TITLE: hideaway
NAME: Norbert Kern
COUNTRY: Germany
EMAIL: norbert-werner.kern@t-online.de
WEBPAGE: not yet
TOPIC: Light and Fog
COPYRIGHT: I SUBMIT TO THE STANDARD RAYTRACING COMPETITION COPYRIGHT.
JPGFILE: hideaway.jpg
ZIPFILE: hideaway.zip
RENDERER USED: 

    Megapov 1.2.1


TOOLS USED: 

    Xfrog, Poseray, Photoshop


RENDER TIME: 

    trace about 4 weeks (4096*2304 AA 0.3)  peak memory 1.2 GB 


HARDWARE USED: 

    3.06 GHz P4 / 2 GB RAM



IMAGE DESCRIPTION: 

A sunlit, secret glade with dense, colorful vegetation and a bright, 
misty atmosphere - my tranquil vision of paradise and my pet subject. 



DESCRIPTION OF HOW THIS IMAGE WAS CREATED: 

From the beginning I had some ideas for the 'Light and Fog' round, 
but only one of them succeded. Main issue were insufficient media 
effects. But after the trials phase s bright sunlit scene with a 
slightly visible background succeeded. It consisted of an empty 
space for a glade and many nearly invisible simple trees behind the 
camera, at the sides and far away in the front for shadows and 
suitable radiosity. At that time it seemed clear, that the final 
render would take weeks for finishing, so I reserved less than two 
weeks to create and optimize the foreground. 
Currently I'm working on a tool for realistic distribution of all 
sort of vegetation, but this will take more time than I had. Then I 
remembered an older image, which fascinated me since I've seen it 
for the first time - 'Stream' from the creators of the plant modeler 
Xfrog (http://web.inf.tu-dresden.de/ST2/cg/gallery/pdfiles.html). 
It wasn't possible to convert the Rayshade scene file to something 
useful, so I extracted camera/light positions and plant coordinates 
for 4 appletrees, 140 bushes, 2550 grasses, 440 nettles, 1560 yellow 
flowers and 60 dandelions.

First I formed a ground object by creating a mesh out of these 4750 
plant positions. I failed to implement a marching cube algorithm, 
but the freeware tool Points2Polys did the job (http://www.paraform.
com/ppdl/index.html).
Since it is impossible to simulate a dense vegetation with such a 
low number of single plants, I put spheres at these coordinates and 
used the trace command to generate more coordinates. Every time when 
a trace on these spheres succeeded, a second trace on the ground 
mesh was done to get the final plant position. By example 175000 
grass positions were generated with this method based on the given 
2550 coordinates. 

It took a long time to find the right combination of plants - in the 
end 32 different plant meshes were used, thereby deliberately 
disregarding botanical accuracy when it seemed appropriate. 
All visible trees, shrubs and plants were modeled with Xfrog 
(http://www.xfrogdownloads.com/). 
The phragmites were created with Xfrog by Gilles Tran (http://www.
oyonale.com/ressources/english/sources15.htm).

To fill some 'holes' clovers and buttercups were planted with my 
own methods from the camera point of view (http://news.povray.org/
povray.binaries.images/thread/%3C43236855@news.povray.org%3E/?ttop=
218917&toff=50). 
Two different ways were used - first to use a mask with white parts 
from ground and black parts from plants and then to 'plant' only at 
the uncovered regions. The second way is to use two traces from the 
camera point - first trace on ground only, second trace on a union 
of ground and vegetation. If the two traces delivered the same 
result, a fill plant was set at this point.

For some additional objects like the roses, shrubs and some 
background trees I wrote a 'position finder' macro which replaced 
the tedious way of trial and error - see http://news.povray.org/
povray.binaries.images/thread/%3Cweb.44784f2141a06b1c9bff7e8b0%40
news.povray.org%3E/?ttop=229646&toff=50 .

First I rendered the scene with 4096*2304 pixels. A second povray 
render was done for some contrast/color correction and resizing.
The pov file included in the zip is renderable, but uses cylinders 
instead of the meshes. They aren't included because of their size.

