta of Photogenics but while trying to adjust the =
gamma value we bumped into a bug .............
=20
One of the things I would love to hear is, how VALVE did that for Lost =
Coast, but I can understand that they do not want to share this :-)

________________________________

From: gda...@li... on behalf of =
Emanuele Salvucci
Sent: Wed 11/30/2005 7:01 PM
To: gda...@li...
Subject: Re: [Algorithms] Dynamic Gamma / Tone Mapping


=20
A question about the model based on the eye perception, and new =
consoles.
=20
Considering X360 and PS3 are HDTV-ready and HDTV screens usually have =
far higher contrast and luminosity than standard PC monitors or TVs, is =
it useful to tone map with that model  ?
=20
At the end of the day, isn't an HDR scene well represented by a 1000:1 =
contrast for most games...?
=20
I'm suspecting it's gonna be a waste of cycles.
=20
=20
=20
Best,
=20
Emanuele Salvucci
Maya|Lightwave
Game Technical Artist
Lscript developer
=20
www.forwardgames.com <http://www.forwardgames.com>=20
ema...@fo...
=20
"#3. He causes things to look different so it would appear time has =
passed."
________________________________


----- Original Message -----=20
From: Wolfgang Engel <mailto:wengel@RockstarSanDiego.com> =20
To: gda...@li...=20
Sent: Thursday, December 01, 2005 1:42 AM
Subject: RE: [Algorithms] Dynamic Gamma / Tone Mapping

Yep, the paper title is "Dynamic Range Reduction Inspired by =
Photoreceptor Physiology". It is published in the IEEE 2005 Vol. 11, =
No.1 January/February 2005, page 13. A short google search only showed =
up the IEEE website where you can buy the paper ... I think I bought it =
from there.
Having bought my third paper from this website now, I am thinking about =
a membership ...

- Wolfgang

________________________________

From: gda...@li... =
[mailto:gda...@li...] On Behalf Of =
Rowan Wyborn
Sent: Wednesday, November 30, 2005 2:32 PM
To: gda...@li...
Subject: RE: [Algorithms] Dynamic Gamma / Tone Mapping


do you have links to either of those papers online? (siggraph or 2005 =
paper)=20

:)

thanks,
rowan





________________________________

From: Wolfgang Engel =
[mailto:gda...@li...] On Behalf Of =
Wolfgang Engel
Sent: Thursday, 1 December 2005 3:52 AM
To: gda...@li...
Subject: RE: [Algorithms] Dynamic Gamma / Tone Mapping


I think good ideas for contrast improvements can be found in Erik =
Reinhardt's 2005 paper ... just forgot the name, but he also talked on =
SIGGRAPH this year about it.
You might for example interpolate the luminance of a pixel with the =
scene luminance or you might store luminance not for RGB per scene, but =
for each color channel separately.
This should help with contrast. Anyone tried this?

- Wolfgang