21
\$\begingroup\$

I want to take some pictures of some crystals. Being crystals, they don't move around much so I can set up my camera and lighting as I like. I have a high-powered halogen light shining on them, and my camera on a tripod above them. When I took photos in automatic mode, they came out being quite noisy, so I did some reading on this site and ... got incredibly confused.

So, what are my best options for taking these photos? Given the set-up, I have no restrictions (that I can think of) on shutter speed, aperture (the crystals are pretty flat so no depth-of-field issues), or ISO. I'm a little loath to do much post-processing to remove the noise since I want to see the fine detail on the crystals.

Here's the crystal with the automatic settings (f/9.0, 1/250sec, 3200ISO).

enter image description here

(The camera is an Olympus E-PL1 (a micro 4/3rds) and I'm using a 14-42mm lens on full zoom to get a decent size image of the crystal.)

(I strongly suspect that this has been asked before, but I couldn't track it down. So if it gets "closed as duplicate" that'll be fine as it'll tell me where to look!)


I clearly need to do a lot of experiments on this, but using a combination of the suggestions, here's a much better picture of a crystal:

better crystal

To get this, I shot it in daytime (benefiting from a bit more light) with ISO100, 1/4s, f/5.6. I shot in RAW and did some post-processing (some adjusting of levels, a little sharpening followed by a small blur to reduce the little noise that was there). As I said, many more experiments needed but at least I now know what to experiment with! Many thanks.

\$\endgroup\$
8
  • \$\begingroup\$ f/9 may be a bit much on an m43 sensor I think. You are getting into diffraction territory which will make the photo a bit unsharp all over. Open up a bit, I think f/5.6-f/8 should be optimal for that system. Not that this has anything at all to do with the noise, just thought I'd mention it. \$\endgroup\$
    – Staale S
    Commented Feb 4, 2011 at 20:36
  • \$\begingroup\$ @Staale f/9 may be the widest setting that provides sufficient depth of field. The difference in diffraction effects between f/9 and f/8, or even f/9 and f/5.6, is still very small. \$\endgroup\$
    – whuber
    Commented Feb 4, 2011 at 20:51
  • \$\begingroup\$ Try to use whiter lights as well, though if you're shooting raw, fixing white balance isn't a big issue. \$\endgroup\$ Commented Feb 4, 2011 at 22:22
  • 1
    \$\begingroup\$ The more light you can throw onto it the better off you'll be. It will allow you to increase your depth-of-field by using a smaller aperture, and, at the same time, reduce noise by reducing your ISO. I'd go for a lower ISO as the priority because it will reduce the noise, allowing you to crop into the image if necessarily. \$\endgroup\$
    – Greg
    Commented Feb 5, 2011 at 2:03
  • 1
    \$\begingroup\$ There is another related question: photo.stackexchange.com/q/11/67 \$\endgroup\$
    – chills42
    Commented Feb 10, 2011 at 15:03

10 Answers 10

30
\$\begingroup\$

The best way is to use the lowest ISO possible (100 is often the best), and slightly overexpose (without clipping highlights), then post process it back down. This will help to decrease noise in the shadows.

Also shoot raw, so that any adjustments can be made before the conversion to jpeg.

\$\endgroup\$
10
  • \$\begingroup\$ +1, and I second the suggestion to slightly overexpose (ETTR, expose to the right) without clipping. It should be noted that if you do slightly overexpose to maximize your dynamic range, you might need to boost saturation a tad in post processing to compensate. Otherwise, the lowest ISO and a longer exposure...get more light down the lens, will help immensely with noise. \$\endgroup\$
    – jrista
    Commented Feb 4, 2011 at 20:48
  • 4
    \$\begingroup\$ Watch the histogram, either on your camera or in your results. \$\endgroup\$
    – mattdm
    Commented Feb 4, 2011 at 21:42
  • 4
    \$\begingroup\$ Just note that exposing to the right at anything other than the lowest ISO is a redundant exercise. Shooting +1 EV at ISO 200 is not as good as +0 EV at ISO 100. When at ISO 100, then shoot +1 EV. But make sure you're not clipping. \$\endgroup\$ Commented Feb 4, 2011 at 22:13
  • 1
    \$\begingroup\$ @Nick, that does not tally with my experience in this matter. A TTR 1600 actually has less shadow noise than an on-meter 800, I have found. There are other drawbacks such as smaller dynamic range, that is true. \$\endgroup\$
    – Staale S
    Commented Feb 4, 2011 at 23:00
  • 1
    \$\begingroup\$ @Nick Bedford I just read that chromasoft article, basically he makes comparisons about noise whilst allowing the G10s built in noise reduction to do unknown things to the images! I repeated his experiment with a DSLR and got the opposite results. I will post the results soon, but basically they confirm ETTR works, even at high ISOs. \$\endgroup\$
    – Matt Grum
    Commented Feb 5, 2011 at 16:04
14
\$\begingroup\$

To minimise noise get as much light down the lens as you can. As you have a static subject using a longer exposure is probably the best option. Setting the ISO to the minimum value will help you let in more light. Select the aperture based on sharpness, I'd go for something in the middle like f/5.6

For the ultimate noise reduction, consider taking multiple shots and then averaging in software. See:

Here several short noisy exposures were combined to match the noise of single exposure with more light. However the methodology can be applied to several longer ISO100 exposures to further reduce noise!

iso_multi1

The top section was a 1 second exposure with a really small aperture so quite noisy. Averaging 16 such photos in Photoshop resulted in the middle section, with much reduced noise. The bottom image, for comparison was the result of a single long exposure.

The last image was shot at ISO100 so I couldn't use a longer exposure without overexposing and clipping highlights (I could have closed the aperture further but that wouldn't have allowed any extra light into the camera). By combining several ISO100 shots of this length I can increase the amount of light, effectively simulating the result of using an ISO setting lower than the camera could ordinarily produce.


A standard piece of advice for minimising noise to "expose to the right". This means place the histogram as far to the right (highlight) end of the scale as possible, without overexposing.

An article was posted in the comments to chills42's answer which claims:

"Expose to the Right" is just plain wrong

Here is the article

I repeated the high ISO part of the experiment with a Canon DSLR instead of the G10 compact, and most importantly I turned off noise reduction (the author let the G10 do it's own thing with noise reduction, with no attempt to keep it consistent).

The part I wanted to specifically test was the claim that by reducing ISO alone you could reduce noise equivalent to exposing to the right. I wanted to test this because this goes against both the theory of how noise originates and all my experience.

Here is a close up of a photo of a colour checker chart, 1/10s f/5.6 ISO1600 :

standard_exposure

And here is the result of reducing the ISO to 800 whilst keeping the other settings the same (1/10s f/5.6 ISO800):

lower_ISO

It looks similar, but there's ever so slightly more noise (see the composites at the end).

Finally here is the result of exposing to the right, 1/5s f/5.6 ISO1600

ETTR

A clear reduction in noise results. Here are the same images, presented in the same way as the chromasoft article. Firstly the camera metered exposure with the lower ISO image offset in the centre:

lowerISO_vs_standard

Finally, the camera metered exposure with the expose to the right in the centre:

ETTR_vs_standard

QED

Finally the article does have some useful things to say about exposing to the right causing problems with colours (you get loss of saturation as you approach the limit of the sensors saturation point). But on noise it is demonstrably wrong.

\$\endgroup\$
7
  • \$\begingroup\$ "effectively simulating the result of using an ISO setting lower than the camera could ordinarily produce" That could be the conclusion we can extract from that experiment... Was that was already known? \$\endgroup\$
    – tomm89
    Commented Feb 4, 2011 at 21:35
  • \$\begingroup\$ As somepoint, longer exposures will increase noise (due to sensor thermal dissapation) \$\endgroup\$
    – Alan
    Commented Feb 4, 2011 at 22:02
  • 1
    \$\begingroup\$ @Alan: That's true, but it takes a pretty long exposure to do that. I would put 30s at the very lower bound of the length of time needed to create enough thermal noise, and exposures around 1 minute or longer are usually needed to exhibit fixed pattern noise and an increased noise floor. \$\endgroup\$
    – jrista
    Commented Feb 4, 2011 at 22:22
  • 1
    \$\begingroup\$ Sure. I just wanted to point out longer exposure doesn't produce a decreasing amount of noise. \$\endgroup\$
    – Alan
    Commented Feb 4, 2011 at 22:49
  • \$\begingroup\$ @Alan: A longer exposure at a lower ISO to maintain the same overall EV does produce lower noise, until you reach an exposure time that begins to exhibit thermally related noise. If the original exposure was 1/250th of a second at f/9 and ISO 3200, you could easily open the aperture to f/3.2, lengthened the exposure to 1/30th of a second, and used ISO 100 for considerably less noise at the same final EV. \$\endgroup\$
    – jrista
    Commented Feb 5, 2011 at 4:56
9
\$\begingroup\$

Set your ISO to as low as you can. And if you'd like, take multiple exposures and average the results.

Because you're on a tripod, the shutter speed won't matter so much. One thing that will definitely help, though, is to use the timer function (ie, take the shot 10 seconds after the shutter is pressed) so that you're not touching the camera when the shot is taken. This time will mean that any vibrations from you touching the camera to set everything up will be damped by the time the shutter fires, so the image should be sharp.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ I particularly like this answer for the timer suggestion. Not something I might have thought of otherwise (fortunately my camera can do 2s delay so I didn't have to wait too long between shots). \$\endgroup\$ Commented Feb 10, 2011 at 22:06
  • \$\begingroup\$ Glad it can be of use to you. I figured it out when taking my own crystal photos. \$\endgroup\$
    – mmr
    Commented Feb 10, 2011 at 22:26
4
\$\begingroup\$

Two things cause noise in your pictures: amplifying the sensor (higher ISO), and longer exposures (> 1 minute). What you want is a good balance between them. With the Olympus E-PL1, you probably won't be able to tell the difference between ISO 100 and 200 but you will start to see noise start affecting your picture as early as ISO 400 (based off of the DP Review). Also, for some reason jpeg seems slightly less noisy than RAW, but jpeg will degrade with every edit.

Keep in mind that at close focusing distances, your depth of field is razor thin. With the camera 10 inches away at 42mm and f/9 your depth of field is .16 inches in front and behind the focal plane. Where you place the focus can be critical with a picture like this.

To get optimal noise results with your camera, you want to put your camera at ISO 200, aperture priority or full manual, f/9 - f/11 and at least 12 inches away, and whatever exposure time is necessary. If your required exposure is approaching a minute (the max on your camera), find some more light. An effective way to add a lot of light is to use a flash, but you'll want it mounted away from the camera so you aren't bouncing light off the mineral into the lens.

\$\endgroup\$
3
  • \$\begingroup\$ I'm intrigued by what you say concerning jpeg versus raw. Surely if the jpeg is less noisy than the raw, then it's a digital after-effect which I could do later on my computer, or am I just showing my ignorance of how modern cameras work!? \$\endgroup\$ Commented Feb 4, 2011 at 21:35
  • \$\begingroup\$ That's based off the samples from DP Review. It definitely is contrary to conventional wisdom. My guess is that there is still some noise reduction happening for JPEG even if you tell it not to. Software noise reduction will reduce your sharpness, so it's usually one of the last things you want to do (before sharpening). \$\endgroup\$ Commented Feb 4, 2011 at 21:57
  • 1
    \$\begingroup\$ Raw is only more noisy if it hasn't had noise reduction. Your camera is most likely applying noise reduction as JPEG is created from the raw data whether in camera or in your photo suite. I would just shoot raw. \$\endgroup\$ Commented Feb 4, 2011 at 22:17
4
\$\begingroup\$

There have been some great suggestions so far. I have a more off-the-beaten-path suggestion, however, in case you want to get the absolute maximum quality, clarity, and detail you can.

It looks like you are doing some macro work, so I understand that your DOF is very thin. At a wider aperture, you should be able to use ISO 100 at decent exposure times, at the cost of even thinner DOF. You can compensate for the thinner DOF by taking multiple exposures with an increasingly distant focal plane, and stitching them together with a depth of field stacking tool like Combine ZM. This is a common technique amongst macro photographers who use the Canon MP-E 65mm 1-5x Zoom Macro lens, which has a DOF as thin as 0.02mm at maximum magnification AND an aperture of f/16.

\$\endgroup\$
4
  • \$\begingroup\$ Useful to know. In this case, the crystals are themselves "wafer thin" (to coin a phrase) so I'd have to be using something like that Canon lens to really notice it. But I have some other similar pictures to take that aren't so thin so I'll bear this in mind for them. \$\endgroup\$ Commented Feb 4, 2011 at 21:38
  • 1
    \$\begingroup\$ If the crystals are thin to start with, I would open up your aperture as wide as you can and still maintain the sharpness you need. You are currently using f/9, which is a significant part of why you need to use a higher ISO. \$\endgroup\$
    – jrista
    Commented Feb 4, 2011 at 22:23
  • \$\begingroup\$ Actually, I partially retract my comment. Even though the crystals are thin, my ability to focus precisely isn't so having a bit of lee-way from the DOF helps considerably. \$\endgroup\$ Commented Feb 10, 2011 at 21:59
  • \$\begingroup\$ Yeah, at very close distances, especially with a macro lens or extension tubes, your DOF can literally become vanishingly thin. Less than a millimeter thick if you get close enough. I would stop down a bit, maybe even back away from the crystals a bit, and expose for longer. You have a still scene, so you should be free to expose for as long as you need to at ISO 100 to get the correct exposure with minimal noise. \$\endgroup\$
    – jrista
    Commented Feb 11, 2011 at 3:50
3
\$\begingroup\$

Keep the temperature of your camera low to keep the thermal noise to a minimum. At the very least, this means not taking pictures after it's been busy for a while. Hot lights and hot rooms also work against you here. Things that draw a lot of power will generate more heat too. The LCD is a factor here; things like Live View (on-screen viewfinding) and video mode can heat up the camera faster.

Noise from heat can increase the image noise significantly. Here's one experiment:

Cooling a Canon 350D results in less camera noise being present in raw images... For the cooled camera the measured noise level after 40 exposures was a whopping 40% less than in the uncooled camera.

(Emphasis mine)

And this isn't "put the camera in the freezer" cooling; the temperature was kept at a room-like temperature of 18C.

\$\endgroup\$
8
  • 1
    \$\begingroup\$ Relevant for astro photography but quite off the mark in this case, I think! \$\endgroup\$
    – Staale S
    Commented Feb 4, 2011 at 20:37
  • \$\begingroup\$ @Staale Makes sense, but are you sure? Has anybody experimented with lowering the camera's temperature to see whether noise improves? \$\endgroup\$
    – whuber
    Commented Feb 4, 2011 at 20:49
  • \$\begingroup\$ Given that the basic issue is noise caused by shooting at 3200 ISO, I will stand by my statement that thermal noise is absolutely and totally irrelevant in this case, yes :) \$\endgroup\$
    – Staale S
    Commented Feb 4, 2011 at 21:12
  • 1
    \$\begingroup\$ No doubt that ISO 3200 is the cause of most noise. But thermal noise is still a factor, and thus I thought it should be mentioned. \$\endgroup\$ Commented Feb 5, 2011 at 1:11
  • 1
    \$\begingroup\$ "At the very least, this means not taking pictures after it's been busy for a while": I think that this piece of can be very relevant here. I just had a longish session of macro photography using live view, and it heated up the body so much that my ISO 100 shots had much more noise than what I would expect in normal conditions. \$\endgroup\$ Commented Feb 5, 2011 at 11:53
2
\$\begingroup\$

Switch to an aperture priority mode, use an aperture between probably 8 and 16 to make sure the whole object is in your depth of field, lower your ISO to 100 (or as low as your camera goes), set it up on a tripod, and click. You'll have a much longer shutter speed, so some stable surface will be necessary.

Its mainly your high ISO on a smaller sensor thats killing you here. Your camera doesn't know that its on a tripod and shooting a static object, so it thinks you need a reasonable shutter speed to handhold, so its increasing your ISO.

\$\endgroup\$
1
\$\begingroup\$

Lots of good answers thus far but I did want to mention one other consideration: if photos with an absolute minimal amount of noise are critical, you might want to consider a different camera. The larger the sensor, the less noise you'll see (all other factors being equal). A micro 4/3rds camera has a smaller sensor than a APS-C (crop sensor) DSLR, which is smaller than a full-frame DSLR.

I'm not suggesting you rush out and buy a new camera, but if low-light, low-noise photos are important the sensor size should be a consideration in your purchasing decision.

\$\endgroup\$
1
\$\begingroup\$

Using a tripod allows you to use slow shutter speeds, thus allowing you to use lower ISO, resulting in less noise. More importantly, a good combination of your light and some reflectors will get you the best image that you want.

Also consider software noise reduction in post processing if you're still seeing more noise than you find acceptable.

\$\endgroup\$
0
\$\begingroup\$

I know of only two ways to really control noise: correct exposure and more pixels. I know this sounds snotty, but it's true. Use logic when you set up. Correct lighting means bracketing; more pixels means framing: frame the shot, zoom the lens and shoot again, or better, if you can, move closer or further away.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.