|
 |
Alain <ele### [at] netscape net> wrote:
>
> The artefacts seems to be related to the max_gradient, BUT, they are not holes
> in the isosurface. Tested with interior_texture{pigment rgb y}finish{ambient 1}}
> and there are no green pixels.
>
I agree.
Those pesky bright spots have continued to bug me (even though they have been
considerably lessened), so I've been running test after test to try and
pinpoint the cause. I hate mysteries! In the process, I've come across
something very odd, concerning media, the "water box" and some of the
isosurface parameters.
After determining (through tests) that I could "strip out" some non-essentials
from the scene, with no ill affect, I decided to run it using some very radical
values, on purpose, to *clearly* show the oddities. I also substituted a simple
small green sphere for the water box, to eliminate
extraneous variables. I've included the test images, and my stripped-down code.
No radiosity or photons in this experiment, and no surface normals.
Some of the "radical" values I used are an isosurface max_gradient of 7(!) and
an accuracy of .00001. I also added an interior_texture to the isosurface, per
Alain's suggestion. All of this is to clearly show any isosurface holes. I
*want* them there, to distinguish them from the bright spots. (You would
think--or at least I would-- that a finer iso accuracy value would actually
clean up the bright spots. Not so; it makes them worse!) I also set the media
samples value to 1(!), because samples affect these results only minimally, if
at all, and it's much faster to render. I originally had it at 100, and the
render tests were taking 1 hour each. Ugh.
BTW, I took a look at the scene with a distant camera, and saw that the media
sphere size was *huge* in comparison to the relatively tiny isosurface box.
Seems kind of wasteful. ;-) So I reduced this sphere size to
6(!)--it still covers the isosurface well--and translated it in +z, to put the
sphere surface very slightly in front of the main camera--so as to put the
camera outside of it--thinking that the bright spots could somehow be an
artifact of the camera originally being inside the media. But that didn't
help.
#1 is the isosurface by itself, no media or sphere object. The holes in the
isosurface are very apparent. But no bright spots.
#2 is with the sphere added. Still no bright spots.
#3 is with media but NO sphere. Still OK.
#4 is with media AND the sphere. Suddenly, there are LOTS of anomolies-- simply
from the addition of the sphere into the scene! Not isosurface holes, either.
That makes no sense--there is no CSG with the isosurface (which would have been
my first thought.)
#5 is that scene but with isosurface accuracy changed to .001 (the default.)
The anomolies disappear! (Well, there are now a *few* new isosurface holes,
nothing terrible.)
#6 is with iso accuracy back to .00001, but max_trace 2 added. Again, the
anomolies disappear. I find that very strange, because again, there is no CSG
in the scene, and AFAIK max_trace and all_intersections are meant for CSG. Yet
it works.
#7 is with iso accuracy .00001, but with all_intersections taking the place of
max_trace. Again, no anomolies.
So in essence, adding the "water box" to the scene--while using media--seems to
be the root cause of the bright spots...coupled with isosurface parameters! I
don't know what to make of all that, but luckily there are solutions. Is there
an isosurface/CSG bug somewhere?
I hope this proves useful. Feel free to keep experimenting with the original
code (and to prove me wrong); there are so many permutations possible that I
could have missed something.
Ken W.
-------simplified code----
#declare IsoOk = 1;
#declare MediaOk = 0;
global_settings {
assumed_gamma 1.0
max_trace_level 100
}
sky_sphere {
pigment {
gradient y
color_map {
[0.0 rgb <0.6,0.7,1.0>]
[1.0 rgb <0.2,0.3,0.8>]
}
warp {repeat y flip y}
}
}
light_source {
<0, 0, 0>
color rgb <1.5, 1.2, .6>*4
translate <0, 20, 10>*1000
}
#if (MediaOk = on)
sphere {0,6
hollow
translate 4.3*z
pigment {rgbt 1}
interior {
media {
scattering { 1, rgb 0.015 }
method 3
intervals 1
samples 1
density {
planar
scale 10
translate -5*y
turbulence .2
}
}
}
photons {collect off pass_through}
}
#end
#default {
texture { pigment {color rgb 1} finish{ambient 0.0} }
}
#declare MountPig =
pigment {
cells
turbulence <1,.1,1>
scale .24
pigment_map{
[0.0 rgb <.6,.3,.2>]
[1.0 rgb <.5,.2,.1>]
}
scale <100,1,100>*3
}
#declare MountainTex =
texture {pigment {MountPig}}
#declare Cany =
function {
pigment {
gradient x
turbulence <.5,.0,.5>
//lambda 0
color_map {
[0.0 rgb .59 ]
[0.29 rgb .59 ]
[0.3 rgb .51 ]
[0.4 rgb .1 ]
[0.5 rgb .05 ]
[0.6 rgb .1 ]
[0.7 rgb .55 ]
[0.71 rgb .59 ]
[1.0 rgb .59 ]
}
}
}
#declare Terrain =
isosurface {
function { y - Cany(x,y*2,z).red*2 }
contained_by { box { 0, <1.3,1,6> } }
accuracy 0.00001 // or .001
max_gradient 7
// max_trace 2
// all_intersections // alternative to 'max_trace'
translate <-.5,-1,-.5>
scale 4
texture {MountainTex scale .1}
interior_texture{pigment{color rgb <0,1,1>} finish{ambient 1 diffuse 0}}
}
Terrain
#declare Obj = Terrain;
#declare Norm = <0,0,0>;
#declare Start = < -0.3, 1000.5, -1.8>;
#declare CPos = trace (
Obj, // object to test
Start, // starting point
-y, // direction
Norm ); // normal
#declare Start2 = <1, 1000.5, 3>;
#declare CPos2 = trace (
Obj, // object to test
Start2, // starting point
-y, // direction
Norm ); // normal
camera {
location CPos + <0,.3,0>
direction 1.25*z
look_at CPos2 + <0,-.5,0>
right x*image_width/image_height
}
// green sphere
sphere{0,.05
texture{
pigment {rgb <0,1,0>}
finish{ambient .5 diffuse 0}
}
translate <.2,-2,.3>
}
Post a reply to this message
Attachments:
Download 'newsgroup_radical_exp.jpg' (353 KB)
Preview of image 'newsgroup_radical_exp.jpg'
|
|
