What's going on here?

The first shot is a classic POV-Ray radiosity shot, to give an impression of
what I'd expect the "final" MCPov results to look like.

The other three shots are MCPov shots of the very same scene...

- after the first pass (34 minutes)

- after the 15th pass (about 7 hours)

- in a low-res version *not* using a portal on the broken glass  (plus some
subtle differences in the height field, but those are minor; something like
180th pass, 11 hours; scaled up here to match the other shots)

Note...

(A) how the broken glass has started out almost opaque in the MCPov high-res
shots, and has developed bright specks since then; the same effect was seen in
the low-res version, too - the interim result indicates that MCPov is trying
hard to make up for the initial error with some really "heavyweight" rays, and
the result will eventually be correct, but the performance is definitely poor;

(B) the arced shadow in the MCPov high-res shots to the right of the broken
glass, that has come out of nowhere; as this, too, gets riddled with bright
specks over time, I suspect it will be cancelled out eventually - but why is it
there in the first place? It's not in the low-res shot, so it must have
something to do with the (bounding box) portal I defined on the glass, but it's
much larger than the box.

Can someone give me a clue how to fix these issues? In this manner, it will take
ages before I get any decent result...

Post a reply to this message

Attachments:
Download 'lostcity-mcpov-montage.jpg' (235 KB)

Preview of image 'lostcity-mcpov-montage.jpg'

"clipka" <nomail@nomail> wrote:
Ah well, looks like I myself am the first one to come up with a (partial)
solution for these issues:

- When I added the portal for the broken glass, I also added
mc_check_portal_overlap; it seems that this also requires using mc_sort_portals
(at least for my shot). This eliminates the arced shadow (which, as it seems,
was due to the glass portal "shadowing" the sun portal).

- The issues with the glass not being sufficiently transparent seem to have been
due to insufficient number of samples specified for its portal, though I have no
clear idea why that should be. Somewhat cranking up the number of samples mostly
solved the problem (though not completely, as it seems).

Post a reply to this message

"clipka" <nomail@nomail> wrote:
> "clipka" <nomail@nomail> wrote:
> Ah well, looks like I myself am the first one to come up with a (partial)
> solution for these issues:
>
> - When I added the portal for the broken glass, I also added
> mc_check_portal_overlap; it seems that this also requires using mc_sort_portals
> (at least for my shot). This eliminates the arced shadow (which, as it seems,
> was due to the glass portal "shadowing" the sun portal).

Are you sure you want a portal for the glass?  For outdoor scenes, I have had
good luck with just a portal for the light source and mc_sky if the sky is
non-uniform.  A portal may be required to get very sharp caustics, but at this
oblique angle, it should not be critical.  You might try just increasing the
primary diffuse samples for the ground instead.

 - Ricky

Post a reply to this message

"triple_r" <nomail@nomail> wrote:
> Are you sure you want a portal for the glass?

No, I'm not sure about anything regarding MCPov ;)

The web page states that a portal is good if you want good caustics, and I do
want good caustics...

Maybe I need a proper definition of "good" though...

Another thought though: Without the portal I had the same problem with the glass
not looking properly transparent to begin with, and then MCPov collecting bright
dots to make up for it. Which is not the most performant way to get a good shot
I guess, so maybe I *do* want the portal.

Post a reply to this message

For caustics you might be better off using standard POV-Ray, as photon 
mapping is implemented very well. Then you could add dispersion to really 
make the glass sparkle.


"clipka" <nomail@nomail> wrote in message 
news:web.49a7f07e336c2d8d8d0daedf0@news.povray.org...
> "triple_r" <nomail@nomail> wrote:
>> Are you sure you want a portal for the glass?
>
> No, I'm not sure about anything regarding MCPov ;)
>
> The web page states that a portal is good if you want good caustics, and I 
> do
> want good caustics...
>
> Maybe I need a proper definition of "good" though...
>
> Another thought though: Without the portal I had the same problem with the 
> glass
> not looking properly transparent to begin with, and then MCPov collecting 
> bright
> dots to make up for it. Which is not the most performant way to get a good 
> shot
> I guess, so maybe I *do* want the portal.
>
>

Post a reply to this message

"Mike Hough" <nos### [at] nospamcom> wrote:
> For caustics you might be better off using standard POV-Ray, as photon
> mapping is implemented very well. Then you could add dispersion to really
> make the glass sparkle.

I do want the caustics, but my main concerns are others:

- Some subtle lighting effects that don't seem to work with radiosity

- Blurred reflections & refractions

- Toy around with MCPov

Post a reply to this message

"clipka" <nomail@nomail> wrote in message 
news:web.49a8936a336c2d8d40ae432c0@news.povray.org...
>
> - Blurred reflections & refractions

MegaPOV used to have blurred reflection and refraction. I rather liked the 
feature because it was easy to use.

Post a reply to this message

"Mike Hough" <nos### [at] nospamcom> wrote:
> "clipka" <nomail@nomail> wrote in message
> news:web.49a8936a336c2d8d40ae432c0@news.povray.org...
> >
> > - Blurred reflections & refractions
>
> MegaPOV used to have blurred reflection and refraction. I rather liked the
> feature because it was easy to use.

"Used to" is the keyword here. The version I use - 1.2.1 - doesn't seem to, and
I'm not going to install yet another POV (derivative) version on my system... I
already have POV 3.6.1c, MegaPOV 1.2.1, POV 3.7.0.beta, MCPov 0.0.5 and about
three or four POV 3.7.0 branches for radiosity development...

Post a reply to this message