I have to be straight up here. I’m writing about digital noise in photos and I am going to tick off some readers who have different perspectives, some of them finely polished and honed by years of incoming BS. I’m not suggesting that these folks are stupid, but that they may have chosen to believe crap from some alleged experts, or are, to be blunt, doing things incorrectly and building skyscrapers out of tiny bits of Meccano.
If you notice the noise, it’s a lousy photograph
That quotation is attributable to my friend Rick Sammon’s dad. It’s about as clear as things can be.
I am a big fan of the work of American photographer Alfred Steiglitz. He did amazing work and I really favour his New York at Night series where he would haul an 8x10 plate camera out into the streets to make images. Given the chemistry available at the time, despite having a massive negative plate, these images are filled with grain, the chemical equivalent of digital noise. You know what? No one complains about the noise. This should encourage us to ask “why not?” Partly in fairness it is because the images are old and not so many people are looking at them anymore, but more importantly, it’s because the images are interesting and filled with the frameworks for stories.
As creatives, we are all storytellers, and if we do a crappy job at telling our stories, viewers are going to start looking around and because they are already subconsciously frustrated, will start to look for what’s wrong. It’s normal and human, as well as being rude. Don’t like something? Fair enough, shut up and move on to more important things. Unless of course you live in a basement and your life is nothing but flaming people from behind an alias on the Internet between bouts of jerking off to Internet porn.
But that’s nobody who comes here.
The more challenging problem with noise is the bad habit exhibited by so many digital creatives these days, and that is over inspection.
Let’s talk about proper viewing distance. The concept is simple. Viewing distance for any work of art is at minimum twice the diagonal size of the image. I am writing this article on a computer with a 27” diagonal display. Basic arithmetic, no longer taught in schools sadly, says that proper viewing distance for an image that is 27 inches diagonally would be 54 inches.
I know of lots of folks with computer displays that are 27 inches diagonally. I know of no one who only judges their images from 54 inches away. Most desks and chairs place the screen about 24 inches from the eye. Yes, I have measured this because quantitative data is useful and assumption is the root of all fuckups. Sometimes at this too close viewing distance, the creator editor may become frustrated because of what they see as too much noise.
The next step taken is not to step back and reevaluate but instead to zoom in on the screen to reveal just how bad the noise is! There’s no way that at some level of zoom that things will not look worse. The math does not work here. Moreover most computer displays are pretty low resolution, as low as 72 dots per inch and very rarely better than 130 dots per inch. A decent print is 300 dots per inch. Are you starting to see a problem here?
Sometimes, this causes people to conclude that the images on their smartphones are better, or at least as good as those taken with their “real” cameras. The truth is that this is not so. Retina displays definitely have higher resolution than a typical computer screen, but they are small screens and these images are rarely displayed on a larger screen or printed so we do not see the actual image quality. That’s ok because the smartphone images are doing what they need to, and in most cases are considered disposable anyway. Simply put, if one wants to get their shorts in a knot about noise, the smartphone image is the fastest way to get that tied because the image quality is actually quite poor, even after the massive amount of processing being done before you see the image.
Another challenge occurs when serious creatives pass judgements based on a JPEG. JPEG formats are destructive. JPEG fine destroys over 70% of the original data BEFORE the first image is ever saved. JPEG normal is over 93% destructive, and this repeats every time you save an image. It’s not hard in three or four saves to make a lovely image look like the cat box.
I suppose I could get frustrated at repeating these scientific facts so often, but accept that I am outnumbered by the sheer volume of alleged experts on the Internet, many of whom could not make an interesting photograph if their lives depended upon doing so. Doing the same thing over and over does not make you better if you are doing it wrong.
Noise reduction software is also far better than it has ever been. Suppose that you have made a RAW image at an ISO that you determine after proper and reasonable viewing of an image that you decide is interesting still has more noise than you might like. Lightroom’s built in noise reduction is one of the best algorithms there is. It works and is subtle. Some noise reduction software creates very contrasty smears, because more contrast creates the illusion of sharpness. That’s bollocks of course, but never let the truth get in the way of marketing. And if you believe any of that AI bullshit, well I expect that I cannot be of assistance to you. See my post on AI here.
As an educator and coach, I have been asked in the hundreds of occasions to help someone with an image that they feel is far too noisy. With proper examination and looking at things as one would a painting, we discover that the offending noise is not so bad at all and if the image is genuinely interesting to a viewer, not seen at all. Sometimes, we are our own worst enemy, determining fault where there is none.
Digital noise does exist. The higher we turn up the sensitivity on our cameras by raising the ISO, the more power is going to the sensor and the signal to noise ratio on output lowers. That’s basic electronics because we are pushing more power. Sort of, but not the same as, why an overdriven guitar amplifier will naturally distort. We can minimize digital noise by using lower ISOs, but perhaps not as low as you think that you need to be. If we are going to regularly be shooting in low light conditions where we cannot supplement the light, we should be choosing cameras with better low light native sensitivity.
Could I have used noise reduction software on the ISO 409600 shot? Sure. Would it have been less noisy. Yes it would and you can see the impact below
Note that the noise reduction did nothing to make the image more interesting or more compelling. This shot was taken a few years ago with a then brand new Nikon D5, the first camera to support a built in extended ISO of over 3 million. Those images looked like hell, but the work made the 10200 ISO images usable. Makers should be working on better low light performance, but most don’t, instead releasing sensors with more megapixels because buyers think that matters. In reality for most use cases, megapixels are utterly irrelevant, but truth doesn’t get in the way of marketing.
What we can learn from this is that getting the best exposure that we can, which is the combination of ISO, aperture and shutter speed is going to have a big impact on how much digital noise there is going to be. Not shooting in JPEG and not editing JPEGs will make even more of a difference, but in the end if the image is not interesting, then fixing the noise won’t matter.
So you might be asking, how often do I use noise reduction tools? The answer is nearly never, and only really when I need to recover as best I can a mistake that I made in the shooting process. Even then though, I only make the decision to use these tools after viewing the image at a proper viewing distance and not zoomed in at all. You will reduce your noise concerns and your post processing time if you do the same. Also if you stop posting anything but your very best to the Internet, there will be less for uninformed troglodytes* to whine about, but the zero value of social media is a screed for another day.
with apologies to any decent troglodytes who may be offended by inclusion
Do you have an idea for an article, tutorial, video or podcast? Do you have an imaging question unrelated to this article? Send me an email directly at email@example.com or post in the comments. When you email your questions on any imaging topic, I will try to respond within a day.
If you shop with B&H Photo Video, please consider doing so through the link on thephotovideoguy.ca as this helps support my efforts and has no negative impact whatsoever on your shopping experience.
If you find the podcast, videos or articles of value, consider clicking the Donation tab in the sidebar of the website and buy me a coffee. Your donation goes to help me keep things going.
I'm Ross Chevalier, thanks for reading, watching and listening and until next time, peace.