Am 10/17/2017 um 20:33 schrieb Bald Eagle:
> I'm always looking to learn how to increase the quality of my images, and reduce
> the amount of work it takes to create a scene (thereby freeing up more time to
> fine-tune it).
> 
> Can you give us a bit of background on the creation of this scene?
> 
Sure. But keep in mind that all I'm doing might very well not be the 
best, easiest or simplest way.


> How long did it take you to model it?
15 years to evolve ;)
But seriously this is hard to say - every few years I did go back to 
this scene and added, changed or improved on a few details.
The pitcher with the cap and handle is a good example. I do not remember 
how long it took to create the original csg/lathe object, but as I 
wanted to smash it, it had to be redone as a mesh. Modeling in Blender 
was about half a hour but finding out how to break it and automatically 
arrange the shards on the floor lasted a whole afternoon. There are 
various ways, but I used the tool "Object:Cell Fracture" that ships with 
blender but needs to be enabled within the user preferences window - 
it's fun!
Or the lion-head-chairs: for the cat shots I did improve the fabric a 
bit by using back-scattering (this was not available the last time I've 
done anything with the scene) and making the fabric slightly translucent.
As meanwhile, to me, this scene is no longer about recreating the 
original painting, more about doing silly things, I also changed my 
original lion heads (friendly looking with a closed mouth) - close to 
the ones in Vermeer's paintings - to some more wild looking animals with 
open mouth where even the teeth can be seen.
Just for the cat shots of course, in the future I'll use the friendly 
lions again ;)


> How many tools other than POV-Ray did you use?
In no particular order:

Lightroom - creating 16bps Adobe RGB texture maps from digital raw photos.
Photoshop - editing texture maps; creating bump, transparency and 
specularity  maps; also maps to drive the fur-mesh generation.
Wings3D - mesh modeling.
Blender - mesh modeling, breaking the pitcher.
Poser*  - cloth room for creating the draping of the table rug.
Daz Studio - rigging and posing of the cat.
PoseRay - mesh conversion.
MS Visual Studio - to write simple tools to help scene development. The 
fur-mesh could not have be done in SDL because it needs access to the 
cat-mesh uv-coordinates for any given point at the surface.

*I don't use Poser anymore and would meanwhile do this with Blender.

> 
> What sorts of simplifying "tricks" did you use (if any), 

Err, well, I tend to make things more complicated than needed ;)
E.g. the pitcher, the pearls and the grapes are all slightly transparent 
and use scattering media for a SSS effect.
I used this method before for some Chinese porcelain (before SSLT was 
even introduced to POV-Ray) but while this worked great for the pearls, 
for the grapes it should be just a little more intense and for the 
pitcher, well, I'm pretty sure I would have gotten away without using it 
there at all - and would have saved a few hours of render time for each 
image where it is seen up close.


> and what interesting
> methods or algorithms did you employ (you already touched on the map for the
> hairs in the cat's fur)
> 
While I was about to write "nothing special" there is one thing that is 
worth mentioning (and I've almost forgotten after 15 years).

Here is the very first incarnation:

http://news.povray.org/povray.binaries.images/thread/%3C3d7e31fa%40news.povray.org%3E/

where I did not even try to recreate the painting but the old Cornell 
University CGI copy of the painting that was the cover art of a book 
that itself was considered 25 years ago the "bible of graphics 
programming", I still have it ;)

To recreate the painting "The Music Lesson" for a raytracer, the 
dimensions of the room are needed. Or, to be more precise, the 
proportions (the dimensions in actual units, let's say cm, are less 
important and can be established later by simply assuming the teacher 
has the size of e.g. 170 cm).
Assuming that Vermeer did strictly follow the (since the Renaissance era 
well established*) rules for perspective painting, I did a few 
orthographic construction drawings (from the right and the top to 
establish the view point, vanishing point, horizon line and such) and so 
could determine at what angle the mirror above the virginals hangs from 
the wall.
Within the mirror is also the edge visible where the rear wall meets the 
floor - so the distance from the front to the rear wall can be 
calculated. Now, the interesting thing is, assuming the window wall is 
symmetrical, this matches exactly with the size for the window wall *if* 
there is a third window and *if* all three windows have the same size 
and *if* the wall size between the center window and the left/right 
window is identical.
A lot of assumptions and ifs (and still no way to get the position for 
the wall opposite the windows besides - by having the proportions so far 
- assuming a room with nice overall proportions) - but I was quite 
fascinated by these results so far.
At this time I was not aware of the book "Vermeer's Camera" by Philip 
Steadman (maybe it wasn't even published), but later after having posted 
an early version here in this newsgroup someone (sorry, don't remember 
who it was) did mention it and I did immediately buy the book.
Philip Steadman did the same calculations I did and also did came to the 
same conclusions regarding the dimensions of the room**. My image 
"Vermeer's Room" (posted a few days ago) shows the result of my "reverse 
engineering" process based on the painting.

And at my homepage are a few of my older renderings:

http://www.lilysoft.org/CGI/Music%20Lesson/musiclesson.htm



former home town) did publish a few books on this topic - himself being 
mostly influenced by Italian painters.

**but I completely disagree with the arguments he makes for Vermeer's 
usage of a camera obscura. These are all just cherry-picking or - in 
other words - a good example for bad science.


> Any special comments on lighting?
> 
Just GI, i.e. radiosity in POV-Ray terms. As a result, no material in 
the scene uses specular or phong highlights but blurred reflections, 
even the carpet (see below) - together with the ultra high radiosity 
settings the main reason for the long render times.
And I do not post the radiosity settings on purpose, they are not 
helpful because even with a powerful workstation every render took up to 
45 hours and with a common high end i7 class machine it would take weeks.


> The carpet is beautiful - how did you texture and drape it?
> 
My wife has some interest in old carpets (not only Persian) and created 
over the years a huge collection of high resolution photographs.
As she (like me) shoots always in RAW format, these are perfect for 
being used as image maps. I searched her collection for a rug that comes 
close to the one seen in Vermeer's painting - regarding color tint and 
ornamentation. Parts of the coloring are selective changed later in 
Photoshop to get even closer.
Draping was done in the Poser cloth room, I think it was Poser 6 at the 
time, I don't have Poser installed anymore.
The rug is also reflective using *very* blurred reflections and I'm sure
anisotropic, blurred reflections would make it even better but this is 
one of the feature that I greatly miss when using POV-Ray as render engine.

> 
> Thanks  :)
>
You're welcome. And feel free to ask if you have any more questions. I 
always wanted to add some scenes or at least some furniture objects to 
my homepage for download, but - lazy as I am - never did it. Maybe some 
day...

-Ive

Post a reply to this message