To evaluate the image quality of the Fujifilm XT2, I’m presenting two sets of results, one for resolution and one for noise. I’ll start with the resolution of a daylight landscape scene, and I have two sets of comparisons. The first is a new comparison made against the full-frame Canon EOS 5D Mark IV. I know they’re completely different cameras aimed at different photographers and at very different price points too, but I’ve received a lot of requests to compare them, so here you go. Below this I’ll repeat my earlier test comparing the Fujifilm X-Pro2 against the XT1. Since the XT2 shares the same sensor and processor as the X-Pro2, the quality results from the latter are applicable to the XT2, and to make my latest comparison more useful, I took care to shoot with the same lens from the same position and take crops from most of the same areas. So first up, the XT2 vs the EOS 5D Mark IV.
I took the scene above with the Fujifilm XT2, fitted with the XF 35mm f2 prime lens, set to f5.6; for the EOS 5D Mark IV I fitted it with the EF 24-70mm f4L IS USM lens, set to just over 55mm f8. The aperture values chosen represent sweet-spots for quality on both lenses. As for the choice of lenses, these were the ones supplied by the manufacturers during my loans and better than any other alternatives I had to hand. I would have preferred to use zooms on both or primes on both (although getting the field-of-view to exactly match with primes would have been difficult). But in terms of quality, both lenses used are very good and at the aperture selected I don’t believe either are holding the bodies back, especially towards the centre of the frame. If you think this is an unfair test, scroll down to my X-Pro2 vs XT1 comparison where both share the same XF 35mm f2 lens.
At the time of writing, Adobe hadn’t yet released support for the 5D Mark IV RAW files, so I’m presenting a comparison of JPEGs straight out of camera; I will update this page with RAW results once the 5D IV is supported in Adobe.
So in the left column you’re looking at an APS-C body with a 24 Megapixel X-Trans III sensor and on the right you’re looking at a full-frame body with a 30.4 Megapixel CMOS sensor.
Judging from the crops below I’d say both are resolving a similar amount of real-life detail – certainly there’s no visible benefit here to the extra 6 Megapixels of the EOS 5D Mark IV, nor its larger sensor, at least for low ISOs on the conditions of the day. If anything, the Fujifilm XT2 crops in the left column are a little crisper and more contrasty, due no doubt to Fujifilm’s in-camera processing defaults when using the standard Provia Film Simulation. I’m not surprised as Canon often reigns back the sharpening on its JPEGs using the Standard style, so once I get a chance to process the RAW files with the same settings, I’m expecting a similar playing field in terms of visible sharpness.
So an approximate draw here on the first page, but the really interesting test is on the next page where I’ve compared the noise levels of the XT2 against the 5D Mark IV, and also included Sony’s A6300 for good measure. So scroll down for my Fujifilm XT2 noise results or skip to my Fujifilm XT2 sample images! But first to see how the XT2 compares against the earlier XT1, scroll down for my X-Pro2 vs XT1 test.
Fujfilm X-Pro2 vs XT1 quality
I’m republishing my earlier X-Pro2 vs XT1 quality comparison below as the XT2 shares the same sensor, processor and image quality as the X-Pro2, so the results are applicable to both models. To compare the real-life quality between the Fujifilm X-Pro 2 and the Fujifilm XT1, I shot this scene with both cameras moments apart using the same lens, the XF 35mm f2 set to f5.6 for optimal sharpness. Both cameras were also set to their base sensitivity of 200 ISO and metered the same exposure. I’m comparing JPEGs here straight-out-of-camera. Note the XT2 shares the same sensor and processor as the X-Pro2, so the results here are applicable to that model.
Below you’re comparing 100% crops from the new 24 Megapixel X-Trans III sensor of the X-Pro 2 in the left column against the old 16 Megapixel X-Trans II sensor employed by the XT1 and many previous Fuji X bodies. Both sensors share the same APS-C area, so in order to squeeze 50% more pixels into it, the X-Pro 2 has simply made them smaller. An increase of 50% in total pixels sounds like a lot, but in actual linear resolution the boost represents a little over 20% horizontally and vertically. The old sensor sports 4896×3264 pixels compared to 6000×4000 pixels on the new one.
So what does the boost in resolution get you in practice? Judging from the crops below, you’ll only notice it in the very finest details. Look closely and you’ll see the X-Pro 2 is certainly resolving finer detail than the XT1, most notably in fencing, railings, grills, brickwork and the lettering on distant signs, but in other areas it’s hard to tell the difference. The good news though is the extra detail is there if you look for it, and the unique colour filter array continues to do a good job at avoiding moire even without a low pass filter.
The bottom line is the X-Pro 2 maintains the satisfyingly crisp, moire-free output of previous models while resolving a little more detail, which should keep the Fuji faithful happy. Once again the increase in detail isn’t huge, but it’s there given the right subject and should allow you to output at larger sizes before the quality is compromised. Numerically at 300dpi, the X-Pro 2 lets you to output 2in taller and 3in wider than older models with the X-Trans 2.
Scroll down for my Fujifilm XT2 noise results or tab to my Fujifilm XT2 sample images.
Fujifilm XT2 noise
To compare noise levels in low light I shot the following still-life arrangement with the Fujifilm XT2, Sony Alpha A6300 and Canon EOS 5D Mark IV at each of their ISO sensitivities. I fitted each body with a high quality prime lens: the Fujinon XF 35mm f2, Sony FE 35mm f2.8 and Sigma ART 50mm f1.4. On their APS-C bodies, the first pair were capturing an equivalent field of view of 53mm, so to match the field of view with the full-frame Canon at 50mm, I positioned it slightly closer.The two APS-C bodies were set to f8 and the full-frame to f11, and each sensitivity used exactly the same exposure. I disabled Auto lighting Optimiser on the Canon and Dynamic Range Optimiser on the Sony and used their standard picture styles and Auto White Balance. I shot all the images in RAW+JPEG mode and as soon as the Mark IV is supported in Adobe Camera RAW, I’ll present a comparison using identical processing settings. For now though, I’m presenting a comparison using the out-of-camera JPEGs, with the cropped region indicated by the red rectangle on the main image below; each is presented in the table below at 100%. Many thanks to Park Cameras in the UK for their support in making this comparison.
In the table below you’re looking at 100% crops made from two bodies with 24 Megapixel APS-C sensors and one with a 30.4 Megapixel full-frame sensor; from left to right, the Fujifilm XT2 (24 Megapixel APS-C X-Trans III), Sony A6300 (24 Megapixel APS-C Exmor CMOS), and Canon EOS 5D Mark IV (30.4 Megapixel full-frame CMOS). The Canon offers extended sensitivities of 50 and 102400 ISO, but I’ve left them off this comparison and just used the full range of the Fujifilm and Sony bodies from 100 to 51200 ISO.
The first thing to note is the Sony A6300 captured slightly brighter images which, given identical lighting and exposures, suggests it’s slightly more sensitive than the other two at the same quoted ISO values.
Looking beyond the brightness to the detail in the out-of-camera JPEGs and they’re all a lot closer than you might think, at least in the low to mid ISO range. Indeed I’d say they’re essentially delivering the same degree of real-life detail within this range and the only real difference – minor though it is – regards processing styles at the default settings.
At 3200 ISO and above, the impact of noise becomes more apparent on all of them, whether through softened details, increased speckles or smearing. At 3200 and 6400 ISO I’d say the Sony A6300 and Canon 5D IV enjoy a minor edge over the out-of-camera JPEGs from the Fuji XT2, but again it’s pretty subtle.
At 12800 ISO, there’s noticeable smearing in areas of flat colour, most notably resulting in lost detail in the leafs. Interestingly I’d say the Fuji now enjoys the edge over the Sony, but I wouldn’t cite the Canon 5D IV as being decisively better than either of the APS-C models, despite the benefit of a larger sensor. Sure there’s arguably a tad fewer noise speckles on the crops at 25600 and 51200 ISO, but nothing really significant. This might surprise you if you were assuming a definitive lead from full-frame over APS-C, but modern sensors keep getting better and the latest generation from Fujifilm and Sony are certainly very good.
Of course we’re not just comparing sensors here, but out-of-camera JPEG processing too using the default settings. I shot this scene in RAW+JPEG mode, so as soon as Adobe supports the EOS 5D Mark IV, I’ll be able to make a comparison with identical processing to level the playing field. But for now in terms of out-of-camera JPEGs, all three cameras are performing similarly in the 100-1600 ISO range with only minor differences beyond there.
Next skip to my Fujifilm XT2 sample images, or scroll down for my earlier X-Pro2 vs XT1 noise comparison, the results of which apply to the XT2.
Fujifilm X-Pro2 vs XT1 noise
I’m republishing my earlier X-Pro2 vs XT1 noise comparison below as the XT2 shares the same sensor, processor and image quality as the X-Pro2, so the results are applicable to both models. To compare noise levels under real-life conditions, I shot this scene with the Fujifilm X-Pro 2 and Fujifilm XT1 at each of their ISO values. I fitted both cameras with the same XF 35mm f2 lens set to f5.6. The full view is pictured below with the red rectangle indicating the area I cropped; note I’m comparing out-of-camera JPEGs here.
Below you’re comparing 100% crops from the new 24 Megapixel X-Trans III sensor of the X-Pro 2 in the left column against the old 16 Megapixel X-Trans II sensor employed by the XT1 and many previous Fuji X bodies in the right column. Both sensors share the same APS-C area, so in order to squeeze 50% more pixels into it, the X-Pro 2 has simply made them smaller. The old sensor sports 4896×3264 pixels compared to 6000×4000 pixels on the new one, so the difference represents a boost of a little over 20% in linear resolution. You saw on the previous page how this allowed the X-Pro 2 to slightly out-resolve fine details compared to the earlier models, but does the smaller pixel size compromise noise and tonal range? The crops below should help answer that question.
First things first: the shots here were taken under constant artificial lighting and the shutter speeds were matched for each ISO value. Judging from the crops below I think it’s apparent the X-Pro 2 is once again slightly out-resolving the XT1, most noticeably in the fine veins on the leafs. But there’s also a slight difference in brightness, with the X-Pro 2 crops looking a little darker than those for the XT1. It’s very subtle, but implies the older XT1 is fractionally more sensitive at the same reported ISO values. The difference could also be down to processing styles, so warrants further investigation.
In terms of visible noise I’d say both cameras are fairly well-matched at the pixel level when viewed at 100% up to 1600 ISO – and up to this point they’re both delivering pretty clean output too with lots of detail. At 3200 and 6400 ISO some noise becomes visible on both cameras although I’d give a very slight edge to the earlier XT1 but the situation reverses at 12800 ISO and above; that said neither look great at this point.
So despite that minor drop in comparative sensitivity (or processing difference), I’d say Fujifilm has got away with increasing the pixel count and devoting more of them to phase-detect AF duties without compromising noise levels compared to the older models with X-Trans II sensors. So while the X-Pro 2 doesn’t improve noise at high ISOs compared to the XT1 in this test, it’s no worse at the pixel level that always comes as some relief when the resolution is increased.
Next scroll down for a comparison of movie quality or head over to my Fujifilm XT2 sample images.
Fujifilm XT2 vs Sony A6300 vs Canon EOS 5D Mark IV movie quality in 4k
To compare 4k movie detail and noise levels in low light I shot the following still-life arrangement with the Canon EOS 5D Mark IV, Fujifilm XT2 and Sony A6300. It may seem odd to compare a full-frame DSLR against a pair of mirrorless cameras with smaller APS-C sensors, but the 5D Mark IV actually takes a crop when filming 4k that’s roughly similar in size to APS-C. So with similar sensor areas devoted to 4k capture and similar field-reductions as a result, I feel it’s a valid and important comparison to make. I matched the vertical field of view on each camera then filmed clips under (UK / 50Hz) artificial light at 25p, adjusting the white balance manually to 3200K. The picture profiles were set to their defaults and any noise reduction or tonal enhancements disabled. I then took frame-grabs of each clip using VLC and have presented crops at 100% in the table below.
Each camera may be devoting a roughly similar sensor area to 4k capture, but with quite different numbers of pixels within it. The 5D Mark IV for example is using a 1:1 crop with no scaling, whereas both the Fujifilm XT2 and Sony A6300 are having to down-sample a higher resolution to generate a 4k frame. In the absence of RAW recording on all three or the ability to output 4k over HDMI on the Canon, I’m having to compare footage that’s already been processed in-camera using the default picture styles and profiles. Indeed judging from the crops below, the biggest difference between them regards their processing styles.
Starting on the left column, the Fujifilm XT2’s standard Provia Film Simulation is applying the highest contrast and sharpening for the punchiest result. In the middle column, Sony’s A6300 arguably has the mildest style of the three, while on the right, the Canon EOS 5D Mark IV has the softest style. In terms of actual detail recorded, I’d say all three are roughly in the same ballpark, although most graders would want to choose a calmer profile on the XT2 and apply greater sharpening to the Canon.
While all three start with roughly similar detail at 800 ISO, by 6400 ISO the noise levels have increased to a point they need to be dealt with. The Fuji and Canon clips become softer, while the Sony A6300 adopts a more hands-off approach with more visible noise but greater retained detail as a result. Indeed I’d say the A6300 delivers the best result here, but that for most of the range all three bodies are pretty close to each other.
This is an interesting result as it proves there’s little to no benefit of having the 5D Mark IV’s full-frame sensor when it comes to filming 4k video. The small crop effectively brings it in-line with APS-C models in terms of 4k quality, and while models like the Sony A6300 may lack a touch-screen and headphone jack, they make up for it with an articulated screen, the chance to film and review using the viewfinder, log profiles and 4k output over HDMI. Plus the small matter of being considerably smaller, lighter and cheaper.
But the 5D Mark IV does use its full sensor width when filming 1080p, so does that mean it might enjoy an advantage on noise levels over the APS-C models? To find out I filmed a second set of clips in 1080 / 25p and have presented crops in the second table below. Scroll down to check it out, or tab over to my sample images.
Fujifilm XT2 vs Sony Alpha A6300 vs Canon EOS 5D Mark IV movie noise in 1080p
To compare 1080p movie detail and noise levels in low light I shot the following still-life arrangement with the Canon EOS 5D Mark IV, Fujifilm XT2 and Sony A6300. I matched the vertical field of view on each camera then filmed clips under (UK / 50Hz) artificial light at 25p, adjusting the white balance manually to 3200K. The picture profiles were set to their defaults and any noise reduction or tonal enhancements disabled. I then took frame-grabs of each clip using VLC and have presented crops at 100% in the table below.
When filming 1080p, all three cameras use their full sensor widths and scale the pixels to generate a 1080 frame. But while the Fujifilm XT2 and Sony A6300 start with APS-C sized sensors, the Canon 5D Mark IV starts with a larger full-frame sensor. Since the original sensor resolutions aren’t too far apart, does this mean the Canon enjoys an advantage on noise levels? Let’s find out!
Judging from the crops below, each camera’s default processing is playing a significant role with Fujifilm again looking crispest and Canon the softest, but I’m confident that selecting more neutral profiles and processing to taste would deliver similar styles. Despite the greater sharpening and contrast to start with though, I feel the Fujifilm XT2 is actually capturing fractionally higher detail than the other pair, and I also fail to see any significant benefit in cleanliness to the Canon.
So for 1080p I’d give the Fujifilm XT2 the nod here, although again all three are pretty close. What it does however illustrate again though is a modern APS-C mirrorless camera can deliver essentially the same movie quality as a modern full-frame DSLR – at least in terms of resolution and noise levels. Of course there are perspective and depth-of-field differences to weigh-up when filming 1080 on the Canon as it does exploit the full sensor area, but again it’s revealing how the detail and noise aren’t decisively better.
Next tab over to my sample images or back to my verdict!