Noise mitigation for OSC (One-shot Color) Cameras
We all know that SNR is the Signal to Noise Ratio, and that more signal and less noise is good, right? OK, if you didn’t know that, you do now.
So we’re done, right? Of course not. SNR is what defines an astroimage’s quality. A higher SNR produces a beautiful, smooth, detailed image. A low SNR produces an image that looks like it was taken with a cellphone in a cave. When you first start imaging, your pictures will probably be closer to the cellphone on the SNR scale, and then gradually slide over to the aesthetically pleasing and awe-inspiring images that made you get into this hobby in the first place. My objective is to make pretty pictures and I need to get rid of or mitigate as much noise as possible. I think my slider is about half way after a bit more than a year of imaging. I’ve come to a point where I have to advance equipment to further move the slider, but more on that later.
I want to be careful about saying that we’re eliminating noise, because it’s more nuanced. In some cases we’re avoiding it, some we’re hiding it, and in other cases we’re just getting more signal to mitigate it. The ultimate goal from a light polluted site like my back yard is to jack up signal and make as much of the noise as possible unnoticeable.
OK, so what is noise?
Craig Stark, from Stark Labs (makers of PHD, Nebulosity) has some great YouTube videos about SNR and how to get the best ratio you can, but it’s kind of a leap forward for a beginner, and very technical from the start. As well it should be, because Craig’s target audience at those presentations is mainly semi-experienced to very experienced imagers.
For our immediate purposes here, we’ll define noise as bad information in any given pixel that interferes with the signal of the object you’re trying to capture. There are three main sources of noise in Astroimaging.
- Read Noise – Inherent dark current noise
- Thermal Noise – duh
- Shot Noise – Mostly affected by Skyglow and atmospheric conditions in astrophotography
Read Noise is the kind of inherent background noise present on every sensor. This is what we take BIAS frames for. That’s me telling you not to skip your calibration frames. I almost never use less than 100 BIAS frames when stacking. They are quick and easy to take and you don’t have to take them very often. Some sensors are less noisy than others, and some respond better to cooling than others.
Thermal Noise is just what it sounds like. The hotter your sensor gets, the noisier your data will be. This is why our dark frames need to be the same temperature. There is no good way to beat this when nights stay warm other than having a cooled sensor, or only shooting in the winter. Last Summer I had excellent skies and countless clear nights. However, it was consistently in the 70’s (20C) overnight and my sensor was nearly 100 (35C) degrees, causing so much noise that I pretty much wasted my time. I am such a cheapskate, but the noise is so unbearable during the Summer that I have invested in a cooled astronomy specific camera. It is a pricey investment, but it’s either I purchase a specialized camera, or build a crude, heavy, inconsistent and inefficient cooling box for my DSLR.
A detailed review of the camera I chose (ZWO ASI071MC Pro) will come once I’ve had chance to use it a bit more.
Shot Noise is what filters are made for – at least for people like me trying to image a couple miles from O’Hare airport and 500 yards from an eight-lane interstate highway! Skyglow via light pollution is going to be your biggest enemy here. Shooting a monochrome camera and narrowband filters is the most effective method to eliminate as much of this type of noise as possible. I have three kids, a wife and two dogs. I image often times in short windows between trees. I’m not doing mono imaging at this time in my life. For now, a one shot color camera is for me. There are Dual, Tri, and Quad narrowband filters for color cameras with very narrow bandpasses for Ha, Hb, OIII and S2, but they run from $380-$1100US. I would love to slap a 48mm quadband filter on my ZWO ASI071MC-Pro and go to town, but that’s a $2500+ dent in my already banged up wallet.
See below for bandpass information.
OK, so what filter is right for you?
A lot of targets we’re trying to image transmit photons from very specific wavelengths of light. Galaxies aren’t in this category, but all of those bright detailed red thingy’s are. All of the extra noise ruining your images is coming from the wavelengths in between the ones you want to collect. There are a few types of light pollution filters that I’ll go through here from boadbandiest to narrowbandiest. The more light polluted your site, the narrower you want your filters. All of the skyglow and pollution noise lives in between those few wavelengths that have meaning to the detail of your images.
L-Pro/UHC/LPS-D1, etc. These style filters are designed to block out the most commonly used Mercury and Sodium street lights. These will allow for the most natural color transmission, but are really for low to light-moderate light pollution. There are exceptions with methods and equipment to use these filters in strong light pollution, but that’s another discussion. This is the only type of light pollutions filters that I’d consider acceptable for imaging galaxies. The bandpasses are broad enough that almost the full spectrum a galaxy puts out will be received – meaning the blues in Andromeda and M51 can come through.
CLS/CLS-CCD (City Light Suppression) These filters aim to drop out everything in the middle of Ha and OIII/Hb, but the bandpasses remain somewhat wide enough to make color correction almost possible in post processing. You’ll be hard pressed to get an accurate color representation for galaxies. For example, you can get great detail on M33, but it will be toward the red side of purple. These are good for moderate light pollution.
IDAS NB1 –This kind of falls in between a CLS and Tri or quad-band narrowband filter. The cutoffs are much narrowbandier than a CLS, but still leave more room than a traditional narrowband filter. Why not go all the way, you say? Well, this type of filter is $239US, and the quad-band narrowband is $1075US. That’s why. With the NB1, you will increase your SNR significantly while still using all of the pixels (colors) in your OSC on each shot and leave your wallet a little heavier. The NB1 is good for nebulae in fairly heavy light pollution.
Dual, Tri and Quad-Band Narrowband Filters. These filters are very specialized and cut down to only two, three or four wavelengths within a few nanometers. The more bandpasses. the more expensive. The dual from STC coming in at $380, the tri from Radian/OPT at $725 and the quad also from Radian/OPT at $1025. These filters are new, but the potential is exciting. I’m sure the price will come down and the tech further developed to produce some amazingly high SNR images from urban locations.
The purpose of all of these filters is to get you more signal and “less” shot noise per frame. There are additional software aided imaging methods to reduce noise, but this is for idiots and about equipment. You can click here to read about dithering. Dithering is a game changer and necessary, but works best when you are starting with a proper base.
For those of us in urban to suburban environments, our objective is to mimic a rural site. Well that’s not technologically possible (yet). Our only option for using a color camera is to filter out as many of the bad wavelengths as possible and hope we can pull back some somewhat natural star colors in processing – I still haven’t nailed this. The narrower the bandpasses get on a well made filter, the more you isolate transmission lines of the object you want while removing the noise-causing skyglow and pollution in the middle.
Getting Your Best SNR Value
We can’t eliminate noise, but we can counter it by boosting signal, avoiding parts of it, or removing it in processing. This whole process can become extremely technical with a heckuva lot of math, but if you’re starting out and just want better pictures, you don’t want to see more math! So take these four pieces of advice to start smoothing out your images.
- Take as many images as you can and stack them. Stacking images allows the stacking program to recognize noise and cancel it out while adding, or stacking, your signal. The quick and dirty recommendation is to take as many images as you can. There is a point of diminishing returns, but at this point I’d recommend getting everything you can and stacking it in DeepSkyStacker.
- Dithering. Dithering is a process that moves the scope a few pixels in between images. The process itself doesn’t remove noise from the exposure. Instead, it randomizes the noise between exposures allowing it to be mitigated when stacking. You can read more about this here.
- Take your calibration frames! Dark and BIAS frames are absolutely necessary for removing thermal and shot noise respectively. When you stack your calibration frames with your light frames noise is removed frame by frame. Without these, you’ll find processing to a smooth image nearly impossible.
- Choose the right filter for your target. If you live in the dry desert, a hundred miles from a city or mountains above light pollution, this doesn’t apply as much. But if you’re like most of us, you have some level of light pollution and sky glow. The more pollution you have, the narrowbandier your filters need to be. Galaxies require a broader spectrum to capture the right colors, and nebulae can get down very narrow as your aiming to capture Ha, Hb, OIII and SII bandpasses. This can get pricey, but it’s definitely necessary to capture high contrast images near a city.
Noise is our enemy, turning smooth detail of beautiful nebulae into grainy fuzz that is harsh to the eye. Hopefully these tips help you reduce the visible noise in your images and move forward in producing the images you want to see. The reduction of noise can get very complicated and technical. For now, get these few things dialed in and then you can advance through the math of noise to starting mitigating whats left.
Clear Skies, Bleary Eyes – KA