Strange detail smearing with Fuji XE-2 sensor?

[albireo]

A longtime Nikon user, I have been contemplating the purchase of a small mirrorless setup for a while. the Fuji X-system of course attracted my attention - it seems like a good compromise of features, weight, lens range and sensor characteristics. So I borrowed a colleague's XE-2 and took a few quick sample snaps to see how the output compares to the Nikon's.

I was quite surprised to see, at 100% magnification, on a good number of my test pictures, a curious pattern. It looks like a soft smearing of detail in busy parts of the photo. In my mind it looks similar to one of those Adobe Photoshop creative filters that appear to blur fine detail for creative effect.

Has anyone noticed the same on their Fuji XE/XT photos?

A few details. I only tested in camera jpegs - no raw. The camera is running a 4.01 firmware, which I'm told is the latest and greatest, and brings its performance and features on par with the X-T10. The lens was a 27mm f/2.8. Disappointed by the first batch of photos, I trawled through the menus and, thinking it was some kind of oversharpening issue, I changed the NR to -2. There seems to be no way of deactivating the Noise Reduction altogether on this particular model. The issue persists, albeit in diminished form, at this low NR level. The following is a resized/cropped test sample with NR set to -2

[/IMG]

and the following is a 100% crop from the original out of camera jpg, showing the curious pattern (please click the yellow bar to resize) I was referring to. Note in particular the dry grass in the middle and the bench on the right:

[/IMG]

Is this normal/acceptable behaviour for this sensor?

 

[Mr Perceptive]

Yes normal behaviour, read up on X series painterly effect. It can be reduced in camera by reducing NR and Sharpening, but it's most common effect is within earlier versions of Lightroom RAW conversion. Read up also on Peter Bridgewoods guide to processing X-Trans files.

It has never bothered me on any of my images, but for some people it appears to be some sort of showstopper.

 

[eternumviti]

Are you saying that the default setting for the camera is an effects one Mr P?

Very odd if so.

 

[JemHayward]

If fine detail is critical (e.g. I'm going to print at A0 or something mad like that) I use Iridient Developer on the .RAF raw file. JPG conversion and in camera processing will always introduce variables, and the default values / behaviour will be a matter of what the programmer thinks you were taking a picture of. If this level of detail is important, you shouldn't really be going anywhere near Jpegs.

 

[albireo]

Hi, I was under the impression, by reading on the X-trans sensor, that out of camera JPEGs were of excellent quality. I thought one of the appeals of the system is that it would allow the photographer to concentrate on composition whilst alleviating the need for extensive (and in my opinion often tedious) post-processing.

Based on the images above this appears not to be the case. The Nikon D5500 (a camera I was hoping to replace with the X-T10) has a number of shortcomings (size for some? poor AF precision?). However, its out of camera JPEGs, again IME, are noticeably superior to the above and extremely sharp down to 100%. This is of interest to some I guess (for example when creative cropping is required).

 

[JTC]

I've not been aware of any such shortcomings with the X-T10. Or perhaps it's one of these things that, once noticed.....??? I can see what you are meaning, no question, but hadn't noticed it thus far on any of our shots. Could it also be a firmware-related thing?

 

[JemHayward]

Must admit I'd never noticed it, and I'm wondering if this is why the images look so good blown up to silly sizes? Sort of built in dithering of detail? The x-trans sensor, for better or worse, is different to the beyer filters used by just about everyone else...

 

[Cesare]

The reason for this sort of complicated pattern issue is that the camera has no anti-aliasing filter, and the lens is sharp enough to get details beyond the resolution of the camera onto the sensor. The jpeg engine in the camera is trying it's best to deal with this, but basically it's having to guess what is really going on, and the results can be at times, well, odd. It's always going to be a problem, and there are different approaches used by different camera makers (especially in the jpeg engines where processing power may be limited), trading off resolution against artifacts.

Imagine a light source which is less than 1 pixel in size, say, a reflection off a bit of shiny metal in the distance hitting the sensor. Depending on exactly which pixel it hits the colour content of the light source will be unknown to the camera (it will just cause a bright spot on a red, green or blue pixel).

So, you could get annoyed by this, or you could consider it a happy problem to have, that the sensor and lens system has such high resolution that you are reaching the boundaries of what is possible.

If you want to experiment, you can actually de-focus a camera a little and basically achieve a similar effect to an anti-aliasing filter. Just focus at 20 feet not infinity, and see what happens ;-)

 

[Cesare]

Actually i've just remembered a better place to see this problem, and that's photographs of choppy water, which tends to give all sorts of odd looking artifacts on most digital cameras, but hey, best not to peek at the pixels as it'll only upset you.

 

[albireo]

Thanks for the insightful comment, Cesare. However I'm still not convinced. I would lie if I said that (based on my entirely person photographic preferences) the issue is not there and not noticeable at all.

Just for fun, here is an out of camera jpg of approximately the same scene taken with a Nikon D3200. The sensor is a respectable 24 megapixel DX made by Nikon, which at the time was really highly rated. The lens I used is a Voigtlander Ultron 40mm f/2, a lens that is apparently so sharp at f4.0+ that photozone.de concluded "surely exceeds the sensor resolution [of the Nikon D200 CCD] here as well as at f/5.6".

[/IMG]

and here's a 100% crop of the bench/grass region:

[/IMG]

I will readily admit the the light conditions were completely different. Is this so called painterly effect present here? I would say, this being a quick handheld snapshot, there is lot that leaves to be desired - however, the overall impression is to my eyes more "correct" that what I see in the XE-2 pictures.

One thing to note: the D3200 sensor appears to have an antialiasing filter, therefore as you, Cesare, say the "painterly" issue might be inherent to the lack of antialiasing filter in the Fuji 16mp sensor. If such was the case, however, why would Fuji market a camera with such an obvious fine detail smearing characteristic? Why not keep the AA filter and avoid this?

 

[JTC]

I suspect it's got nothing to do with the AA filter and everything to do with whatever RAW->JPG conversion is being done. Interestingly enough, I really don't see this effect in the X-T10 in-camera JPGs....