Long Exposure In-Camera noise reduction - is it worth it?

TylerW

I recently came across this following photo and comment, and since, its been a topic bouncing around in my tiny little mind:

http://www.flickr.com/photos/mumbleyjoe/2043508173/

Cliff notes version: the photographer had a five minute window to get a shot. He took a 2 minute exposure, which then called for an additional 2 minutes for the in-camera noise reduction. Meaning that while his exposure was only two minutes, he really only had one chance to get the shot.

My question is this - is the in-camera noise reduction that much better than anything you would get out of NeatImage / Noise Ninja? Or are you better off shooting noisy images, having the flexibility to work more quickly, and then taking your images home and tidying them up in post?

schmoo

Personally, the one occasion that I used te in-camera noise reduction program I ended up with a much grainier photograph. This photo was a 30 min exposure, as was this one, but the first was using noise reduction so it was a full hour before I was able to review the shot. When I view it at 100%, noise reduction seems to have cause little black dots/lines all over the image. I wasn't impressed and I never used it again.

I regularly take shots as long (or short) as 2 min and it surprises me that he would find it necessary if he exposes it right. Given his comments in the gallery about how he doesn't tend to do much PP, I'd think he meters correctly the first time. Then again, perhaps someone with more experience can chime in and say that it's more effective for shorter exposures or something?

I've never used Noise Ninja so I can't speak for that. It's been something I keep meaning to look into, however....

Beautiful images in that gallery!

ian408

I don't know about the in-camera. I don't think you get better than say
Noise Ninja or Neat Image. Plus the processing time is far too long.

ziggy53

In-camera, long-exposure noise reduction is simply an automated method of "dark frame subtraction". When a digital sensor is left on for a long period of time (time exposure), it heats up. How much it heats depends upon how long the exposure lasts, ambient temperature and the heat dissipation properties of the system involved. A heated sensor generates visible noise in the form of "hot pixels". Which pixels get visible are usually repeatable and consistant in both position and intensity.

The problem is mostly visible where there is a large dark field in the image, like with astro-photography.

You can accomplish pretty much the same thing by taking a series of long exposures followed by the taking of a "dark frame" of the same duration and then manually applying "dark frame subtraction" as a part of post-processing (pretty easy). This way you avoid the need for a dark frame after each exposure and it both speeds the process of acquisition and it saves battery life.

You can determine later if you don't need to process using the "dark frame subtraction", and just skip the step.

More information on "dark frame subtraction" here:

http://www.astropix.com/HTML/J_DIGIT/JPG_DFS.HTM

HarlanBear

Ziggy has the line on info in this area, for sure. But I never use in-camera nosie reduction using a Nikon D70 (which is prone to noise) and I do get it, especially in night skies with long exposures or high ISO shots. I have tried the "dark frame" technique Ziggy describes, but I prefer using Nosie Ninja and love the results. Plus I feel I have control of the process. There are times I will then need to mask off the results for detail in buildings, etc., or play with the strength. But for my workflow and semi "old school" shooting, I don't like the in-camera system and prefer to deal with it in post.

YMMV

ChrisJ

I would love to see a side-by-side test of a 30 or 60 minute exposure. But it takes so long (and uses batteries)! I've used the correction sometimes and other times not. I haven't noticed it hurting my images, but if it helps, it's marginal.

Nikolai

1) If I only had 5 min window to get a critical shot, I'd turn off everything that would slow me down. Noise is definitely something I can deal with in post.
2) Having said that, Canon 40D has some different way to kill the noise. It's also a dark frame method, but it's notwhere near the original exposure timewise. Thus far I never went for a super long ones with it, but on 30sec exposure the lag was around 5-7 seconds only (as opposed to full exposure time on 20D and 30D).
40D also has high-ISO noise reduction, which improves the noise handling even further, making ISO3200 fairly usable and ISO1600 almost a no brainer.
Just my 0.000002 of the f/stop

TylerW

Nikolai wrote:

2) Having said that, Canon 40D has some different way to kill the noise. It's also a dark frame method, but it's notwhere near the original exposure timewise. Thus far I never went for a super long ones with it, but on 30sec exposure the lag was around 5-7 seconds only (as opposed to full exposure time on 20D and 30D).

Interesting Nik, with my 40d, if I take a 30sec exposure, its down for another 30 secs of noise reduction.

Nikolai

TylerW wrote:

Interesting Nik, with my 40d, if I take a 30sec exposure, its down for another 30 secs of noise reduction.

Weird! Maybe a combo of HighISO+long exposure wroks differently?
I was at ISO800 and 30sec

TylerW

Nik, I think you have a magical 40d. I did three 30 second exposures, all with long exposure NR. One at iso 100, one at iso 800 w/o high iso NR, and one with high iso NR. All of them took around 30 seconds of noise reduction.

Nikolai

TylerW wrote:

Nik, I think you have a magical 40d. I did three 30 second exposures, all with long exposure NR. One at iso 100, one at iso 800 w/o high iso NR, and one with high iso NR. All of them took around 30 seconds of noise reduction.

Interesting.... I think I need to check mine.
I suppose you were RAW/full manual? And what firmware version?
I had 1.0.3 in Glacier (when I shot at night), now I'm with 1.0.5 as of few days ago..

ian408

So another way to do this is to make multiple short exposures and stack them.

This is a technique used by astrophotographers to help with noise. It's also
true that a good number also use cooled sensors

Nikolai

Ooops.. I guess I never had that setting ON.
Yes, I tried 30 sec on 40D tonight, and sure it took another 30 for the dark frame

ziggy53

ian408 wrote:

So another way to do this is to make multiple short exposures and stack them.

This is a technique used by astrophotographers to help with noise. ...

That technique helps with "random" sensor noise, but hot pixels due to thermal noise are not random and would be accumulated along with the image.

A dark frame is just a mask built against the thermal noise. The subtraction process negates the dark frame from true image pixels.

Rhuarc

Well, I know this is pulling up an old thread, but I had some questions related to this. I am interested in getting into some night photography, and was wondering if I could pick people's brains regarding this whole Dark Frame, Hot Pixels, etc...

I have the XTi, and would appreciate any information people can give me in relation to this camera. I will kind of list a few of the questions I have and people can answer as they will.

1) Does the Long Exposure Noise Reduction in the XTi actually use the Dark Frame method and take a dark exposure after the light exposure for the same period of tiem then subtract it? (I take a 5 min. shot, then the camera would take a 5 min. dark frame, and then subtract the two.)
2) If I want to use the Dark Frame method myself does the dark frame need to be the same length exposure, or if it is take right away will the sensor still be "hot" and then I can just use a short exposure time?
3) Do I need to take a dark frame after each shot, or only after I change the exposure times. (So if I was using a 5 min. exposure time I would take a single dark frame for 5 min. and then use that for all exposures for that night of 5 min. If I switched to 3 min. I would take another dark frame for all exposures of that length)
4) What exposure times does the XTi use the Noise Reduction for? (I realize it should actually be called Hot Pixel reduction, not NR)
5) If I want to do the Dark Frame method manually, what exposure times should I start taking Dark Frames for? If I take a 5 sec. shot will there be hot pixels? 10 sec? 30 sec? You get the point! LoL

6) On a slightly unrelated note, does it really work for reducing noice on a High ISO shot to take 3 bracketed shots that are 1/3 apart, and then stack them? I read that since true noise is random, that if you take 3 shots then stack them using the average in Photoshop then in each frame different pixes will have noise, and so the chances of having every pixels exposed correctly in the final combined shot go up.

Ok, well, hope all this made sense! Any help anyone can give me would be greatly appreciated!

ziggy53

Yes, I believe that the XTi uses a "dark frame subtraction" method for long exposure noise reduction.

Yes, it is similar to what you can do manually and that is what people did before the method was automated in the camera.

You can do a single dark frame to use in all subsequent long exposures, and that's how some people still do it, especially for "extremely" long exposures (typically 1/2 hour to hours long exposures).

Using a single dark frame will probably not be as accurate as individual drak frames except when the camera reaches equilibrium, and that is anyone's guess.

Yes, it is pretty important to use a dark frame of the same duration as the actual exposure. The sensor will heat differently through time and even depending on the environmental conditions at the time of exposure.

More information:

http://www.dgrin.com/showpost.php?p=332163&postcount=5

---

Yes, people do use multiple stacked exposures to control "random" noise. Obviously there are limitations on when that method is appropriate. There are also studies to show that multiple different exposures used for some HDR technique can also help reduce high-iso noise besides the direct reduction from not "boosting" dark noise. I am not sure if any HDR automation software exploits that property.

Rhuarc

Thanks for the help ziggy!

Ok, so from what it sounds like unless I am planning on doing many long exposures of the same length I would be better off to just let the camera do it's thing in camera. If I am planning on doing 10 or 20 of the same length then it may be worth doing it manually just to save time? I suppose what time I do save would be wasted later in post, plus if i let the camera do it then the file I get to work on is still RAW, but if I do it manually then I am no longer working with a RAW file.