Tag Archives: long exposure noise reduction

Dark Frames And Your Night Photography

9 Replies

In an upcoming webinar (Down with the Noise) I explain a lot about noise: causes, contravention and cures. This is a bit of a prelude and addresses the questions:

  1. What is a dark frame?
  2. What do I do with my dark frame?
  3. What do I do if I don’t have dark frame(s)?

What Is a Dark Frame?

A dark frame is one or more images taken at the same exposure length, ISO and ambient temperature as the light (normal) frames but with the lens or body cap on the camera to prevent any light from reaching the sensor. When doing many kinds of night and low light photography dark frames can be quite helpful. And when doing star trails or other night imagery dark frames may save your bacon. A dark frame is what your camera does after a long exposure when long exposure noise reduction is turned on. But you’ll be far more efficient if you take those frames yourself. If you’re taking 100 light frames, e.g. for a star trail, you can take 3 or four dark frames and waste 50% less time (and not have gaps!)

Contrary to popular belief dark frames and long exposure noise reduction do little to reduce the random noise that is present in every exposure. That random noise is most pernicious in dark photos and shadow areas.  Dark frames, however are good for the following things:

  1. Reducing or eliminating hot pixels and amp glow
  2. Removing any “bias” in your image – that is bringing the black back.
    Want me to translate that: an unexposed area on your sensor should read as “0,0,0”  for Red, Green and Blue but I will bet you you don’t get zero!

This would probably be a good place to show you what a dark frame looks like. But you’ll be disappointed. Dark frames are usually quite black.  So instead of showing you JUST the dark frame, here is the dark frame boosted to show the speckles from hell – though they may not be obvious. Here I have made the speckles more obvious by boosting the darks using Curves in Photoshop. At this level of detail there are not any obvious hot pixels.

Dark Frame Overview – Boosted to show details. Note where the markers are – they are shown in the next frames.

And next is the same dark frame zoomed to 3200% unaltered. Hover your cursor over the image to see the same area boosted using curves.

Single Dark Frame (Linear Mode) Unmodified – or cursor over to see boosted version.

When inspected carefully, and with the dark level significantly increased it is possible to notice the hot pixels and possibly banding in a dark frame.  While there were plenty of red speckles and obviously green and blue as well in my stable of dark frames, the “hot pixels” didn’t leap out at me.  If you look carefully at the image you’ll notice I also used the color sampler tool to provide RGB values for 3 different locations on the image.  Of note is location 1 where the R (red) value is 14.  What is particularly worrisome about that value is that even after the entire frame has been boosted to the equivalent of 1.6 stops, you’ll notice that a value of 14 is still larger than all three colors at spot 3. After boosting, that red pixel really stands out with a value of 44.  Our first take away is that boosting the brightness boosts the noise.  The second thing to notice is that the red spot is NOT a hot pixel.  How do we know? Compare 4 dark frames (all boosted)

Boosted Dark Frame                             Click these–>   Frame 1 ~ Frame 2 ~ Frame 3 ~ Frame 4

Takeaway 2: There really is randomness!

Now that we have noticed the randomness, we realize that if we average enough of these frames together we can get the average “bias” – that is the amount of offset above zero in the image.  And if there are hot pixels, the good news is they will be in there too.

But How Do I Use a Dark Frame?

The simplest answer is to feed your dark frame(s) to a program that already knows what to do with them like StarStax, StarTrails or Image Stacker.  But you can do it yourself, and perhaps more elegantly using Photoshop.  How?  Place the dark frame as a layer over the image you want to correct and change the blend mode to Subtract (or difference). Adjust the opacity of the blend until it looks just right.

But I Did Not Take a Dark Frame, Now What?

All is not lost. If you have enough frames you can create a unique kind of dark frame. I took over 300 28-second exposure for a star trail along Lake Gaston in North Carolina.

In the image below I created the top frame using the Brighten mode in StarStax. I could just as easily have created the top frame using the StarCircleAcademy Stacking Action.

I used Darken mode to create the middle frame by feeding it my 100 darkest images. Using Darken mode as the stacking option means that hot or stuck pixels that are in every image as well as the lowest value of sky glow will be collected into a single result.

I then loaded the light (Brighten Mode) and the dark (Darken mode) frames into Photoshop. I placed the dark image over the brighten stack and changed the blend mode to Subtract.

What Happened Here?
Dark Frame Substitute process

Several interesting things happened:

  1. The hot pixels were almost completely annihilated
  2. The sky gradient caused by lights glowing in the distance was also almost eliminated.
  3. The contrast in the sky and elsewhere was improved
  4. The red bias on the railing was mostly removed.

A few less desirable things happened, too. The bright red glow on the railing once subtracted caused some of the railing to turn green. And the subtraction created some “holes” and “halos” in the image – especially where the brightest lights are found.  With some minor touch up, most of those issues can easily be fixed.

Is this the end? By no means! There are a LOT more interesting techniques to follow. Stay tuned.

SHARE this on Facebook or Google+ You know you want to help your friend out anyway, right!?

Trouble with Long Exposures – Part 2 of 2

14 Replies

In the previous article I discussed 4 of the 6 most common problems that occur with long exposures.  Those problems are:

  1. Poor Focus
  2. Dim Stars (low contrast)
  3. Strange Colors
  4. Purple or Pink Glow

In this installment we tackle these two issues

  1. Gaps in Star Trails
  2. Lots of Noise (Colored Speckles)

Gaps in Star Trails

To oversimplify a bit there are four causes for gaps in star trails created from successive exposures:

  1. Camera limitations
  2. Camera or intervalometer misconfiguration
  3. Processing choices
  4. Weather conditions

Camera limitations: I described this issue in my article “How long does a 30 second exposure take?”  All the Canon cameras I own – including the top of the line 5D Mark II require 32.8 seconds to complete a single 30 second exposure. Well there you go: almost 3 seconds of time where there is no exposure. This problem can be compounded by two common misconfiguration blunders:

  • Failing to allow enough time between exposures when using an intervalometer. Or using the wrong drive mode on the camera.
  • Failing to turn off long exposure noise reduction.

To avoid intervalometer misconfiguration I operate in either continuous exposure mode or bulb mode. I use continuous exposure mode when my exposures will be many and a maximum of 30 seconds – e.g. when trying to capture meteors or planning for a time-lapse animation. In continuous exposure mode I set my intervalometer with a start delay and then program an exposure time of several hours… AND I put my camera in Manual, high-speed continuous exposure mode with a typical exposure of 30 seconds. You do not really need an intervalometer for this – a locking cable release is sufficient.

When I operate in bulb mode, I try to get a moderately long exposure. Usually in the 4 to 10 minute range depending on the sky conditions. In this setup it is very important to put the camera in Bulb exposure and program the intervalometer to leave a 3 second gap between one exposure and the next. I have recently discovered, however that the Canon 5D Mark II will work with my intervalometer set to 1 second intervals. That’s goodness. I am still trying to work out whether the problem is due more to the timer or the camera. I do know that in continuous exposure mode all my cameras require 32.8 seconds per each 30 second exposure. Failure to allow a long enough pause between exposures can cause unexpected results.

Photo 1: For the first half of the evening I mistakenly left long exposure noise reduction on. The result was that half of my shots occured at every-other eight minute intervals.

The “dotted lines” in the circle above were caused by leaving on long exposure noise reduction. The result was that the intervalometer timed an 8 minute exposure, waited three seconds and then pressed the shutter for the next 8 minute exposure. However 3 seconds after the exposure completed it was still doing long exposure noise reduction so that cycle was skipped until the intervalometer released the shutter for the next 3 second “off” interval.

I have gotten into the habit of setting my exposure length to 3 seconds less than what I want… e.g. 9:57 for a 10 minute exposure. I then set a 3 second inter-shot interval. I used to set a 10 minute exposure plus a 3 second gap – but the predictability of starting a new exposure every 10 minutes makes it easier to monitor what is going on.

Another cause for gaps: changing the battery. I can offer the following important tidbits when you need to change the battery.

  • Do not wait for your battery to be exhausted. A partial exposure may not stack well or be completely written to your card. Battery exhaustion will likely occur at an inopportune time.
  • Have everything at hand in advance of the change. For example, keep the battery in your front pocket where your body heat will keep it warm.
  • Practice a battery change BEFORE you start your exposures. Only by practice beforehand will you be able to discover that the battery compartment is blocked by your tripod, or impossible to reach, etc.
  • When you DO change batteries beware! Your camera settings may change dramatically!

Processing choices you make when stacking the star trails also affect whether your gaps will be inconspicuous. Do not do any sharpening until you complete your stacking – and even then avoid sharpening the star trails themselves. The method used to stack trails is significant. However, I have observed that people do not notice gaps even in this image of 19 8-minute exposures printed out at 20×30 inches.

Photo 2: Even though it is composed of 19 eight minute exposures the gaps are never noticed even when printed at 20x30.

Weather conditions can also introduce gaps. In a truly dark sky where clouds are not lit by city glow, moonlight or twilight, clouds become “black holes” and block starlight. Low or fast moving clouds can obscure some, most or all of one or more images in the set. This can be perplexing if you happen to be sleeping during exposures which started and ended with clear skies.  Another problem is dew which may form a fog that diminishes or eliminates some or all of the exposures. Vigilance with a rag, the use of a hood or a dew heater are your only weapons against dew.

Lots of Noise (Speckled Colors)

I purposefully left the noise in Photo 1. It’s quite noticeable in the rock silhouette at the lower right and appears mostly as red specs. Annoying? Well, yes, but it is not the end of the world.  In order of effectiveness here are your best approaches to keep the noise manageable:

  1. Shoot at a lower ISO (100 or 200)
  2. Shoot and stack shorter exposures – longer exposures generate more noise.
  3. Capture the foreground and the star trails separately. A better lit foreground will exhibit less noise.
  4. Shoot during colder seasons – lower temperatures result in lower noise.
  5. Control stray light with a lens hood – and close or cover your viewfinder while exposing.
  6. Use high ISO noise reduction
  7. Use noise reduction post processing tools. Chrominance noise is usually most in need of correction.
  8. Use long exposure noise reduction.

Hopefully you noticed that long exposure noise reduction (LENR) is last on the list. If you are trying to stack star trails it is impossible to get continuous trails with LENR on. It is also the least effective unless you are only going to shoot one shot.

Before we go much further, it is worthwhile to note that there are 4 causes of “noise” and each has a different source. The random speckles are usually what is meant by noise. Those random speckles are created by heat, limitations in the electronics, and things as bizarre as electromagnetic phenomenon like sunspots. No kidding. True noise is by nature random and LENR can not do a thing to combat random noise except to diminish it by reducing the luminance of the offending pixels – which also reduces the sharpness of your image. But there are 3 other kinds of noise that are not random though often lumped into the same general category: hot/stuck or degraded pixels, local heat noise (sometimes called amp glow), and high ISO noise. LENR is effective for these because they are not random.

Hot or stuck pixels usually appear as bright pink, red, blue, green, white or purple spots. They are caused by either electronic problems on the sensor chip or by the dyes used to detect the color.  A pixel detects the intensity of the color red by use of a red dye (inkjet droplet) over a sensor site. If that red dye is insufficiently thick, or missing altogether then that pixel location will always read hot if there is any light falling on it – and if the problem is electronic it may read hot even if no light is striking it. Dead or degraded pixels are just the opposite. Too much dye or dead electronics at a pixel site. Degraded pixels are stuck black or darker than the surrounding pixels and are seldom if ever noticed in night photography.

Locally caused heat noise is noticeable in some cameras and is due to the heat of electronics in proximity to the sensor. In my opinion this problem is a design flaw in the camera. However this kind of noise is repeatable so LENR can help correct it. The “Pink or Purple Glow” that results from this flaw was discussed in Part 1.

High ISO noise has an understandable parallel in the world of audio. Take nearly any cheap radio. Turn it up. At some point the sound will become distorted and harsh. This harshness is because there are limitations in the signal, the amplifier circuitry and the speaker used to produce the sound.  Increasing the ISO in your camera is the photographic equivalent of the audio scenario.  At some point amplifying the light measurements made at each pixel makes the noise more obvious.