Refraction in objects

what i believe is usually done is just forget about the outgoing refraction.

Really? I think outgoing refraction can change a lot the look of the material. Its harder to see what is inside a swimming pool than behind a glass bottle.
Well, if anybody knows or has implemented that type of refraction, let me know.

Quote:

Original post by YengaMatiC Light refracts twice, so you dont know the distance that light refracted vector travells within the object.

I remember in physics exercises I compute that distance... maybe you mean to do that in real-time?

However, in a physic exercice (not solved), the book asks to dimostrate that when the light passes through a piece of glass with parallel plane surfaces the outgoing ray is parallel to the incoming one. It says that the distance between the two parallel ray is x=t*theta*(n-1)/n where t is the width of the glass, theta is the incident angle respect to the normal, n is the refraction index. This is true for little value of theta (so in general I imagine is sen(theta) instead of only theta). So the only thing you have to do is to move a little what is seen throught the glass (per pixel, so I think you can do that with a pixel shader, maybe applying directly the formulae above (with some optimization since t and n are constants), calculating theta for each pixel (or for 2x2 or 4x4 pixel or more, depending on the size of the glass surface and on the distance of the viewer).

I hope that helps a little.
[Note: I may be wrong, but that's the way I'd do it! [grin]]

I'm not exactly sure what you are looking for but this site may help you.

Quote:

Original post by YengaMatiC Really? I think outgoing refraction can change a lot the look of the material. Its harder to see what is inside a swimming pool than behind a glass bottle. Well, if anybody knows or has implemented that type of refraction, let me know.

true.

but since when is rasterizing about physical accuracy?

as long as your view though the object looks all warpy, youve got a nice refrection going on as far as immersiveness is concerned.

This is a really good question, and it has been considered before. This paper has some good results and describes exactly how you might do it: http://www.cs.uiowa.edu/~cwyman/publications/files/approxISRefract/UICS-TR-05-01.pdf

Wow! Thanks a lot, it looks like the way i thought it out. First rendering back facing geometry and storing the normals and depth, then rendering front facing geometry and getting distance modulated by the normal and a factor (not refraction factor).
Also it could be computed by not using snell's law, but an approximate method like adding the scaled normal to incident vector.
I'll look deeper in the paper and figure out how shadowing can be integrated with this algoritm (i suppose a third pass from lights pov).

Amazing though, thanks for the url :D

You're welcome. I found the paper because I had the same idea, and figured someone must have thought of it before. No doubt there are lots of cool things you could add, I would like to see the effect using the same distance trick to reflect into a real scene instead of just an environment.