Image quality

Fujifilm GF 110mm F2 | ISO 100 | 1/125 sec | F9
Photo by Carey Rose

The GFX 100's image quality is ridiculously good, as you might expect from the largest modern sensor we've yet encountered.

Key takeaways:

  • Excellent image quality with impressive resolution in addition to the tonal quality and increased DR its larger sensor should bring
  • Superb JPEGs offer an alternative workflow in circumstances that demand fast proofs or don't require extensive image manipulation
  • Image stabilization increases the likelihood of realizing camera's full potential but doesn't particularly extend range of usability
  • Good dynamic range but with usability limited by phase-detection banding artifacts
  • No demonstrable benefit to 16-bit files over 14-bit Raw output

The 100 megapixel resolution of the GFX 100 very quickly highlights any shortcomings in your lenses or technique. To get the maximum sharpness at the center of the test scene meant allowing the corners to drop-off a little: stopping down brought the corners up, but resulted in reduction of center sharpness. That this large format system is pushed by the challenges of 100MP suggests it'll be difficult for smaller formats to follow the Fujifilm's path.

The most obvious distinction between the GFX 100 and its rivals is that it captures a higher level of detail. The Phase One XF 100MP (which has a larger sensor) appears to capture similar levels of detail but exhibits more false color. However, the Phase One was processed with different software so we can't be certain how the demosaicing algorithms or levels of applied sharpening compare. The Panasonic S1R is able to capture more detail using its high-resolution feature but only in circumstances where it's practical to shoot eight consecutive images in electronic shutter mode - with a tripod.

The GFX's larger-than-full-frame sensor also means lower noise at any given ISO, which will tend to be perceived as improved tonal quality and gradation at low ISOs. Matching the depth-of-field by opening the FF camera up by 2/3rd EV helps to narrow this difference but only in situations where you can also drop the ISO by 2/3EV. This initial Adobe Raw color profile isn't as saturated as usual, but there's no reason to think that the camera's underlying color response is any different to the 50MP models.

While JPEGs may not be core to the workflow of every medium format shooter, the GFX's are eminently usable. The color is attractive, the detail level is very good and the noise reduction strikes a well-judged balance between noise suppression and detail retention. Having a JPEG engine that produces attractive results is certainly no bad thing.

Image Stabilization

Putting 100 megapixels onto a sensor is one thing, but being able to repeatedly capture images that exploit that full resolution is quite another. Autofocus is one critical component, a low-vibration shutter is another (we saw tiny amounts of shake at some shutter speeds, even in electronic first curtain mode). But the GFX 100's main support mechanism for ensuring high-res consistency is its image stabilization mechanism.

The GFX 100's image stabilization system genuinely helps you get the most out of the camera - though it still took a handful of shots to get a result this sharp at such a slow shutter speed. Processed in Adobe Camera Raw 11 with minimal noise reduction applied.
Fujifilm GF 45mm F2.8 | ISO 800 | 1/15 sec | F2.8
Photo by Carey Rose

Fujifilm claims 5.5EV of stabilization (in the pitch and yaw directions with the 63mm prime), but we wouldn't focus too much on this number.

Both our usage and testing support the idea that IS is an essential feature for offering 100MP on a camera. You'll still need to think carefully about your shutter speeds (don't imagine yourself shooting at or below 1/focal length with much success) but, combined with a high shutter speed and good technique, IS significantly increases the number of 'perfectly sharp' images you'll get. And, as we'll see with autofocus, 100 million 'perfectly sharp' pixels is a challenging target.

We didn't so much find it hugely extending the operating envelope of the camera, but we did find that it helps raise the likelihood of you getting the most out of the camera, within the expected constraints.

Dynamic Range

The use of a modern BSI CMOS, dual-gain sensor means the dynamic range is excellent. The widest possible dynamic range is at its base ISO (100) but if you have to shoot in a situation that would usually require a significantly higher ISO, you'll be better-off raising the ISO to 800, where the camera moves up to its second gain step. This has less absolute DR but lower noise.

ISO invariance: Despite this, we don't see much difference in terms of noise, between shooting at ISO 6400 and giving ISO 100 the same exposure and brightening the resulting file. This indicates that very little noise is being added by the camera's circuitry, since adding amplification doesn't show in a significant improvement.

Exposure latitude: Lifting the deep shadows (as you might if you'd lowered your exposure to protect the highlights but don't want your shadows completely clipped to black) shows less noise than we saw in the GFX 50R but with additional impact being seen from the phase-detection banding. As you'd expect, there's a slight benefit in deep shadow noise over its full-frame peers, but banding in the GFX files means that in use the difference is minimal.

PDAF banding

As we've seen in a number of cameras, there are slight imperfections in the way the rows of pixels with phase-detection elements in are incorporated into the image. PDAF pixels are partially masked, so receive less light than regular pixels. This difference needs to be accounted for when rendering the image, and any slight rounding error can become visible if you try to brighten the dark tones of the image (which is an inherent part of trying to exploit a camera's full dynamic range).

Whether you ever use this much of the camera's DR will affect if you ever notice it. There's also a chance that future improvements in processing software will correct for the differences better. For now, though, the PDAF bands beginning to become visible is likely to act as a limit on how much dynamic range is actually usable from the camera.

16-bit vs 14-bit

The GFX 100 is one of the first cameras we've tested to have a sensor with a 16-bit ADC readout mode and the ability to retain all this data in 16-bit Raw output. However, the main benefit of additional bit-depth is the ability to retain the deep shadow information of its images with greater precision (an area of the image with the lowest signal-to-noise ratio, meaning much of this information risks being noise).

Most brighter regions of an image are already over-encoded, meaning that they're already encoded with more precision than can actually be distinguished between, at the point of capture.

Comparing the deep shadow regions of 14 and 16-bit Raw files, brightened as if to be included in an image. We cannot find a visible difference, and mathematical analysis of the Raw data shows a <0.1EV difference. These are the tones several stops below the point the clip to black in the JPEG and hence represent the additional dynamic range available to you.

Our testing shows there to be less than a 0.1EV difference in DR between 16-bit and 14-bit mode, because the camera's pixels aren't producing a signal that warrants the additional encoding precision. So while there is, in principle, a <0.1EV difference between 14 and 16-bit modes, this will only exist at base ISO and will only come into play in tones already impacted by the PDAF striping. In short: we could not find any reason to use 16-bit mode.