Topaz Upscale versus Nikon Z8 Pixel Shift for Hi Resolution

I tried out the Topaz Photo ‘upscale’ feature (like Gigapixel), and then compared it to my Nikon Z8 ‘pixel shift’ feature. The upscale option lets you use AI to increase the resolution of a photo, up to 6X.

Camera manufacturers that offer similar pixel-shift capabilities on some of their models include Fuifilm, Sony, Canon, Nikon, Olympus, Panasonic, Pentax, and Hasselblad.

My Nikon Z8 pixel shift feature was configured for 32 shots. This pixel-shifted photo is 11,008 X 16,512 resolution, or 181MB.

I configured Topaz Photo to upscale 2X, or 11,040 X 16,560 resolution. The original raw photo (the first in the 32-shot series for the pixel shift) has a resolution of 5,520 X 8,280 pixels.

I used my Nikon Z8 with the Meike 85mm II SE f/1.8 lens at f/2.8, ISO 500, 1/80s. I was exactly 10 meters from the target. This distance guarantees that the target in the photo (a Siemens Star) will look really bad. The details in the target simply go beyond the capability of the Z8 sensor.

I used NX Studio to combine the raw pixel-shifted shots into a single high-resolution photo, which is then in Nikon’s NEFX format. I converted the NEFX file into DNG format using Adobe DNG Converter.

Locate the Topaz Photo ‘Upscale’ option

As shown above, I opened up my raw-format shot in Topaz Photo version 1.6.1, and then selected “Upscale”. The program automatically ran the ‘RAW denoise’ filter as soon as I added the raw shot to Topaz.

What’s shown above is an extreme crop from the original photo. This camera/lens combination is quite sharp, but the Siemens Star plot at this distance goes beyond what my 48MP camera sensor can handle.

The ‘Upscale’ selected options

I selected the ‘2X’ up-scaling option, and I let the Topaz artificial intelligence select the “High fidelity v2” algorithm, along with its choice for the denoise magnitude, deblur, and compression. There are 9 total up-scaling algorithm options to select from.

Topaz Photo can output the up-scaled photo in DNG format, which can then be edited in almost any other editor.

The original raw-format shot, zoomed to 400%.

My original raw shot (the first in the 32-shot pixel-shift series) is shown above, zoomed in to 400 percent. The colors in the center of the Siemens Star chart photo are caused by aliasing of the camera sensor; the chart itself is pure black-and-white.

The up-scaled Topaz ‘High fidelity v2’ result, zoomed to 200%.

Since the photo was up-scaled by 2X, the photo needs to be zoomed by 200% to get the same size on the screen as the original photo zoomed by 400%.

The photo definitely looks higher-resolution, but the chart center looks a bit weird. The whole photo has that ‘fake’ look about it.

Topaz ‘Standard MAX’ algorithm at 200%

For this chart shot, I like the ‘Standard MAX’ best of all of the various Topaz algorithms. It looks the least like AI.

Topaz ‘Standard v1’ algorithm at 200%

32-shot pixel shift result, zoomed to 200%

It’s pretty obvious that the 32-shot pixel-shifted result is vastly better than the Topaz results. It has true increased resolution, compared to the Topaz artificial intelligence guesses at how to increase the resolution. The aliasing has been completely overcome.

If you need higher resolution than your camera natively provides, pixel-shift is the way to go (assuming your camera has that option). Most camera manufacturers offer pixel shift, but on a very limited number of models.

If you have subject motion, then the advice reverses. Use Topaz in that case, because you only get a smeary mess when you try to combine the pixel-shifted shots. The Topaz resolution results are much better than the original raw shot for several of its algorithms; it just can’t compete with the pixel-shifted results.