Sorry for taking three weeks to reply to this, but I was out, got busy, had assignments, and so on.
Anyhoo, here's where we are now.
First, in response to your request that I post my images showing the noise (and the lack thereof), I have a number of images to produce. I want to also take this opportunity to help clarify what may be a drastic misimpression by some people about what "noise reduction" really means vis-a-vis Canon's 1DsMII (or the other models, for that matter). I'll get to that last, but the first stuff is needed to establish the groundwork.
The first photo is the original star-trails image that I posted in my previous note, where I exposed for TWO hours outside in 40F temp, followed by the two-hour dark frame exposure inside at 75F temp:
http://www.danheller.com/tmp/noise/clean-noise-1.jpg
If you exmine it closely in photoshop, you'll see a pretty clean/smooth image. To give a contrast, here's the second image I mentioned in my last post, which was a 30-min exposure at 50F, where the temp dropped to the 30s during the dark-frame:
http://www.danheller.com/tmp/noise/clean-noise-2.jpg
Here, I've labeled noise "A" in the photo. These are bright red spots--slightly bigger than "dots".
*THIS* is the noise that the Canon NR is supposed to minimize. Now, despite the fact that there are two red spots, this is still a dramatically clean image for the shooting conditions. It's easy to clean this stuff up. The noise indicated by "B", however, is subtle splotchy reddish stuff that is most likely NOT the type of noise that canon's NR will do anything about. It's an entirely different sort of thing that readers should be aware of.
To illustrate this, the third image is a two-hour exposure with NR turned OFF entirely:
http://www.danheller.com/tmp/noise/no-NR.jpg
Here, the red dots are EVERYWHERE--because NR was off--but there is NO appreciable increase in the amount of the other type of noise. There's no difference at all, IMHO.
What's more, this third image illustrates this "normal grain" as a standard shot taken of my son:
http://www.danheller.com/tmp/noise/normal-grain.jpg
(Hint: my son is on the RIGHT.)
Here, the noise appears much like standard film grain, because the ISO rating is 1600. It's a pretty nice grain for a 1600ISO shot, IMHO, but I use this image to further demonstrate to many readers who may have a misimpression of what canon's noise reduction does. It does NOT reduce this type of noise--or, what I call "graininess." Canon's NR serves only one purpose: to reduce the big, ugly red spots in long exposures.
Now I can directly reply to Chuck's comment here:
> I must comment that Canon Inc. has never made any claims about the efficacy of its noise reduction technology for exposures longer than 30 seconds. The fact that there is an improvement during exposures longer than 30 seconds is quite welcome, but it's an extra benefit, not an advertised feature that we must support.
The fact is, the type of noise that NR actually improves rarely even SHOWS UP AT ALL at exposures less than 10-20 minutes. I've experimented with this extensively in the past (because I'm a huge fan of long exposures), and it wasn't till I got involved in this thread that it occurred to me to piece all this data together to come to this realization:
Canon's NR is fantastic technology, and it does a great job at what it DOES. But here's where we can come to conclusions:
1) The need for dark-frame exposures is unnecessary. The camera should cache a number of dark-frame datasets that would be applied to the same in-camera firmware algorithms to remove noise in long exposures that it currently does. How many frames it needs to cache is a matter of performing tests that no one other than canon can do: apply a ten-minute dark-frame dataset to a 12-minute exposure, and so on (using the same type of regression testing methods) across a spectrum of about 6-8 hours.
2) If (1) is done, then NR needn't be an option that you turn on or off--it simply applies the appropriate dark-frame dataset to any picture that the algoritm is programmed to accept. Obviously, there is some some performance hit because of this, then the option to turn it on or off could remain, but I suspect that there would be no adverse user experience or image degradition by applying NR algorithms accordingly.
3) Canon should clarify (using pictures) what exactly NR is intended to clean up, and what it is not. (A clarification that seems to be necessary given the misimpressions out there by a large segment of the community.) Much of that info is implicit on its website (yes, including pictures), but it's deeply embedded under intimidating "techie" papers (even though nothing really "techie-like" is really discussed. (Still, it's fine work! It's just not as complicated as the paper portends itself to be.) Using the same photos in the general end-user doc (along with simplistic definitions) would be adequate for most consumers (who are buying an $8K camera).
Lastly, there's this:
> As long as we substitute "Canon Inc." instead of me personally, I totally agree...
It should be noted that at no time have I (or anyone else, given what I've read) felt that you were personally responsible for any of the tech matters regarding these things. You're our ombudsman to Canon, and I see you as our main conduit of communication to the shadowy decision-makers who reside high above in their Ivory Towers. We are all prepared to make human sacrifices for the benefit of better equipment, user interfaces, and all-around pictures for our livelihoods, creative outlets and who knows what else.
In other words, nothing bad at all is directed towards you--we just hope that we can help you communicate voices of reason back to Canon.... and, if possible, communicate their thoughts back to us. :-|
dan
Linear Mode
