Welcome to Fractal Forums

I know, this is an hard thing to do, and it's really too much to ask such things for a lovely free software... :embarrass:

There are different reflection styles, the one built in MB is specular and smooth, but also exists a "fuzzy" one - see this image

http://fractalyst.deviantart.com/art/Ornamental-Orb-287200951

The material on the back reflects smoothly near objects and blurs nicely faraway ones. Probably too hard to do... :-\ but lovely effect

Also I noticed that hard shadows always have "hard edges", a softening would be also nice like this

(http://upload.wikimedia.org/wikipedia/commons/5/55/Radiosity_Comparison.jpg)

And no there is no hurry to do this - and feel free to send me to hell for asking :embarrass:

as you know, i'm no programmer, but i have some expierience with several 3d rendering engines (cinema 4d, pov and some low level renderers built in cad software).
i am courious, if this is possible in a program, that does not provide a real 3d mesh for rendering.
the shadows in your image sample are not only softened (which should be easy), they are area shadows. the ammount of blur is related to the distance between shadow caster and shadow reciever (easy to see when you follow the shadow of the pillar in the middle of the room).
in 3d mesh renderer it increases rendering time moderate, according to the precision used. the softening reflection is also related to distances, so should be similar.

the other effect in your sample - radiosity - is far more expensive. afaik there are several methods of creating radiosity (sample rays, stochastic modes etc.), but in all cases it increases the rendering time dramatically. i guess this won't be very different in 3d fractal renderers, if possible at all.

when someone manages to implement those features, i'll be the first to be thankful trying the possibilities - even if it takes weeks to finish a rendering :beer:
after area shadows and radiosity, there's not much left to create ultra realistic images. maybe lighting with hdri maps... (just a joke)

If I'm not mistaken, syntopia is already working on it.
regarding global illumination, lycium did an hyper super realistic rendering of power 2 mandelbulb (http://www.fractalforums.com/index.php?action=gallery;sa=view;id=160).
I also remember a wonderful mandelbox render (http://www.fractalforums.com/3d-fractal-generation/a-mandelbox-distance-estimate-formula/msg17078/#msg17078) by Hamilton.

Yes, Fragmentarium (in recent GitHub versions) supports both soft shadows via (circular) area lights, and Image Based Lighting using HDRi's (using panoramic .hdr RGBE images for specular, diffuse, and background). There is also simulated Depth-of-field and HDR tone mapping.

For some test renders see, e.g.:
http://www.flickr.com/photos/syntopia/7270132662/in/photostream
http://www.flickr.com/photos/syntopia/7376230282/in/photostream/
http://www.flickr.com/photos/syntopia/6919800648/in/photostream

Wrt speed: Image Based lighting is fast and doesn't slow rendering much (well, not unless you want to calculate shadows based on the HDRi images, in which case rendering will be *much* slower).

Soft shadows and (simulated) Depth-of-field on the other hand requires many samples for each pixel. As rendering time is proportional to the number of samples, this means an image might be 100x slower to calculate (which still might be only a few seconds for simple geometry such as the knot).

It is true that the color bleeding (e.g. red shadows on the white column) on DarkBeams images requires multiple bounces - which again increases rendering times. I plan to do a path tracer implementation at some point, which will allow this - not sure much is gained when working with something with such complicated geometry as fractals, though.

I know it is different, anyway when I render with reflects putting a dark reflective colour some "color bleed" appears, making very nice effects... Don't know if it's the same I guess no? :) Anyway just a bit of blurring of shadows would be more than enough, I never use hard shadows btw ;D

it's not only the shadows, that make the difference in radiosity, it's also the color interaction between close surfaces. hdr shadows are already simulated in ambient occlusion and - depending on the scene - not so bad. but the "color bleeding" is something different...
(http://upload.wikimedia.org/wikibooks/de/b/b4/Blender3D_RadiosityTwoMeshes.jpg)

I know, but this kind of color interaction can be simulated quite easily with path tracing (for a nice WebGL demo see here: http://madebyevan.com/webgl-path-tracing/) - radiosity is just another way of doing it.

I'm not sure I understand you with the shadows: ambient occlusion is different from soft shadows - AO is not dependent on the position of the light sources and will not be directional. It is also possible to estimate shadows based on light distributions from HDRi maps, but this is computationally demanding.

i think you understood,what i mean. sure, ao is not sufficient to simulate hard contrast maps, but a sky for example with relative constant light is well represented with ambient occlusion - the sunlight is not. but sunlight is also not well represented with global illumination, as the sun normally casts hard shadows.

fantastic realtime demo. i see, that i'm not up to date in those things. my (a bit aged) vesion of cinema 4d needs minutes to produce similar results.
i thought gpgpu is the future, but it's present time too  :dink:

Very interesting! So the fuzzy mirror is called "glossy" ;)

Btw, shooting hundreds of rays per pixel is probably not a good idea for M3D (where this feature request was posted).

But the color bouncing/bleeding effects can be simulated in screen space in a way similar to SSAO - this is called SSDO:

http://www.geeks3d.com/20100619/ssdo-screen-space-directional-occlusion/
http://cg.alexandra.dk/tag/ssdo/ (sorry, in danish)

I'm not sure it adds much to the realism, but at least it is fast to compute :-)

Looks like an awesome idea, thanks! :beer: :beer: :beer:

In the meantime I made a room Difs formula with differently colored walls... ;D


Mandelbulb3Dv17{
Z.....x/...r3.../....2....URw8.Q.xCpzqSzkBKh6./Eus1dtEUAIz9qX0XQDgmgysrUO8uaqRsD
................................UcA7IS5db.2........A./..................S/2...wD
.........cks/.../M.0/.....UF0...8.....E4.....M1iZvSkeSlD/......./K3o0dkpXm1.QJJV
./EnAnQD12..2..dDIRZanH1.lhoC1mtCxzjx7yQCif9K.A............u1....y1...sD...../..
.z1...sDV.lKPUB1YwnayWlLcAbMyqAp2k5p3Onj2dfAa1qWEqHTSr/foQ8CzqLjjMw6mEaD6M5uxFRV
qwfCoXHF3r4Iy0uGga/qAEmj......IX..............kD.I....sD..E.0...................
.............oAnAt1...sD....zUmbo9.mTGj.c/8x0UEco9.eVGj.678x0Uico9....../....k1.
..................kz.wzzz1.U..6......M4...EB....W..........F....8/...I1.....SN52
...U.qFG9yzb2zzzRYoWzz7lz16/4IZzzzzz..EzRQrzS4uD3Yl1z/xv9z9.s6ZeQQLlz2.28.kFrA0.
.Ub96aAIVz9.1se7Umvxz0........../EU0.wzzz1...........s/........./.........E.2c..
zzzz.............0...................2./8.kzzzD.................................
/EU0.wzzz1...................................wzzz1kzzzjvz10U.06.zzzzS9R7z1...wzz
zTi0g.6U.0kzzzjvxy1U.06.zzzzi1.I.06U.wzzzviHF/6U.0kzzzjvvJ3U.06.zzzzi5wJ.06U.wzz
zvCmw/.U..kzzzjv..EsUa3.eeWCNq0.IJ36wk8.wyLsUa3.................................
E....2....E.....I....o....UQjxKPXlaQ7NoI.A5...........................U1C....6..
..........UaNaNaNaNiz..........E................................OaNaNaNaty1.....
..............zD........kz1........wz..........E...................wz...........
........................}

Maybe you noticed that i already did some monte carlo simulated renderings in a hidden function in m3d because of lack of user interface and like Syntopia mentioned, the long render times.
So this image took 25 minutes on my older laptop, whereas the usual rendering would take 5 to 25 seconds dependend on the calculated options.
(I sell the activation procedure of the mc easter egg in m3d for the most bidding one :dink: )

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

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

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.


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.


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  :dink:

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"

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;)

Another possible solution; use "obscurance" method :D

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

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.