Too Much Noise About Noise

How many times have you heard a photographer complain about digital noise, but if it had not been mentioned, you would never have noticed, and if you did notice, it was irrelevant if you enjoyed the image?

The answer is way too freaking often.

What Is Digital ISO

As the ISO setting is increased on a digital camera, more power is applied to each photoreceptor (frequently called a pixel). All pixels have a signal to noise ratio, that is how much noise accompanies how much signal. The relationship between signal and noise changes as power to the pixel increases or decreases. It takes considerably less power to make a photoreceptor responsive to an EV value that would be suited by ISO 100 than ISO 3200.

Indeed, completely unrelated to film photography, in digital the ISO number is simply an indicator of how much power is being delivered to each pixel. Thus an ISO of 3200 for example tells us that each pixel is receiving more power than a setting of ISO 100

In an ideal world we would have accepted that ISO worked for film and there was some other name to describe light sensitivity in digital sensors, but manufacturers worked very hard to equate photoreceptor sensitivity settings to match the chemical receptivity to light of film. That they did made the transition easier for film photographers coming to digital, but for those who have only ever used digital photography, the designation is just another moniker, poorly defined and often in terms of film. Which is factually incorrect.

Digital Noise is Actually Irrelevant

All the whining about noise makes film photographers laugh in their beverage of preference because in the days of film, you could buy film with an ISO of 1000 that had mammoth grain and terrible colour, or black and white film that you could push to ISO 1600 for images that appeared as if they had a layer of sand on them. Photographers at that time learned that grain (how film emulsions responded and having no relationship whatsoever with noise) was a fact of life and a compelling photograph did not concern itself with grain.

I often quote my friend, photographer and educator, Rick Sammon. He shares that his father told him that if a photograph is noisy (grainy) to the viewer, it’s a lousy photograph because the only thing the viewer sees is the noise. This is such an accurate statement, it should be taught on day one of any photography course, but is most often ignored, and frequently willingly by companies that want to sell you something by making you insecure about your work. They cannot actually make you insecure, that is a personal thing, but they do find effective triggers.

Signal to Noise Ratio

Thus as we increase the ISO setting to variables unattainable a couple of decades ago in film and even ten years ago in digital, we are choosing to push more power to each photoreceptor. As the photoreceptor receives more power its signal to noise ratio falls. Simply this means that the amount of noise generated at ISO 1000 is greater than the noise generated at ISO 100 as any pixel peeper will gleefully tell you. Fortunately pixel peepers while common in photography are most non-existent in reality and normal folks do not encounter them and certainly do not have to hear their braying, foolish donkeys that they are.

A signal to noise ratio as noted represents the ratio between the signal from the photoreceptor (call this the good stuff) and noise from the photoreceptor (call this the bad stuff). In the chart below, courtesy of and property of DXOmark, we see the signal to noise ratio of three different cameras. You can see how the signal to noise ratio drops off as ISO increases. (Be aware that an ISO 50 film performed in terms of grain at the same level as one of these digital sensors does at ISO 6400 under optimal conditions). What the chart tells us, and DXOmark is very conservative in their gradients is that with any of these sensors, you will get usable results at ISO 25600.

Copyright DXOMark

As photographers and as videographers, we have the ability to make images and clips at light levels never before usable. This is absolutely marvellous. While it is popular to say that a digital image made at ISO 3200 on a sensor released 10 years ago is more noisy than a digital image at ISO 3200 on a new sensor released a year ago is true, the statement leaves out specific data, so adhere to it with a five pound bag of salt. Check your particular sensor to see if it is valid. The most important measurement being, can you see it at normal viewing distance and if you can, does it matter? I will come back to Rick’s father’s statement. If you see noise in an image as being dominant, it’s a crap image.

Some five years ago, I was a photographer for high school football in my town. In Canada amateur sports does not get the level of attention or fan investment that is found in the United States. In Texas there are stadiums with professional lighting dedicated to high school football. In Canada, the teams are lucky to get a field with any lights at all. In order to make images of championship games, played at night, long after sunset, to maintain appropriate shutter speeds and even with marginal depth of field, I was shooting at over ISO 21000 on a Canon 1Dx. To the parents, players and coaches, the digital noise was irrelevant when I did my job and produced interesting or compelling images. When I failed to do so, no one saw those images except me as they made their way to the digital trash can. DXOmark said that for low light sports, the sensor was best at ISO 3200, however in practice, ISO 21000 worked out at a more than acceptable level. At ISO 25600, DXOmark noted that this was the maximum acceptable ISO before noise took over. I concur and have the shots to prove it. That sensor from 2014 performs as well at high ISO as the newly released Canon R8, or my recent R5 purchase.

I recently did a seminar on Noise Reduction and used modern noise reduction techniques on some of those images. This included the more traditional noise reduction tools including the modern, marketed as AI but really just sample infusion, tools. While there was a reduction in digital noise, it also brought a reduction in sharpness. This is expected as noise reduction is the enemy of sharpness. That I could see the loss of sharpness on a high end 4K display (all displays have fairly shitty resolution - this one being 4K still only delivered 163 pixels per inch) informed me that the use of noise reduction would compromise the overall quality of the image and its value to the viewer. I did take it to the point where I used masking not to apply noise reduction to the jerseys and only to the background and then when I share the original edited image beside the edited image with the noise reduction applied, the non-photographer viewer only said that the numbers on the uniform were less crisp than the numbers of the photo on the left (the original edit). The display was a 27 inch diagonal 4K display and the viewing distance at 1:1 was about two feet. This was well inside the proper viewing distance of about 40 inches for that size display. At the proper viewing distance, the additional noise reduction did not make any difference to the viewer.

I find the psychological impact on the uninformed and insecure of all this noise reduction bullshit extremely troubling. Folks making images that they love are being convinced that they are lousy by marketers telling them that without their secret squirrel noise reduction the images will suck, internet pundits say the same thing, and pixel peepers who pounce on images and zoom them to 300% and then inspect them from two inches away followed by their whining at length all damage the creative process.

Of course if you have the sense to know that the only opinion that matters is your own, and you do not become infected with the pixel peeping disease, you would spend less time editing, perhaps more time making images and certainly have a better time.

Consider this alternative. Don’t spend money on dedicated noise reduction software. Avoid any software that calls itself AI that is really just replacing your image data with image data stolen from another photographer without their knowledge and just go out and challenge yourself to make images that please you. An image that does not please you cannot please any other rational viewer, and what value do the opinions of the irrational have anyway?

There’s too much noise about noise reduction. Be your own person and do your own work and if all the whiners were to die off and vanish, no one with any intelligence would miss them at all.

Thank you for reading. Please subscribe to be notified of new articles and podcasts. if you shop at B&H Photo Video, I ask that you do so through the links on the site as your purchase experience is the same and I receive a tiny commission that helps me keep things going here. Until next time, peace.