hey guys!

this nice little paper about water droplets on glass did really fascinate
me. I started right away yesterday evening (in spite of learning for a class
test *g*) and had half of the algorithm implented till I went to bed.
I rendered a small animation of it while I was at school this morning. You
can download it from

http://www.torfbold.com/droplets.avi (about 330K,100 Frames)
(its compressed using divx 5, you can download it from http://www.divx.com)

I'm afraid it is not water on glass but water on something non-transparent.
I didn't have the time to set up a nice demo-scene, suggestions welcome *g*.
For this animation, I used a grid of 120x120 cells. It took about one hour
to render it on my Athlon 600 256 MB SDRAM, half of the time was spent
parsing, the other half tracing. I think the algorithm could be made faster,
but I'm not a specialist in this field.
I'll release it after I coded the rest of the algorithm, so anybody can have
a try.

I didn't used the approach of using a mesh which was suggested by the paper,
but reduced it to a simple n*m grid. Due to this, it's only possible to
render droplets running down even plates like windows. These plates may have
nevertheless all possible inclination angles.

There is still much room for improvements, but this is as far as I came
yesterday:
- The basic algorithm for animations is implemented
- the masses of the droplets and their paths are already calculated
correctly

What is still to be done:
- calculating the correct size of each droplet according to its mass (at the
moment i'm using a simple linear connection)
- calculating the correct velocities of the droplets. I'm not sure about
this part of the paper as it is not clearly stated how these velocities are
used.

I think the result in the middle of the animation is pretty nice, but in the
end there are many non-moving masses around and you can see that they are
arranged in a grid. The blob-parameters i used were imho not that good, the
droplets lack "blobiness". I'm not very familiar with blobs, but I hope I'll
find some better values for them.
The behaviour of the droplets are dependent on some parameters. Perhaps
other values for these parametres will lead to nicer results, but i had not
yet time to try this.

Comments & ideas welcomed!

greetings:florian

Post a reply to this message