On 20 Feb 2008 15:09:52 -0500, Jim Henderson <nos### [at] nospamcom> wrote:

>
>I've seen photos that don't look photorealistic to me.  It most certainly 
>is a matter of opinion.

As an aside is a black and white photograph, photorealistic, or a sepia one? Are
the Pre-Raphaelites or chocolate box paintings?

Regards
	Stephen

Post a reply to this message

>   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.

I simply calculate the standard deviation of the pixel values. A decent metric
is to see how the change in variance goes down as image converges.

Maybe this should be done locally where there are area of same color.

Anyway, it can be done.

Post a reply to this message

Severi Salminen wrote:
>> FWIW, I think it would certainly be interesting to have unbaised
>> rendering as an option on POV-Ray. But the amount of work required is,
>> realistically, prohibitive. You'd have to basically rewrite the whole
>> program. And I don't see that happening any time soon...

> How wrong you are :)

Perhaps.

> Just check out the images group. Fidos already implemented simple but fully
> working brute force to Pov.

Really? That's interesting. So what's the catch? ;-)

> So no, There is no need to rewrite the whole
> program. Brute force might make it actually a lot simpler.

Yes, I see how two algorithms instead of one would be a lot simpler... 
Oh, wait...

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

Post a reply to this message

"Gilles Tran" <gitran_nospam_@wanadoo.fr> wrote:

> The area_illumination feature that Warp introduced in the latest 3.7 beta is
> one of those things that the previous versions couldn't do: of course one
> could simulate it with grids of point lights, but it was just too
> impractical for common usage. Another feature that is sorely missing is
> efficient blurred reflection. There's a trick to do that in POV but the
> results are usable in only certain (limited) circumstances.

This is exactly what I don't like in POV. You can't just throw a scene to it and
let it render it accurately. You have to enable many kind of features and you
have to guess which features your scene actually needs. POV can't decide it for
you.

This is a problem both for developers and users. Developers have to implement
many special cases and tricks to get the desired effect. Brute force renderers
don't need them because many lighting effects are "automatically" generated.
With brute force you don't need area_illumination kind of keywords. You can use
any kind of light source and the result is always correct.

The same goes with blurred reflections. It is not too difficult to have any kind
of BRDF and the result is still properly antialiased etc. No need to have
special tricks to mimic blurred reflections. But I will verify this statement
when I have implemented them myself...

Post a reply to this message

Severi Salminen wrote:

> This is exactly what I don't like in POV. You can't just throw a scene to it and
> let it render it accurately. You have to enable many kind of features and you
> have to guess which features your scene actually needs. POV can't decide it for
> you.

This is what I don't like about GPUs and scanline rendering. Everything 
is textured polygons; the rest is lashings and lashings of deceptive 
trickery to make it *look* like the real thing. But with POV-Ray, if I 
ask for a sphere, I get a sphere. Not some polygon mesh approximating a 
sphere, but AN ACTUAL SPHERE. You can construct shapes of arbitrary 
complexity. Surfaces and textures can be magnified arbitrarily and never 
look pixellated. Reflections JUST WORK. Refraction JUST WORKS. Etc.

I can certainly see the advantage of a "I just throw objects in and it 
works" approach to lighting. But then, that's more or less how POV-Ray's 
radiosity feature works. You usually don't have to twiddle the settings 
all *that* much - it's more a question of how many years you're willing 
to wait for the result. And that's the kind of worrying part - how many 
years will you have to wait for the result from an unbiased renderer?

(OTOH, the fast preview you can get sounds like a useful feature. Ever 
wait 6 hours for a render only to find out that actually it looks lame? 
It's not funny...)

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

Post a reply to this message

> > Just check out the images group. Fidos already implemented simple but fully
> > working brute force to Pov.
>
> Really? That's interesting. So what's the catch? ;-)

I don't know of any catch. The source is still not available, though. The images
look good to me.

> > So no, There is no need to rewrite the whole
> > program. Brute force might make it actually a lot simpler.
>
> Yes, I see how two algorithms instead of one would be a lot simpler...
> Oh, wait...

Of course I meant that bruteforce might be simpler alone than present mix of
many different special cases which need to be taken care of. Of course having 2
methods is more complex than having just one...

Post a reply to this message

Does anybody, anywhere, have any details of how the algorithm actually 
works? As in, how is this different to a normal ray tracer?

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

Post a reply to this message

scott wrote:
>> FWIW, I think it would certainly be interesting to have unbaised 
>> rendering as an option on POV-Ray. But the amount of work required is, 
>> realistically, prohibitive. You'd have to basically rewrite the whole 
>> program. And I don't see that happening any time soon...
> 
> Not just the rendering method, but things like different reflection and 
> lighting models, newer methods of increasing the efficiency of ray 
> tracing (I posted a link in the pov4 group), etc.

OK. I wasn't aware that any existed, but hey.

> Given that most 
> people seem to think SDL is POV's strongest point, why not improve the 
> SDL to be more flexible?

Sounds good to me...

>>> You do realise that the big rock in the Cascades demo is an isosurface?
>>
>> Really? I thought it was just a tellesated triangle mesh based on an 
>> isosurface? (Remember, I haven't actually been able to watch the demo 
>> yet.)
> 
> Well yes, of course, nothing can directly show an isosurface, even POV 
> has to sample the function to generate pixels.  But my point was it 
> shows an isosurface in realtime, in fine detail.

My point is that usually, no matter how closely you look at a POV-Ray 
isosurface, it will always be beautifully smooth. Every NURBS demo I've 
ever seen for a GPU has been horribly tesellated with sharp edges 
everywhere. Sure, if you had several billion polygons, maybe you could 
almost approach what POV-Ray gives you... but presumably that would 
require slightly more than 256 MB of video RAM.

POV-Ray, of course, gets round this problem by using more sophisticated 
mathematical techniques than simply projecting flat polygons onto a 2D 
framebuffer. I've yet to see any GPU attempt this.

>> OOC... Clearly Crysis has some pretty serious graphics. But is it 
>> *fun* to play?
> 
> I only played the demo, and it was in a bit of a rush, seemed pretty 
> similar *gameplay* to FarCry, which isn't a bad thing.  Played fine on 
> my nVidia 7900 card, I think I had low or medium detail and it was above 
> 30fps most of the time.  I think you'd need to play a few levels before 
> any new gameplay became apparent, just IMO.

Mmm, OK. Well my graphics card is only a GeForce 7800 GTX, so I had 
assumed it would be too under-powered to play it at much about 0.02 FPS. 
 From the way people talk about Crysis, I was under the impression that 
you need a four-way SLI rig of 8900GTX GPUs just to make it get out of 
bed... Apparently it's not as bad as that.



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

Post a reply to this message

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

The Hulk character and rendering is quite good.  It's so good in fact 
that it's minor details that will set it apart from a real person.  That 
is, apart from the fact that he's a big green hulking jumping like a 
grasshopper! :D

it's the so called uncanny valley effect.  Read more on wikipedia...

Post a reply to this message

> Reflections JUST WORK. Refraction JUST WORKS. Etc.

Reflections and refractions don't just work. You see no caustics at all unless
you enable them and tweak settings. You see incorrect shadows if you don't use
the proper keywords/settings. Etc. Many things don't work if you don't enable
them.

> I can certainly see the advantage of a "I just throw objects in and it
> works" approach to lighting. But then, that's more or less how POV-Ray's
> radiosity feature works. You usually don't have to twiddle the settings
> all *that* much - it's more a question of how many years you're willing
> to wait for the result. And that's the kind of worrying part - how many
> years will you have to wait for the result from an unbiased renderer?

I have seen so many scenes where you still get radiosity artifacts no matter how
hard you try. I'm not an expert so I might be wrong. And with max settings the
speed difference is not _that_ big after all.

> (OTOH, the fast preview you can get sounds like a useful feature. Ever
> wait 6 hours for a render only to find out that actually it looks lame?
> It's not funny...)

Yeah, since each pass in brute forcer is pretty fast, you get to see something
very quickly. And you can quickly decide if the scene needs more tweaking.

Post a reply to this message