Mike Williams wrote:
> What happens is that not all of the 2d UV-map is used.
> The front of the body is mapped to part of the texture
> space that's shaped like the front of a body, and the
> back of the body is mapped to a separate area. It's
> quite common for the body to be split into several
> pieces, with parts like eyes, teeth, tongue and hair
> mapping to disjoint regions of the map.

And also the back and the front of the body I assume? And here, it is
not just the map that is split up, but the actual mesh object is split
into several meshes right? Because that would be the only way to make
those "jumps" in the UV-coordinates as far as I understand.

> The usual method is to attempt to blend the textures
> near the seams to something rather plain.

Ah, indeed I couldn't imagine what the alternative to that would be. It
sounds like a rather big limitation though, in cases where you want a
detailed and non-plain texture covering the whole object.

What I had been wondering was if there were smarter techniques invented.
For example I could imagine a feature to layer textures on top of each
other, each texture with a independent set of UV coordinates. That way,
the seams for the different texture layers could be at different places,
and the seams at the lower layer(s) could be covered by the upper
layer(s). But I guess an approach like that would be too much trouble in
most cases, and require too much data (the separate sets of UV
coordinates).

When I get time I'll try experimenting with UV-mapper to get a feel for
the things myself. For now, thanks for the explanations to all who
replied.

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message

3d73d492@news.povray.org...
> And also the back and the front of the body I assume? And here, it is
> not just the map that is split up, but the actual mesh object is split
> into several meshes right? Because that would be the only way to make
> those "jumps" in the UV-coordinates as far as I understand.

Not always. For example, the vertices for the "eye" part may refer to
another map that the vertices for the "lips" part, while sharing the same uv
plane.

> Ah, indeed I couldn't imagine what the alternative to that would be. It
> sounds like a rather big limitation though, in cases where you want a
> detailed and non-plain texture covering the whole object.

From a practical point of view it's not such a limitation. It is for
textures that have some regular pattern if you want the pattern to be 100%
continous over the shape (such a checkerboard pattern). Fortunately, for all
practical purposes, those sort of patterns are rarely used. Natural patterns
like skin have a lot of randomness that make seams quite invisible. Man-made
objects are even less of a problem. Personally, I find the stretching /
pinching artefacts more problematic and actually quite visible, even on
production-quality 3D.


G.


--
**********************
http://www.oyonale.com
**********************
- Graphic experiments
- POV-Ray and Poser computer images
- Posters

Post a reply to this message

Wasn't it Rune who wrote:
>Mike Williams wrote:
>> What happens is that not all of the 2d UV-map is used.
>> The front of the body is mapped to part of the texture
>> space that's shaped like the front of a body, and the
>> back of the body is mapped to a separate area. It's
>> quite common for the body to be split into several
>> pieces, with parts like eyes, teeth, tongue and hair
>> mapping to disjoint regions of the map.
>
>And also the back and the front of the body I assume? And here, it is
>not just the map that is split up, but the actual mesh object is split
>into several meshes right? Because that would be the only way to make
>those "jumps" in the UV-coordinates as far as I understand.

It's only the maps that are split. The entire model (front, back, eyes,
hair, teeth and tongue) is a single mesh2 object (or single Wavefront
OBJ file in UVmapper). 

In POV, the uv_indices feature is used to manage the "jumps" - the
situation where one vertex maps to more than one position on the UV map.

Wavefront OBJ files don't have the equivalent of uv_indices. I think
they just declare some vertices more than once, because each vertex
declaration can only correspond to one texture locator.

>> The usual method is to attempt to blend the textures
>> near the seams to something rather plain.
>
>Ah, indeed I couldn't imagine what the alternative to that would be. It
>sounds like a rather big limitation though, in cases where you want a
>detailed and non-plain texture covering the whole object.

The alternative, in such cases, is lots of hard work. This leads to
situations like that of the wonderful new DAZ dragon, where the model is
free, but a texture map for the model costs $12.97.

One trick that's being developed by some members of the Poser community
at the moment is to arrange for the seams to be located in places where
they tend to be out of sight much of the time. The existing Poser models
have two particular seams that tend to be rather noticeable, one around
the neck and one down the outside of the arms. The "Free Market Family"
models are being designed so that the arms are mapped cylindrically and
only have a seam down the inside.

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message

On Mon, 2 Sep 2002 17:32:49 +0200, "Rune"
<run### [at] mobilixnetdk> wrote:

>Never having used modelers or imported UV-mapped meshes, I don't know
>much about how UV-mapping works for objects of complex topology.

Who said that all of the UV map has to be used? :)

Remember the free fish model site Gilles posted a while back? Download
one of the models and look at the .bmp - a picture is worth a thousand
words.


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

Mike Williams wrote:
> It's only the maps that are split. The entire model
> (front, back, eyes, hair, teeth and tongue) is a
> single mesh2 object (or single Wavefront OBJ file
> in UVmapper).
>
> In POV, the uv_indices feature is used to manage
> the "jumps" - the situation where one vertex maps
> to more than one position on the UV map.

Ah, of course. Always having used one UV vertex per point vertex in my
own work, I overlooked this this possibillity...

> Wavefront OBJ files don't have the equivalent of
> uv_indices. I think they just declare some
> vertices more than once, because each vertex
> declaration can only correspond to one texture
> locator.

Makes sense.

> The alternative, in such cases, is lots of hard work.
> This leads to situations like that of the wonderful
> new DAZ dragon, where the model is free, but a
> texture map for the model costs $12.97.

Oh, ok. But surely, they must have some clever ways to come around the
problems? I could still imagine a tool that would let you work with
several overlapping texture layers with different UV coordinates, but
which would put it into a single (seamless) layer in the final step of
production. Still hard work of course.

> One trick that's being developed by some members of
> the Poser community at the moment is to arrange for
> the seams to be located in places where they tend
> to be out of sight much of the time.

:)  I guess UV mapping is indeed an advanced skill to master...

Now to figure out how to implement advanced UV mapping in POV-Ray for
shapes with complicated topology made of several different types of
bicubic patches. I do have a few ideas actually, but they will be
difficult to carry out...

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message

Gilles Tran wrote:
> Not always. For example, the vertices for the "eye"
> part may refer to another map that the vertices for
> the "lips" part, while sharing the same uv plane.

Yes, I understand that now... :)

> Fortunately, for all practical purposes, those
> sort of patterns are rarely used. Natural patterns
> like skin have a lot of randomness that make seams
> quite invisible.

I was more thinking of clear lines, such as lines in clothes, and lines
separating clothes texture and skin texture. Making those match up
completely continuously would seem difficult to me.

> Man-made objects are even less of a problem.
> Personally, I find the stretching / pinching
> artefacts more problematic and actually quite
> visible, even on production-quality 3D.

I can imagine. One thing I can't understand is the limitation in POV-Ray
that UV vectors have to be UV (2d) vectors instead of 3d vectors. The
textures are 3d anyway, even image_maps, and one could easily take
advantage of 3d procedural textures if 3d coordinates were allowed in UV
maps. (I know that UV means 2d, but what things are *called* isn't
really the issue here. I see no practical reason for the limitation.)

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message

Peter Popov wrote:
> Who said that all of the UV map has to be used? :)

I don't know. Who said that?

I think UV-mapping sounds like it involves many problems as it is.
Having to use the entire rectangular area of the map would simply be
insane.

> Remember the free fish model site Gilles posted
> a while back? Download one of the models and
> look at the .bmp - a picture is worth a thousand
> words.

Thanks, I think I'll do that when I get time. :)

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message

3d74f4c1@news.povray.org...
> I was more thinking of clear lines, such as lines in clothes, and lines
> separating clothes texture and skin texture. Making those match up
> completely continuously would seem difficult to me.

Well it's easy, since clothes are by definition made of 2D flat material cut
and folded at the right places... uv maps for clothes look just like regular
cloth patterns.
In hi-res models, cloths and skins are different objects, so there's no
problem here either. Really, uv mapping is not such a complex problem (to be
frank I didn't believe it until I got my hands on obj2pov in 1998).

> I can imagine. One thing I can't understand is the limitation in POV-Ray
> that UV vectors have to be UV (2d) vectors instead of 3d vectors. The
> textures are 3d anyway, even image_maps, and one could easily take
> advantage of 3d procedural textures if 3d coordinates were allowed in UV
> maps. (I know that UV means 2d, but what things are *called* isn't
> really the issue here. I see no practical reason for the limitation.)

The fact is that UV mapping is a surprisingly robust and efficient
technology. The mesh2 format is just an implementation of it that works fine
and is 100% consistent with industry formats like OBJ or 3DS.
Regular procedural pigments works fine with too uv mapping. They're just
taken in the <0,0,0>, <1,1,0> space and are very easy to control. Some
normal{} statements, for instance, are typically difficult to set up on
classic CSG (quilted normal on a curved object anyone ?) but are totally
natural to uvmapping.  Just try it :-)

Btw, what effects do you think that could be done only with 3d UV mapping
(just curious) ?

G.

--

**********************
http://www.oyonale.com
**********************
- Graphic experiments
- POV-Ray and Poser computer images
- Posters

Post a reply to this message

Gilles Tran wrote:
> Well it's easy, since clothes are by definition
> made of 2D flat material cut and folded at the
> right places...

Hmm. Maybe so...

> The fact is that UV mapping is a surprisingly
> robust and efficient technology. The mesh2 format
> is just an implementation of it that works fine
> and is 100% consistent with industry formats like
> OBJ or 3DS.

I fully agree. I just don't see the reason for the limitation when it
has some disadvantages and no advantages. Specifying 3d vectors would be
optional, as 2d vectors would be perceived as 3d vectors with the third
component being 0.

> Regular procedural pigments works fine with too
> uv mapping.

Yes, but is confined to the x-y plane.

> (quilted normal on a curved object anyone ?)
> but are totally natural to uvmapping.  Just try it :-)

Hey, I'm not talking about a replacement! Regular UV mapping is indeed
very useful. :)

> Btw, what effects do you think that could be done
> only with 3d UV mapping (just curious) ?

Well, here's a bad but illustrative example. Say you have a statue made
of marble. A 3d solid procedural texture is perfect for this purpose
(because in reality marble is solid) and much easier to use than being
forced to use a UV map on the x-y plane (because in reality the marble
pattern isn't wrapped around the object). Then suppose you want the
statue to come to life and move and of course still have the texture
stick to the surface. You need UV mapping for that, but it would be nice
if you could just use 3d coordinates for the UV map so you wouldn't have
to find a way to wrap the x-y plane around the object and make it look
solid.

Actually, make it a wooden figure that comes to life. Isn't it very hard
to "simulate" 3d solid texturing when using 2d UV maps?

Basically, think of it as 3d solid mapping that stick to the surface
when it moves. Most useful for animations of course.

There are also other uses, but this was the first that came to my mind.

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message

3d751704@news.povray.org...
> Basically, think of it as 3d solid mapping that stick to the surface
> when it moves. Most useful for animations of course.

OK I understand better what you mean.
I suppose that for the usual mesh-based modellers and renderers, it's not
really a problem of uv-mapping, but a question of defining the right map by
plotting the right pigment at the right place. I guess that  the various 3D
painting software have a feature that lets the user define a 3D procedural
texture, run it through a 3D object and then create the map accordingly.
Perhaps that's the way it's solved usually, or perhaps it's considered as a
minor issue to be solved by usual 2D mapping techniques... I don't know so I
can't elaborate on this.
Now in the case of procedural mapping done in POV-Ray, such 3D uv coordinate
system would make actually sense for animations like your animated statue
scenario, once a POV-Ray mesh animation feature exists that can take full
advantage of it, that is... Seems a very long shot to me, but well, why not.


G.

Post a reply to this message