Volumetrics - fast real-time volume rendering

[xadhoom]

So what is your conclusion of the results? I´m very interested, as you can see...

BTW: I hope spookyboo follows this thread, too. It would nicely fit as a ParticleUniverse particle type.

xad

[Oogst]

I have not had time for a serious analysis of all the results yet, but the conclusion from the results on the main testing machine (the one furthest on the left in the test results) is:

-That volumes are slower than standard particles, but fast enough for usage.
-Doing volumes through slices is hardly ever worthwhile, as either the number of slices is too low for good quality, or the framerate is too low.
-Of the different ways to implement soft particles, using the alplha texture for the thickness of the volume is best: an artist can capture the complexity of the cloud in the 3D shape of the volume, and the framerate is almost exactly the same as doing constant thickness (standard soft particles, with which the volume is a cube).
-FInally, if you already had the depth map for something else (depth of field, screen space ambient occlusion, deferred shading, etc.), then such volumes are pretty cheap and you should really use them.

I will try to put the full chapter online on Sunday, if I can get it ready by then. The images in it really explain this stuff much more clearly.

[Oogst]

I finally got all the results from this testing in. In fact, I finished the complete draft of my Master's thesis. The part on the volumetrics is chapter 6:

Using raycasting in GPU shaders to enhance real-time rasterised graphics.pdf

If anyone happens to read anything from this and you happen to see typing/grammar errors, omisions or things that are just not true, then please let me know, since right now I can still fix them before I hand in the final version! Thanks!

Two things have not been done yet in this version:
-image placement needs to be fixed for better page layout (lots of white space right now)
-references still need a logical ordering (normal numbers and in alphabetical order)

[xadhoom]

Hi Oogst!

I´m happy to see your thesis evolve.

I found a small bracket mistake on page 67: "at (ceil(y / d) - 1) · d."

By taking a glance over the pages it looks very promising...

xad

[Oogst]

I've been looking at this for five minutes now, but I really don't see it. What's wrong with the brackets?

[xadhoom]

Ouch! Sorry Oogst! The brackets are correct it just looked as if there was one too much because of the indention.
Maybe you could use a better font for formulas? Especially the "5" looks shifted downwards. An "0" looks a bit like an "o".

Sorry, again and good luck!

xad

[Oogst]

Yeay, I had accidentally removed the courier font for formulas, which I noticed when you mentioned the brackets. So you did find a mistake nevertheless!

[Crazy_Yellow_Rabbit]

your work is really impressive.

[tdev]

has anyone ported this great idea to the default ogre particle system?

[Oogst]

Not that I know of. I think the only thing that would have to be added to make that work, is that the particles need to store the position of the particle and its size in the vertex data.

[mike_cyber]

Trying to compile that on Eclipse/Linux and get:

../Renderer.h:45:19: error: ‘Ogre::PixelFormat’ has not been declared
make: *** [Material.o] Error 1

Many thanks for help
Michael

[Oogst]

The source shared in this topic is from 2008, are you compiling it with an Ogre version from 2008? Ogre has changed a bit in the four years that passed since.

[mike_cyber]

True If I want volumetric fog ie small fog clouds hanging around hills where to look for a tutorial etc.?
Thanks

[Oogst]

I don't know where you'll find a tutorial for that, but something like the technique in this topic would work fine for that. There are many other alternatives as well, depending on how much time you want to spent, what quality and style you need, and how technically difficult you are willing/capable to do. I don't think there is one best practice that everyone uses for this topic.