What exactly is it that you really are asking? FFT is just an algorithmic transform to an abstract mathematical space which has any kind and any number of uses, one of which may be noise removal. Still, even seemingly trivial noise removal may involve amillion other techniques and algorithms - anything from simple thresholding to convolution operations - so I'm not quite sure what the question behind your question is...
Mylenium
Ditto what Mylenium has said... Do you want to achieve something in particular? If so, please say what that is. I'm sure we can help you achieve it.
No one but the Adobe engineers know the underlying algorithms, and they're not going to be likely to share much of the technical detail - it's proprietary.
-Noel
NR not very trivial. IMO anyway.
I had much the same questions when I moved from frequency domain to time domain. A strong preparation in analog engineering treated time domain sparsely, particularly noise. We dealt with detector noise in a nuclear physics lab by running the detectors in pairs at cryogenic temps. The detectors were connected to a coincidence circuit (and gate to you digital folks!). Even then, the output of the gate had to be carefully amplified, then subjected to an A to D converter where the output was displayed as a histogram. I'm sure things have progressed mightily since then, (ca 1950's).
I used the coincidence idea back about 2004 or so when I needed an image absolutely as noise free as possible by shooting the object (an exposed wooden frame for a building constructed from Port Orford Cedar.) twice, then constructing a curve applied to each image to allow adding them together to match the colors and values of a single shot. Extended versions of PS now allows this without any curve inventions!
Viva Al Gore Rhythms! (Didn't Bill Clinton play for him? )![]()
>> I'm sure things have progressed mightily since then, (ca 1950's).
Um, no, not really. Better scintillators, better light detectors, better amplifiers, better collection electronics... but mostly the same issues. You'd probably recognize 98% of the tech in the large hadron collider calorimeters.
I know that, but I was mostly thinking about the efforts at today's digital cameras. At one point, I really expected for really good noise control we would have to have a Peltier cooler attached with a battery pack to power it.
The mult-channel analyzer was quite the instrument. I attempted to convince the group to let me figure out how to attach a microphone to the rig and see what we might capture. No dice.
At the moment, even with the D90, I generally match the ISO to the scene dynamics, not noise considerations.
I just printed a 36" long print stitched from 5 panels from a D800. Gonzo print!.
Yet, the same number of panels from the D90 is equally (almost!) impressive.
The big advantage to sensors like in the D800 up is the fact you can throw away gobs of data and never miss it. I have a RAW preset called Dramatic B&W. From a D90, dropping the sky that far introduces nasties, with which I have to deal. From the D800, no problem!
Yes and I opened it up and messed with it. I don't recall much about it other than it was pretty damn fabulous! It's harder to evaluate a camera with images I did not make, as I don't have a complete understanding of what went in it. In any case, color is not my main goal in photographing these days. I'm back big time to grey scale! Now, instead of worrying about setting proper white balance, it becomes another tool for messing with grey value relationships. You would not believe how a standard color image looks after messing with it in B&W, then turn the color back on. B&W: dynamic. Color: garish.
That stitched pano came out at 249M native file size (no upsampling). Print size10"x25" at 720dpi
North America
Europe, Middle East and Africa
Asia Pacific