Yeah, it’s not really related to OGL… but I was wondering if anyone has seen any good papers on the subject, or is aware of any commercial (or otherwise) programs out there that implement it (other than Genuine Fractals).

Btw, it seems everyone’s interested in using it for image compression (which is fine, but) I’m hoping to use it for image upscaling and enhancement. Some add-ins do this with genuine fractals, but I’m convinced the method they’re using (to compress the image) isn’t giving the best/most accurate or asthetically pleasing results.
Anyone tackled this project before or have any comments? Papers? Programs? nay-sayers?

I’m not usually one to respond so glibly, but have you tried Google? Doing a search with the keywords “Fractal Interpolation” brings up a large number of results, some of which appear to be papers on the subject.

Hmm, I’ve been reading everything I can find (unfortunatly I can’t seem to find Jacquin’s original paper.)
The more I read the more discouraged I become - it appears fractal compression, or more specifically PIFS (partitioned iterated function systems) perform what I had in mind, just (because of speed requirements) do a less than optimal job of it.
I’ll keep checking and try to have a decent program coded up before too long.

Unfortunaly the patent on the fractal technology is rather all inclusive, so most anything i come up with my end up falling into the categorization of their patent…

Hmm… any input is greatly appreciated, I suppose I’ll leave ya’ll alone now,

You’re trying to invent detail where there is none to begin with?
Sounds to me like that can only reasonably done if you know something about the basic texture of the object photographed.
For architecture you’d want straight lines, for clouds you’d want something fuzzy.
To be honest I doubt that fractals are the best way to come up with that sort of stuff.

You could check out wavelet transforms - using a coarser version of the wavelet on the analysis side and a finer version on the synthesis side. (Although I’m not sure how one would take wavelets into the third dimension - I’ve only checked them out for sound data).
What’s nice about wavelets (nicer than fourier transforms) is that you can use different wavelet shapes.

The idea is that there is multiscale detail in the image, wherefore, given some input waveform, the same waveform (albiet with some transformations) exists at a larger scale (which, with a regular grid of datapoints, means (possibly) higher frequency data). The tricky part is deciding which transformations should be “ok” to perform, and finding the best matching data.

Imagine an image of a field - because of perspective the grass blades get smaller as they approach the horizion/get further away. A good algorithm would match the low detail data of the grass with a scaled (and perhaps other transformations - rotation, mirroring, chroma scale or shift, lumniance shift or scale) version of the high detail data.
This is how fractal compression works - except they try to keep the number of source data “blocks” down, and limit the transformations to something that can be described briefly (bit data/flags)
My understanding of IFS is that they take it a step further in that (well, Iterated Systems actually only deal with the luminance, then add chroma (it appears in the patent, anyway)) they can iteratively apply the function to that block, in successive size and amplitude scales to “add” detail.
Perhaps Iterated Systems hasn’t done that (it kinda doesn’t look like it in the patent info I skimmed over, and Jacquin’s PIFS stuff) or perhaps they have. To tell you the truth, I’ve only seen input and output images, I haven’t been able to work with Genuine Fractals.

Anyway, the idea is to manufacture higher detail from existing detail in the source image.

You’re right though, there’s only so much you can do with data interpolation. (Of which I’ve seen at a moderate deal.)