Le 04.12.2008 17:10, clipka nous fit lire :

> Just wanted to point out that the current approach is too limited in my opinion.
> There should be separate (!) options to gamma-correct (or reverse
> gamma-correction) for...
> 
> - preview output window (to adjust for the PoV-ray user's display of course, if
> necessary)

I'm a strong believer that the combination of graphic cards and display
should provide a gamma of 1.0: linear curve;
CRT with 2.2 imply a correction at the graphic cards of 0.45; from there
  the system then has a 1.0 display subsystem and all should be linear.
Opposite of <1.0, 0.5, 0.0> should be <0.0, 0.5, 1.0>, whatever the
number of bits per channel (I like 8, but you need a CRT; LCD do not
have so much bits for all colour... and they might not be linear)

> 
> - image output file (to adjust for the "end viewers'" displays, or
> post-processing steps that may expect a certain gamma correction)
> 
And for image format which support exposing gamma, another one ?
Picture_gamma & Correction_gamma... well find a pair of better names.

> - image input file, on a per-image(!) basis (to compensate for input files that
> may (or may not!) be gamma-corrected for different viewing conditions)

That's a post load gamma correction (beware, PNG for instance already
have one factor from picture: do you want to be able to ignore it ?)

> 
> - color literals, (a) on a scene-global scale and (b) on a per-color basis
> (maybe using macros; to compensate for colors "picked" in an image processing
> software from gamma-corrected pictures)

Global scene should remains linear.
A function/macro for gamma-correction of a colour might be interesting.
(given the computation, a function might be better, but I'm biaised)

> 
> Also note that *all* these settings shold allow for *arbitrary* corrections, not
> just the choice of either "no gamma correction" or "corrected for a gamma of
> 2.2".

Of course!
And please, by default, have a 1.0 everywhere.

Post a reply to this message

Le_Forgeron wrote:
> I'm a strong believer that the combination of graphic cards and display
> should provide a gamma of 1.0: linear curve;

I thought that too, and tried configuring my graphics card that way.
Everything looked washed out. Because almost nobody who makes images or
chooses website colors or whatever has the gamma set that way.

Post a reply to this message

Le_Forgeron <jgr### [at] freefr> wrote:
> I'm a strong believer that the combination of graphic cards and display
> should provide a gamma of 1.0: linear curve;

I'm more of a realist than a believer in this, as unfortunately there is a
*significant* difference between "should" and "is".

If you want a raytracing software for an ideal world - fine. I want one I can
use in the real world.

> > - image output file (to adjust for the "end viewers'" displays, or
> > post-processing steps that may expect a certain gamma correction)
> >
> And for image format which support exposing gamma, another one ?
> Picture_gamma & Correction_gamma... well find a pair of better names.

No, why should there be another parameter for that? It's just another image
output format. If it records the gamma it is assumed to work on without
additional correction, then all the better.


> > - image input file, on a per-image(!) basis (to compensate for input files that
> > may (or may not!) be gamma-corrected for different viewing conditions)
>
> That's a post load gamma correction (beware, PNG for instance already
> have one factor from picture: do you want to be able to ignore it ?)

Possibly. I'd expect reality to come up with PNGs which actually specify a wrong
factor.


> > - color literals, (a) on a scene-global scale and (b) on a per-color basis
> > (maybe using macros; to compensate for colors "picked" in an image processing
> > software from gamma-corrected pictures)
>
> Global scene should remains linear.

One simple question: Why?

Some people may prefer to sacrifice the ability to do simple color additions,
and instead prefer to be able to pick all their colors from other programs (all
the more since PoV-ray doesn't come with a built-in picker).

Internal calculations should remain linear, but why scene? Why not allow people
the freedom to use a different gamma as default for all colors they enter?

Maybe one reason why so many people misused the gamma stuff is because PoV-ray
documentation makes a big fuss about it, saying "hey people, very important
thing: get your gamma things right!" - without PoV-ray providing enough options
to *really* get them right...?


> A function/macro for gamma-correction of a colour might be interesting.
> (given the computation, a function might be better, but I'm biaised)

A macro should definitely suffice - if you want to do calculations, even I agree
that they should be done in 1.0 gamma. So all that's needed should be a macro to
fix the input.


> > Also note that *all* these settings shold allow for *arbitrary* corrections, not
> > just the choice of either "no gamma correction" or "corrected for a gamma of
> > 2.2".
>
> Of course!
> And please, by default, have a 1.0 everywhere.

Personally I don't care about the defaults - as long as they work for people who
don't know what they're doing, and I can set them however I like.

Post a reply to this message

Nicolas Alvarez <nic### [at] gmailcom> wrote:
> Le_Forgeron wrote:
> > I'm a strong believer that the combination of graphic cards and display
> > should provide a gamma of 1.0: linear curve;
>
> I thought that too, and tried configuring my graphics card that way.
> Everything looked washed out. Because almost nobody who makes images or
> chooses website colors or whatever has the gamma set that way.

Yep. Which is because nobody who *views* websites has the gamma set that way.

Simple as that. There's a de-facto standard out there in real life. It's rather
vague, but it's definitely *not* 1.0.

One more reason why I'd like PoV-ray to support arbitrary gamma for output. It
may be wrong from a sort of religious point of view, but it works.

Post a reply to this message

clipka wrote:
> Nicolas Alvarez <nic### [at] gmailcom> wrote:
>> Le_Forgeron wrote:
>>> I'm a strong believer that the combination of graphic cards and display
>>> should provide a gamma of 1.0: linear curve;

That would probably not look good with just 8-bit per color. But I do 
agree that would be the best. I don't think it would happen because in 
the transition from today's default of 2.2 everything would look ugly.

>> I thought that too, and tried configuring my graphics card that way.
>> Everything looked washed out. Because almost nobody who makes images or
>> chooses website colors or whatever has the gamma set that way.
> 
> Yep. Which is because nobody who *views* websites has the gamma set that way.
> 

Most people view with something close to 2.2 I guess.

> Simple as that. There's a de-facto standard out there in real life. It's rather
> vague, but it's definitely *not* 1.0.

The de-facto standard on the web seems to be sRGB with a gamma close to 
2.2 (The true gamma is more complicated, see 
http://en.wikipedia.org/wiki/SRGB).

Thats what my monitors are tuned to anyway.

-- 
Daniel Nilsson

Post a reply to this message

clipka wrote:
> Nicolas Alvarez <nic### [at] gmailcom> wrote:
>> Le_Forgeron wrote:
>> > I'm a strong believer that the combination of graphic cards and display
>> > should provide a gamma of 1.0: linear curve;
>>
>> I thought that too, and tried configuring my graphics card that way.
>> Everything looked washed out. Because almost nobody who makes images or
>> chooses website colors or whatever has the gamma set that way.
> 
> Yep. Which is because nobody who *views* websites has the gamma set that
> way.

Correct. Typical catch-22, particularly hard to get out of...

Post a reply to this message

Nicolas Alvarez <nic### [at] gmailcom> wrote:
> > Yep. Which is because nobody who *views* websites has the gamma set that
> > way.
>
> Correct. Typical catch-22, particularly hard to get out of...

.... or, in this case, a catch-2.2, so to speak...

Post a reply to this message

> -----Original Message-----
> From: Stephen [mailto:mcavoysAT@aolDOTcom]
> Not everyone can use a newsreader at work on a locked down company
> machine.

Do I really need to point out that usually companies who lock down their
machines like this also have policies about using the computer for
personal use?

If you're going to circumvent your company's policies, then be prepared
for trouble.

...Ben Chambers
www.pacificwebguy.com

Post a reply to this message

On Sat, 6 Dec 2008 18:44:11 -0800, "Chambers" <ben### [at] pacificwebguycom> wrote:

>> -----Original Message-----
>> From: Stephen [mailto:mcavoysAT@aolDOTcom]
>> Not everyone can use a newsreader at work on a locked down company
>> machine.
>
>Do I really need to point out that usually companies who lock down their
>machines like this also have policies about using the computer for
>personal use?
>

No, you don't need to point it out.

>If you're going to circumvent your company's policies, then be prepared
>for trouble.
>

Of course! My captain.

-- 

Regards
     Stephen

Post a reply to this message

> I'm a strong believer that the combination of graphic cards and display
> should provide a gamma of 1.0: linear curve;

No, because it's better to have an 8bit number with bigger steps in 
brightness at the higher end, because that's how your eyes work.  If the 
difference in brightness between 4,4,4 and 5,5,5 were the same as 
254,254,254 and 255,255,255 (ie gamma 1.0) then we would see even worse 
"colour banding" for darker colours than we do currently with 8 bit 
hardware, and at the top end there would be an even less visible change for 
each step.

> number of bits per channel (I like 8, but you need a CRT; LCD do not
> have so much bits for all colour... and they might not be linear)

BTW inside most LCDs, the 8 bit input signal will be used to index a look up 
table (which contains the gamma and LC response curves), where the entries 
are 10 or 11 bit numbers that then drive the DACs that directly give the 
voltage to each pixel.  Very cheap LCDs only use 6 bits to index the look up 
table, and then have 8 or 9 bit DACs, but they are becoming rarer and rarer 
nowadays.

Post a reply to this message