A different style of reflection

[Jesse]

considered trying this ssdo procedure?

Unfortunately it is not that easy to implement in m3d, not only a little extension of ssao.  No time at the moment for it, btw i had already started to do something like this a while ago but i saw that it would be more effort i thought.

The color diffusion looks very subtle btw I don't notice a diffusion of the green?

red is (255,0,0) and green (0,128,0), and with internal gamma of 2 the reflected intensity of green is just 1/4... dunno if this explains it or maybe there is something not 100% exact in the coding.  Attached the pic with internal gamma of 1.

And, what if you realize a "stoppable" rendering? I mean, that you can halt anytime like in the realtime script

This would be common for such rendering methods, the longer you calculate the less noise you have.  But it needs a better interface, a new fileformat etc.  You might be surprised that i have also other things to do 

[DarkBeam]

It's more visible now thanks

Another ssdo paper; http://www.mpi-inf.mpg.de/~ritschel/Papers/SSDO.pdf

Also, found this http://dudka.cz/rrv/ that presents an interesting "fast radiosity"

[Alef]

In feature request maybe I would stand not for better shadows or reflections, but for more colouring methods and colour methods with parameters;) Yeah, and for british english, too.
(I think, I don't nead another thread.)

Exponent smoothing with amazingbox and mandelbulb works nicely, through with no more than first 15 iterations or it gets noizy. If you iterate exponent smoothing with z=cabs(z+ parameter) before summ=exp(-z) +summ, you can move colours around and make dotted mandelbulb. Well, single plain exponent smoothing, maybe with some iteration normalisation, would be well enought.
 
Well there are lots of 2D colour methods, probably one can use Chaos Pro or some Ultra Fractal .upr formulas without built in colouring to test, how and which work in 3D. This should be hudge field of future research for years to come.

Yeah, and with british english, becouse there mostly are Эuropeans;)

[DarkBeam]

Another possible solution; use "obscurance" method

http://ima.udg.edu/iiia/ggg/usersdocs/mateu/obscurances.pdf
http://www.graphicon.ru/2006/fe10_17_MendezSbert.pdf

The MC method seems to be the slowest of all

[Alef]

I think, renders are alredy very realistic, and I think, best improvement step would be couloring methods, maybe implemented like fractal formulas _{(so that user darkbeam could make lots of them)}, or maybe just more. That would add another dimension on the amazingbox.
Another exponents moothed bulb + trap center cabs(z+2). This is pretty simple.