"clipka" <nomail@nomail> wrote:

> I guess the mapped image is considered something in 3D space (which just
> happens to be identical along a certain axis), so first you have to get a slope
> out of the bump map; but that slope already needs to take into account the
> orientation of the plane that should be "bumped" if I'm not totally mistaken.
> Maybe the developer of the algorithm forgot about that.

I seem to remember, in the dim past, posting a problem that also concerned (or
*possibly* concerned) the default slope_map that is built into some/all pigment
patterns. Wish I could remember when, and what it was all about; I'll post a
link if I can find it.
>
> And nobody ever noticed, because such strange effects in bump mapping are
> visually hard to identify...

Absolutely!!

> The bump map usually doesn't look plain wrong,
> because our brain blames it on a different lighting direction. Which gives us
> some impression that "something is weird", but it's hard to pinpoint unless
> you design a picture specifically for that purpose.

Yes, it sometimes QUITE difficult to construct just the 'right' scene. I had to
do similar detective work in trying to nail down a radiosity-from-media
problem. (Hmm, I need to post that experiment...sometime soon...)

Unless my eyes are playing tricks on me (which is possible, of course!), I still
cannot get your red and green areas to look *wrong*, no matter what I try. Blue
and gray, yes; red and green, no.  Although I can't imagine how, I wonder if
there's a clue hidden somewhere in this? (BTW, I tried eliminating the scale
block from your blue texture; but that didn't help it.)

I even set up an animation (using monkeyjam,
a great little Windows app for doing quick tests.) Using a single 'orbiting'
point_light, and your seven cubes rotated into all sorts of
orientations, I never saw anything amiss with red and green. What's your
opinion?

Ken W.

Post a reply to this message

I've had a good long look but I don't see any bumpmap problems. The blue bump
does indeed appear to be the reverse of the others, but this looks like a
normals issue to me, as Tim mentioned. I see no 'sideways' distortion at all,
and all the coloured bumps appear to be consistent between the various copies.

Apart from the direction-reversing, this looks perfectly fine.

There is, however, a problem with your vertices - assuming you were trying for 8
squares per quarter-face, these vertices aren't correct. Some of the vertices in
the coloured squares are off-line, producing a chevron-like indent on many of
the quarter-faces. I see this clearly in the image too.

Post a reply to this message

"Kenneth" <kdw### [at] earthlinknet> wrote:
> Unless my eyes are playing tricks on me (which is possible, of course!), I still
> cannot get your red and green areas to look *wrong*, no matter what I try. Blue
> and gray, yes; red and green, no.  Although I can't imagine how, I wonder if
> there's a clue hidden somewhere in this? (BTW, I tried eliminating the scale
> block from your blue texture; but that didn't help it.)

That's really weird! Which version of POV are you using? Because I actually did
have ONE face which seemed right for every orientation as well, but TWO showed
problems with either the UV-mapped or "flat-mapped" version of the bump map, or
even both, in some orientation or the other.

I'll post a shot from my scene in the binary newsgroup soon (can't do it right
now because my monitor has just been fried, so I can't access my private
computer), along with the original bump map I used.

> What's your opinion?

Mysterious thing. That's my opinion :)

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:

> I've had a good long look but I don't see any bumpmap problems. The blue bump
> does indeed appear to be the reverse of the others, but this looks like a
> normals issue to me, as Tim mentioned. I see no 'sideways' distortion at all,
> and all the coloured bumps appear to be consistent between the various copies.
> Apart from the direction-reversing, this looks perfectly fine.

It gets weirder and weirder... I see only two instances of "outward" bumps in
the whole picture, and they're NOT consistent between copies, and I do see
"sideways" bumps (looking like they were illuminated from the side instead of
head-on) in exactly 50% of all bump-mapped dents.

Which POV version are you using?

I first tested for (and clearly saw) these bump map problems with some official
3.6.something (can't tell exactly which at the moment) for 32-bit Windows. I
also tested with the latest beta (.29) for 32-bit Windows, and with MegaPOV
1.2.1 binary for 32-bit Windows - all showed exactly the same strange effects.

I even think I recall that back in the mid-90's, I encountered similar problems
with the then-up-to-date version of POV.


> There is, however, a problem with your vertices - assuming you were trying for 8
> squares per quarter-face, these vertices aren't correct. Some of the vertices in
> the coloured squares are off-line, producing a chevron-like indent on many of
> the quarter-faces. I see this clearly in the image too.

Never mind the chevrons: I did a subdivision step when exporting the mesh to
POV, which introduced this distortion. I didn't bother about fixing that.

Post a reply to this message

"clipka" <nomail@nomail> wrote:
> It gets weirder and weirder... I see only two instances of "outward" bumps in
> the whole picture, and they're NOT consistent between copies, and I do see
> "sideways" bumps (looking like they were illuminated from the side instead of
> head-on) in exactly 50% of all bump-mapped dents.

I'm now having trouble seeing consistency again, my mind is playing tricks. I
think we're probably all seeing the same thing, but interpreting it differently
(there's a couple of Escher drawings that do this deliberately). I'm going to
play with the lighting... I'll get back to you.

> Which POV version are you using?

3.6.1c.icl8.win32 - from the splash screen. Far more info than you need, I'm
sure. :)

> Never mind the chevrons: I did a subdivision step when exporting the mesh to
> POV, which introduced this distortion. I didn't bother about fixing that.

Ah, right. Just thought I'd check.

Bill

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> 3.6.1c.icl8.win32 - from the splash screen. Far more info than you need, I'm
> sure. :)

I can't find any suitable expression for my puzzlement anymore: I can't check
right now, but that version string looks VERY familiar to me! I'm pretty sure I
had run the scene through EXACTLY the same version, with the same queer effects
as with the 3.7 beta .29.

And you have NO dents at all that look as if they were illuminated from the
side?

Did you use the same image format I did, or did you change that in the scene
code?

Can it be some INI setting??

Post a reply to this message

I am not sure about what you are refering at. The bumps look ok to me, 
except that there is an *illusion* of outside inside due to the lighting, 
but when you evaluate carefully you will see that the bumps are correct.

I loaded your scene into Poseray and noted that all normals are consistent, 
but facing inwards. However, I don't think this is a problem here.

Thomas

Post a reply to this message

"Thomas de Groot" <tDOTdegroot@interDOTnlANOTHERDOTnet> wrote:
> I am not sure about what you are refering at. The bumps look ok to me,
> except that there is an *illusion* of outside inside due to the lighting,
> but when you evaluate carefully you will see that the bumps are correct.

The bumps look *all* correct - as long as viewed by themselves; but they look
like the light is coming from different directions for the different bumps,
while they *should* look like they were all dented "inward", with the light
coming from the camera.

When I render it, some look as if they were dented "outward" instead, and some
even look as if the light was coming from sidewasy - but there is *no* light
source there: All the light sources are near the camera.

In this shot, I marked the bad dents (the scene is slightly different, using
colored little balls instead of colored faces to mark the "X", "Y" and "Z"
sides, but the basics of the scene are the same, and the cube is the same mesh
as well):

http://www.postimage.org/gxNhlii-2ff61a18c010d6fe99b9e86819bcfe62.jpg

Post a reply to this message

"clipka" <nomail@nomail> schreef in bericht 
news:web.4936b4a7f7160dffbdc576310@news.povray.org...
>
> In this shot, I marked the bad dents (the scene is slightly different, 
> using
> colored little balls instead of colored faces to mark the "X", "Y" and "Z"
> sides, but the basics of the scene are the same, and the cube is the same 
> mesh
> as well):
>
> http://www.postimage.org/gxNhlii-2ff61a18c010d6fe99b9e86819bcfe62.jpg
>

Hmm, yes. I see what you mean. Very strange. I shall have to experiment some 
more I am afraid.

Thomas

Post a reply to this message

"clipka" <nomail@nomail> wrote:
> "Kenneth" <kdw### [at] earthlinknet> wrote:
> > Unless my eyes are playing tricks on me (which is possible, of course!), I still
> > cannot get your red and green areas to look *wrong*, no matter what I try. Blue
> > and gray, yes; red and green, no.
>
> That's really weird! Which version of POV are you using?

32-bit 3.6.1c on 32-bit Windows, but running on a 64-bit AMD machine. (The
pvengine is v3.6.1.3)

I'll try another animation, with a different light orbital plane, and see if
that produces anything on the green and red ones.

BTW, I'm definitely seeing that odd 'light angle misdirection' on the blue and
gray areas--and it's 'consistent' in my animation (i.e., traveling around with
the orbiting light as it should.)

KW

Post a reply to this message