A generic rule of thumb if you rolling your own. Stars in the real are
visible by virtue of their brightness not their angular size. In the POV
world it is the angular size that makes an object visible. 

-- 
I am not sensitive. I failed sensitivity training.
Sue me. 
	-- The Iron Webmaster, 150

Post a reply to this message

"Matt Giwer" <jul### [at] ijnet> wrote in message news:3A1639FF.50EFCF48@ij.net...
> A generic rule of thumb if you rolling your own. Stars in the real are
> visible by virtue of their brightness not their angular size. In the POV
> world it is the angular size that makes an object visible.

That's an untruth actually.
Take a granite color map, for example, which disperses a star-like pattern with parts
(dots) being scaled too
small to be seen (when AA is used especially, not sure at the moment if this does in
all cases) and then try an
increase in color value.  I think the previously invisible pixels will then appear.
You might be right though about it when non-AA rendering.
Same goes for objects too I think.  Seems I've encountered that before.

Bob

Post a reply to this message

"Bob H." wrote:

> That's an untruth actually.
> Take a granite color map, for example, which disperses a star-like pattern with
parts (dots) being scaled too
> small to be seen (when AA is used especially, not sure at the moment if this does in
all cases) and then try an
> increase in color value.  I think the previously invisible pixels will then appear.
> You might be right though about it when non-AA rendering.
> Same goes for objects too I think.  Seems I've encountered that before.

Correct, it only works with AA. Still, though, only to an extent and some still
disappear entirely.

If a star is "between the cracks" AA will completely miss it, because neighboring
pixels will be black. In the case
of getting caught and aa'd, the aa will of course decrease brightness as the star is
only part of the pixel.
Angular size is still the predominant factor.

It is very annoying to create good stars in POV when changing quality settings from
one render to another!

--
David Fontaine  <dav### [at] faricynet>  ICQ 55354965
My raytracing gallery:  http://davidf.faricy.net/

Post a reply to this message

"Bob H." wrote:
> 
> "Matt Giwer" <jul### [at] ijnet> wrote in message news:3A1639FF.50EFCF48@ij.net...
> > A generic rule of thumb if you rolling your own. Stars in the real are
> > visible by virtue of their brightness not their angular size. In the POV
> > world it is the angular size that makes an object visible.

> That's an untruth actually.

	In what way? 

> Take a granite color map, for example, which disperses a star-like pattern with
parts (dots) being scaled too
> small to be seen (when AA is used especially, not sure at the moment if this does in
all cases) and then try an
> increase in color value.  I think the previously invisible pixels will then appear.
> You might be right though about it when non-AA rendering.
> Same goes for objects too I think.  Seems I've encountered that before.

	Let me try it again. There are several ways to go for a starfield. The
optimize around "not doing dumb things that cost parsing/rendering
time." 

	I have played with several and tried my own. I have found two
approaches. One is to simulate movie starfields and the other is to
simulate real starfields. 

	Movie type starfields are ones that have just enough more than from
earth but not like too many. Real starfields are printed by exposure
time. 

	Granite and other patterns requires lots of playing with the scaling
and still colors tend to group and look like patterns. 

	What I learned from all of this was to create a simple rule of thumb
while in progress of trying to create starfields in a manner emulating
reality. 

	So the governing rule is, regardless of the distance it must be large
enough to render even though in reality brightness is all that matters. 

	I am not ready to go beyond this rule as yet. I could fake it several
ways from Sunday with processing time penalties getting worse with every
"improvement." 

-- 
Torturing the data until it confesses is not considered the 
scientific method. 
	-- The Iron Webmaster, 289

Post a reply to this message

David Fontaine wrote:
> 
> "Bob H." wrote:
> 
> > That's an untruth actually.
> > Take a granite color map, for example, which disperses a star-like pattern with
parts (dots) being scaled too
> > small to be seen (when AA is used especially, not sure at the moment if this does
in all cases) and then try an
> > increase in color value.  I think the previously invisible pixels will then
appear.
> > You might be right though about it when non-AA rendering.
> > Same goes for objects too I think.  Seems I've encountered that before.
> 
> Correct, it only works with AA. Still, though, only to an extent and some still
disappear entirely.
> 
> If a star is "between the cracks" AA will completely miss it, because neighboring
pixels will be black. In the case
> of getting caught and aa'd, the aa will of course decrease brightness as the star is
only part of the pixel.
> Angular size is still the predominant factor.
> 
> It is very annoying to create good stars in POV when changing quality settings from
one render to another!

	That is exactly what I am trying to eliminate. The best solution so far
results in a 9.7M tga file at level 6 compression -- whatever GIMP means
by 6. However it is some five times faster to render than a random
generation of many fewer stars. However it is a poor solution in general
as it still requires odd things to be done. 

-- 
He who casts a vote decides nothing. 
He who counts the votes decides everything. 
Proprietary, secret code counts the votes in the US. 
	-- The Iron Webmaster, 210

Post a reply to this message