This thread was an interesting read, and then I saw this and had to reply:
But I have in mind a wonderful head made by small and countless iterations of the same head with all its details, hairdo, etc, moreover, I want different hues and shades for them, moreover, its outer cover is transparent and let see some organic inner structures made of different materials.
It´s nearly imposible to achieve such an image with standard programs, or if it can be done it will cost a lot of time and effort, but this would be nothing for this tool .
In your terms, it´s a kind of universal and editable formula, if I can say it.
Well, I must say that I have created just such a human head as you have described, in that everywhere you look you see smaller and smaller versions of the same head! (without the transparent surface or hairdo though). I don't have anything to post just yet because I am preparing a high-quality render for the contest. But I can tell you a little about how it was made, and I think it relates to your Gnomon idea. The concept I used was a literal interpretation of a Multiple Reduction Copy Machine (MRCM). This is similar to an IFS, in that you define transforms, but IMO is much more intuitive and flexible. The transforms were defined using the collage theorem. In other words, the first iteration is hand-placed. I made the head in a standard 3d package with just a few lines of scripting, and while the fractal rules are certainly simple, defining the transforms took quite a bit of artistic (
not mathematic!) input to achieve a human likeness. It was akin to doing any other artistic 3d modelling. I was able to intuitively "tweak" the details until I was satisfied with the result. Here's a render of some broccoli I made with the same method. Please note that this is polygon based, so is easily amenable to the techniques for lighting/texturing/rendering available in general 3d packages. In this case I chose a simple Ambient Occlusion look
I have also implemented 2d versions of an MRCM in Photoshop without a single line of scripting, and without a single calculation on my part. It doesn't get much more artistic than that! The process involves recording an Action to define all the transformations, then playing back the action as desired. To define each transformation, the base layer is copied and then manipulated. The final step in the Action is to collapse all the layers into a single base layer again. This process gives you easy access to powerful (non-affine) transformations as well as color modulation (and any other effect recordable in an Action) that will be repeated automatically at ever iteration.
This simple smiley fractal I whipped up real quick shows what I mean. The layer copies were deformed (transformations involving more than just scale/rotate/translate/skew), and the outer "border" smileys were darkened every time. This is not possible with standard IFS approaches. The absolutely amazing thing is that now that the action is defined, I can start with literally
any image and it will converge on these smileys after a few iterations!