Nekar wrote:
> "Darren New" <dne### [at] sanrrcom> wrote in message
> news:47bca80c$1@news.povray.org...
>> Nekar wrote:
>>> Do scientist completely understand all the
>>> properties of light?
>> Pretty much, yes. :-)
> 
> But not _perfectly_. if they understood light perfectly there wouldn't be
> any search for unified field theories, etc. :o]

They're trying to unify the stuff they understand less (nuclear forces, 
gravity) with their understanding of light. QED's theory matches 
experiment to 15 decimal places.

Of course, the "why does it do that" will (almost certainly) never be 
understood down to the lowest level. But the properties of light? Yeah, 
I don't think there's any inability to predict what light will do under 
various circumstances. Unless, of course, someone comes up with 
experimental evidence that doesn't fit, which is always a possibility.

-- 
   Darren New / San Diego, CA, USA (PST)
     On what day did God create the body thetans?

Post a reply to this message

> I have to say, I've yet to see any unbaised renders where I instantly look 
> and go "wow! POV-Ray could never do that..."

I don't see much nowadays where I say "wow, a modern GPU could never do that 
in realtime" ;-)

It's just a matter of realism and effort.  At the top you have these 
unbiased renderers that take almost no setting up (apart from the scene 
itself) and give a scientifically perfect result after a few days.  At the 
bottom you have GPU rendering, that takes a lot of effort to code, gives a 
pretty good result and takes a few ms to render.  POV is in between 
somewhere, taking a medium amount of effort to get looking good, and taking 
a few hours (not days or ms) to render.

Post a reply to this message

scott wrote:
>> I have to say, I've yet to see any unbaised renders where I instantly 
>> look and go "wow! POV-Ray could never do that..."
> 
> I don't see much nowadays where I say "wow, a modern GPU could never do 
> that in realtime" ;-)

How about, say, volumetric fog? ;-)

I have yet to see any GPU of any description get that right...

For that matter, I've yet to see any GPU do physically correct 
reflections [although surely it can't be *that* hard?], nor global 
illumination that isn't pre-computed [and hence won't change when 
objects move around].

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Warp wrote:

>   I wonder if an automatic measurement and then a threshold couldn't be
> developed. For example, if a given pixel hasn't changed color for the
> last n rays which have affected that pixel, then that pixel is done.
> When all the pixels fulfill this requirement, the image is done.

I don't know about unbaised rendering, but the Fractal Flame algorithm 
is somewhat similar. (You stocastically create an image that starts off 
grainy as hell and eventually progresses to being smooth.)

The author's rendering tool allows you to specify a "quality" setting. 
However, the actual quality of the image isn't real closely related to 
this. It turns out that the quality setting is actually just the average 
number of samples/pixel that must be reached before rendering stops. 
This probably wouldn't be a very good measure for unbiased rendering...

A better measure would seem to be to measure the signal to noise ratio 
of the image - but then you'd need a way to seperate "signal" from 
"noise"...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

> How about, say, volumetric fog? ;-)
>
> I have yet to see any GPU of any description get that right...

A couple of papers show how to do realistic physics-based smoke/cloud true 
volumetric effects (not just static fog):

http://developer.download.nvidia.com/presentations/2007/gdc/RealTimeFluids.pdf
http://www.markmark.net/dissertation/harrisDissertation.pdf

Also real-time volumetric lighting is no problem, in fact Crysis uses it and 
I'm sure there's an nVidia demo kicking about somewhere.

> For that matter, I've yet to see any GPU do physically correct reflections 
> [although surely it can't be *that* hard?],

The reason there is no effort put into doing this is because doing the 
reflection/environment map method is good enough.  In fact with DX10 and 
recent cards the emphasis has gone into speeding up the generation of the 
environment map.  Doing physically correct reflections is hard because you 
need to write a ray tracer inside the pixel shader - it's *a lot* of work 
for very little improvement in realism, which is why hardly anyone bothers, 
unless you are writing some proof-of-concept demo.

> nor global illumination that isn't pre-computed [and hence won't change 
> when objects move around].

http://realtimeradiosity.com/demos/

Real time ambient is done in the "Cascades" demo from nVidia.  Each voxel on 
the surface traces out (32 IIRC) rays to work out the GI at that point.  OK 
so it's not perfect (other geometry doesn't affect it) but it's certainly 
better than no GI.

What you have to bear in mind with all these things, is that they have to be 
designed to work when a frame takes 15 ms to render.  If reducing the 
accuracy by 10% speeds up rendering by 50%, the speed-up option is taken. 
Always.

Post a reply to this message

nemesis wrote:
> Jim Henderson wrote:
>> On Wed, 20 Feb 2008 16:44:59 -0300, nemesis wrote:
>>> like I said, photorealism is not a matter of opinion.
>>
>> Sure it is.  It all depends on one's perception of the image.  You 
>> might look at an image and say "gee, that looks really 
>> photorealistic", and I might respond "are you on drugs?  Look at that 
>> shadow, that's clearly not right".  You might disagree about that 
>> particular shadow (whatever it is).
> 
> if I saw at a sharp shadow and said it was photorealistic, I'd sure be 
> on drugs. ;)

Any shadow viewed from sufficient distance will appear razor sharp. 
Using sharp shadows for large-scale objects and scenes can be completely 
photorealistic.

>> Have you ever been to a movie with someone who thinks the CG effects 
>> are outstanding and "the most realistic effects they'd ever seen", 
>> only to tell them that they were crap effects?

I've certainly heard people say they thought the effects were rubbish 
when I thought they were exemplary... the Hulk, for one.

Post a reply to this message

Warp wrote:
> nemesis <nam### [at] nospamgmailcom> wrote:
>>>   POV-Ray is simply a great tool to create "3D'ish" images like those.
>>> It's easy and fast.
> 
>> so, pov-ray will end up its days as a 3D button design tool?  Are 
>> pov-ray users ok with that evolutionary idea?
> 
>   Why are you trolling?

I'm making justice to my nickname.  I want povray to evolve.  Being 
provocative is a way to do that.  You may call me names, I don't care.

Post a reply to this message

nemesis wrote:
> Sure, you can endlessly tweak the povray scene or settings in other 
> biased renderings to get very photorealistic results.  You may also 
> endlessly tweak the light sources and radiosity/photon mapping settings 
> to get the illumination just right.  You can do nothing at all about 
> aliasing of edges against very bright backdrops.
> 
> Or you can just buy top hardware, model, texture and drop accurate 
> lighting in your scene and let an unbiased rendering method handle it 
> overnight.

I find that most of my tweaking in POV is for 'artistic' reasons - 
making sure shadows don't line up with edges, or that the light is in 
the right place to make the objects look right/good, or that objects are 
placed and textured naturally. Only occasionally do I find myself 
tweaking for image quality.

Post a reply to this message

scott wrote:
>> How about, say, volumetric fog? ;-)
>>
>> I have yet to see any GPU of any description get that right...
> 
> A couple of papers show how to do realistic physics-based smoke/cloud 
> true volumetric effects (not just static fog):
> 
> http://developer.download.nvidia.com/presentations/2007/gdc/RealTimeFluids.pdf 
> 
> http://www.markmark.net/dissertation/harrisDissertation.pdf
> 
> Also real-time volumetric lighting is no problem, in fact Crysis uses it 
> and I'm sure there's an nVidia demo kicking about somewhere.

All I know is that every time I play HalfLife 2 (hailed for it's 
ground-breaking graphics), I'm struck by how cheese and lame all the fog 
effects look.

Crysis? Isn't that that game that requires a small render farm to play? ;-)

>> For that matter, I've yet to see any GPU do physically correct 
>> reflections [although surely it can't be *that* hard?],
> 
> The reason there is no effort put into doing this is because doing the 
> reflection/environment map method is good enough.

Probably.

>> nor global illumination that isn't pre-computed [and hence won't 
>> change when objects move around].
> 
> http://realtimeradiosity.com/demos/
> 
> Real time ambient is done in the "Cascades" demo from nVidia.  Each 
> voxel on the surface traces out (32 IIRC) rays to work out the GI at 
> that point.  OK so it's not perfect (other geometry doesn't affect it) 
> but it's certainly better than no GI.

Cascades sounds interesting. I'd certainly like to watch it. However, it 
requires a more expensive GPU and a more expensive OS before it will 
even consider running, so that's kind of the end of that.

> What you have to bear in mind with all these things, is that they have 
> to be designed to work when a frame takes 15 ms to render.  If reducing 
> the accuracy by 10% speeds up rendering by 50%, the speed-up option is 
> taken. Always.

Yes. And that is why POV-Ray can do things that a GPU can't. POV-Ray 
isn't *trying* to be real-time. ;-)

I could add things like isosurfaces to the list. (Have you ever seen a 
game where the water *actually ripples* rather than just surface normal 
tricks?) The point is, POV-Ray has vastly more time to spare, so of 
*course* it can do a better job than any GPU can. Hardly surprising, really.

Just don't try to tell me a GPU can do everything POV-Ray can. ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

nemesis wrote:

> I want povray to evolve.  Being provocative is a way to do that.

No - being provocative to the point of trolling is a way to get 
everybody to completely ignore you. ;-)

> You may call me names, I don't care.

Oh the irony...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message