Image Resolution versus Increased ISO

Do you lose image resolution as you increase your ISO? Is there a maximum safe ISO that still retains good resolution? Does resolution loss happen before you get unacceptable image noise at high ISO’s?

I had always assumed that image resolution would tank if I shot at high ISO’s, but I never got around to testing that assumption. It’s easy to notice noisy, grainy effects with high ISO, but it would be useful to know what happens with resolution, as well. I am especially interested in knowing if lens resolution test results are significantly impacted when I shoot in dim, versus bright, lighting conditions.

The internet is full of discussions about higher ISO’s causing noise and less dynamic range. Image resolution impacts are rarely, if ever, discussed. Let’s find out if there are more image quality sacrifices being made than you think.

If it’s not obvious, what counts in your photographs is the combined contribution of each element in the image chain. Lens resolution by itself isn’t going to be meaningful if you have image blur or a low sensor resolution, or possibly a high ISO. That’s why this article title refers to image resolution instead of lens resolution.

Increasing ISO will drastically reduce your camera’s dynamic range, but shots ruined by blur are much worse than losing some shadow information or less clean details.

As usual, I’ll be using the MTFMapper program to measure my resolution chart images. Studies have shown that its resolution results are just as valid as programs such as Imatest. The key to good resolution measurements is to use a large, high-quality test chart (besides using a sturdy camera support, careful alignment, and correct focus). To get measurements results that are reproducible by others, it’s also important to use raw-format, unsharpened image files.

The Test

I shot my test chart at ISO 100 through 6400 in 1-stop increments with aperture priority. I left the aperture at f/2.8 in each shot and only changed the ISO (so that the shutter speed would adjust for each different ISO).

I haven’t bothered to show each ISO, because there was so little change between each test. I decided to just show the results at ISO’s of 100, 400, 1600, and 6400. I find quality beyond ISO 6400 to be unacceptable (mostly due to color noise and loss of dynamic range), so I stopped at 6400.

ISO 100 resolution baseline

The shot above is the un-sharpened resolution measurement near the very center of the test chart. The peak resolution is 2088 lines per picture height (Nikon D610). I used my Nikkor 24-70 f/2.8 VR lens at 50mm and shot wide open. The view shown above is at 100% magnification.

ISO 400 resolution

Except for a miniscule change of about 4%, the resolution at ISO 400 is the same as ISO 100. I wasn’t expecting much of a change with this modest increase in ISO, so the results seem consistent with expectations.

ISO 1600

I can barely see a change at ISO 1600. There is some extra image noise, but it’s still mostly ignorable. There’s NO change in resolution from ISO 400. In the old film days, the picture would have already fallen to its knees at this point.

ISO 6400

Image noise is rearing its ugly head at ISO 6400, but note that the resolution is about the same as lower ISO values! I was totally expecting resolution to crumble, but was pleasantly surprised that it didn’t.

I’m not denying that dynamic range takes a huge hit, and color noise is now too objectionable for me to use ISO values like 6400 unless there is no other choice. One thing you don’t have to worry about, however, is resolution loss at higher ISO values.

MTF contrast plot, ISO 100

The MTF contrast plots created in MTFMapper include a shaded band around the plot lines that show the range of individual measurements. You can think of this shaded band width as an indicator of image noise.

MTF contrast plot, ISO 6400

Note the increase in thickness of the shaded bands for the higher ISO, which represent the individual chart edge measurements. You’ll always see this trend at higher ISO values. Also note, however, that the averages (the solid and dashed lines) show very little change at the high ISO.

Summary

Image quality is so much worse beyond ISO 6400 that I stopped testing at that point. I would have thought that resolution loss correlated closely with something like color noise or luminosity noise, but it certainly doesn’t.

It’s comforting to know that my lens resolution test results aren’t impacted by ISO, since tests at f/16 often compel me to increase the ISO. I tend to avoid slow shutter speeds while testing resolution, due to the risk of vibrations affecting the resolution results. By the way, cameras that feature "electronic front curtain shutter" are wonderful in this regard; combining live-view with EFCS and a remote shutter release virtually eliminates vibrations.

It’s a great time to be alive as a photographer. If somebody from the 1970s could see the kind of quality we’re getting today (at prices that are dirt cheap compared to the 1970s, too) I think that they would be absolutely flabbergasted.