« Random Snap | Main | Beautiful Books »

Thursday, 28 May 2015


Feed You can follow this conversation by subscribing to the comment feed for this post.

It is not about the sensor to me it is about a great camera in the old style with proper aperture and speed controls, EXCELLENT lenses at reasonable prices in a form that is good in your hands. most cameras havs good sensors and if you shoot RAW even better. Please stop this tech stuff,,,,it's about the final photo in the end. The best photo I ever took was out of focus, over exposed, badly composed but wonderful.
Give me a camera, any camera and I am ready.Take the tech out of photography put the seeing in. Forget this foreplay.

Ah! That explains why I like my X-T1 files so much. Now I guess I can boast about my cleverness in choosing Fuji.

From two plus years of shooting an X-trans sensor Fujifilm camera, I've noticed a few things about sharpness/resolution on the system. You are absolutely right about very fine details — you mentioned leaves. I just made a very large print from a Fujifilm photograph that included a construction crane, and the details of that structure became very soft. Oddly, however, the files produce very sharp boundaries between larger areas — for example, in that photograph, the edges of buildings against sky are very crisp.

I also agree with something you wrote earlier about Fujifilm sensors/cameras producing very nice black and white quality – I end up taking quite a few of my Fujifilm images to black and white.

"you mentioned leaves. ... Oddly, however, the files produce very sharp boundaries between larger areas"

I have seen this in my X30 images and read about it online. Fuji cameras seem to have more trouble resolving green details than other colors. Something about the custom filter array and the algorithms not matching up well.

"This feeling for the user's images seems to missing in the other camera makers who only made cameras and never made film."

Agree 100 percent. I like to window shop for cameras by downloading a few RAW sample files and opening them up in Lightroom. I am amazed at how bad the color and tonality looks in some of the files with default and near default RAW settings.

Fuji's camera profiles in Lightroom aren't perfect, but they are very good and they are consistent from one camera to another.

Besides the Fuji magic, there is really something to the X-Trans color array. Here is an interesting academic article from 2008:
"Frequency selection demosaiking: A review and a look ahead" by Alleysson & Chaix de Lavarene, http://david.alleysson.free.fr/Publications/VCIP08.pdf
which discusses the benefits of more random patterns. (And I am pretty sure, the Fuji engineers based their X-Trans design on this paper)

They say in this paper that a "non periodic arrangement of size 6x6 chromatic samples gives the best visual result (RAW data) reconstruction...also, the signal to noise ratio with the pseudo-random arrangement is not very different than that of Bayer. But the noise in the case of a random arrangement loses its spacial coherence and becomes less visible making the appearance of the image more pleasant." and the paper shows how one get better separation of luminous and chroma noise than with standard Bayer. One can clearly see this in the raw's of Fuji X-Trans which has far less chroma noise than regular sensors.

Interestingly, the paper also discusses that for this separation to work best, one needs to compute specific 'filtration' values to do optimal RAW processing. This calculation is not cheap and I think this might be the main problem with most current RAW processing for X-trans files where the filtration values are not calculated in an optimal way (while the Fuji JPG engine does do the right thing and thus looks so good)

Anyway, there is much misinformation on the net about this -- often people say that Fuji "smooths" the raw while this is just the absence of chroma noise due to the X-Trans pattern. Similarly, many (all?) RAW converters do not take the right approach to X-Trans and do not calculate filtration values.

Here is a link to a RAW comparison between some other cameras where one can clearly see the lack of chroma noise on the X-Pro:

As regards the dpreview test, to be fair, in this case it looks like a depth of field thing to me. The playing card is in the background, I would say it's out of focus. Compare it with other bits of the scene, such as the white cross directly in front of the card, the Baileys label, the fine etching on the Roman statue and the gold coins on the Martini label. No problems in these areas.

If I understand correctly, Fuji is doing the exact opposite of exposing to the right and this results in better-looking photographs (to your eye). You're not going to make any friends with this.

At risk of a) repeating what other yet-to-be-approved comments will say and b) starting a big argument, I'm not so sure the X-Trans acuity issue is quite so, er, black and white.

DPReview's studio comparisons are done using Adobe Camera Raw, and it's fairly common knowledge amongst Fuji shooters that it doesn't do a particularly great job with X-Trans files, at least on its default settings. In Lightroom (and perhaps ACR) it's apparently possible to overcome this to some extent by maxing out the Detail slider (somewhat counterintuitively, since doing this has nasty results on Bayer sensors).

Alternatively, and perhaps not quite as conveniently, one can try alternative raw processors – Iridient, Photo Ninja and Capture One, in particular, are said to do a good job, and I can personally vouch for C1 as I use it myself with my XE1, and am very happy with the results.

Personal experience with Olympus and Fuji head to head is that the Fuji jpegs usually do the job, but a LOT of work on Olympus RAW files can get close to the same smoothness/color. All of the points made about the Fuji "smoke and mirrors" I do agree with, but that look suits a lot of people and can be a lot work from other brands. Also their lenses are cracking!

DR100 is the default or base mode and DR200 gives one (not two) stops adjustment and DR400 gives two (not three) stops. You need to set ISO400 to enable DR200 and ISO800 to enable DR400.

Thank you BH, Glenn, Adam, et. al. I certainly couldn't have said it better.

My first TV job at ABC in the mid sixties, I was curious about the huge and complex looking TV cameras. These "larger than a man" monsters, gliding on weighty stands, with motorcycle-like grip controls on handlebars, and turret mounted triple Fujinon lenses. Without modern computer aided lens design and the relatively poor B&W broadcast quality of the time, the lenses really had to be top notch.

They've been doing this a long time. As Glenn said; "forget the foreplay". Enjoy the experience.

Once Fuji improves fine detail resolution of foliage for landscape photographers, they will have the best all around system out there in my opinion, the best lenses, and the best bodies. The thing holding them back is the way the sensor resolves fine detail. It drove me to the Sony A7 series, even though I like the bodies and lenses of the Fuji system better. I need the fine detail of the A7r for printing large. Fuji is taking years to release the replacement for the X-Pro1. The X-Trans files seem to favor straight lines, urban environments, the files are very sharp when photographing architecture and details of buildings and signs. But point it at foliage and fine branches and they fall apart.

I'm worried! Rod said in his post that his Fuji lenses were cracking!!! I cant find any cracks in mine. Where do the show up first?

There's another advantage to Fuji's current approach - I have an X100s, XE-2, and an IR converted XM-1 - all 3 share the same sensor, so I don't have to rethink too much when swapping cameras(okay, the IR is a different beast, but the limits of the sensor are the same). This familiarity lets a photog read the light much the same way you used to do with a film you knew well.

This essay completely ignores the whole controversy about Adobe not doing a very good job on opening raw files for Fuji. There are probably a dozen comparisons online that essentially show the Adobe is the worst of the raw converters for Fuji and the problem is the handling of green, fine detail is one of the issues. It is an Adobe problem, not a problem for Capture 1, Photo Ninja or Iridient Developer. Does anyone think that dpreview is not using the same developer - Lightroom that they use for all their files? That takes care of raw.

What about the soft jpeg on the dpreview site with Fuji's built in conversion? As a long time Nikon user I know there is little relationship between a manufacturer's choice for their standard jpeg file and what can be gotten from a raw file and dpreview does not optimize settings. For many years Nikon made it clear that they did not want to bake in settings that could not be undone and so their jpeg files were soft and low contrast especially with the crop sensor cameras. Think that might be what Fuji is doing? There are a lot of assumptions being made to draw any conclusions at all.

Here we go with another Fuji bash. I shoot with both the Fuji X-T1, and a Nikon D800. And YES there is a difference in resolution. I don't rightly care if the sensor is made by Fuji, Sony, or Bob in his basement. The Fuji files have some serious mojo and horsepower, without all the post-processing required of Nikon and the like. Fuji files, even at a measly 16mp are incredibly rich and detailed. I use the latest version of LR (5.7) and I've learned the best way to process a fuji raw file. Though, in most cases, the jpeg works just as well. If you truly want to get the ultimate resolution from ANY Fuji raw file try Iridient Developer. That thing is awesome. I use it, I've also used Silkypix which came bundled with my old(er) X-E1 and it does wonders. However, to keep it all in my LR workflow, I just work with the raw file in LR. I'm not sure what the problem is with you guys and knocking Fuji but I'm sick of to here, and on other websites.

[Did you actually read the whole post, Jorge? Nobody's knocking Fuji in our posts. Quite the opposite in fact. --Mike]

For those of us not lucky enough to be using Fuji cameras—can we get a similar effect by just dialing in -1EV exposure compensation, and bringing the files back up in post?

I routinely keep my E-M1 set to -1/3EV, but maybe I should go further. Will experiment.

Mike -

I hate to sound like some old repeating record, but the Adobe products do not do the best job of sharpening the non-Bayer Fuji files. And the examples referred to are processed in Adobe Camera Raw. There are a number of ways around this such as using Photokit Capture Sharpener’s setting for scanning backs in Photoshop. Even adjusting Lightroom’s sharpening radius downward and its detail slider upwards from the standard settings will help. But the easiest and probably the most effective is to use two alternate imaging programs, Iridient Developer and PhotoNinja, to process the raw files and convert them to tiffs which can be imported back to Photoshop or Lightroom for final tweaks and printing.

I am holding out on Fuji. I had the Fuji X-E1 and couldn't get over the smearing of files and slow AF even with the firmware updates. I know that the Fuji X-T1 is improved and lovely but I honestly don't want that X Trans sensor. I am vacillating between the Sony A7II and the Fuji X-T1 or new Fuji. (X Pro 2?) I will be waiting a long time though. I have yet to find a camera that has as lovely a 3D look or detail as my Sigma DP Merrill cameras. Everything file from other cameras look pale by comparison. Sigma Merrill cameras are a pain in the butt to use and not for the faint of heart. I am hoping one day to get another Fuji, but will wait for the sensor improvements. It's because I love the retro design of the camera, not so much the smearing of files. I love the Fuji lenses and I certainly hope they aren't cracking like mentioned in this comment section.

Semi random comments:

It wouldn't hurt for many here to reread Ctein's 2012 columns, Why ISO isn't

and RAW is not Raw

The first clarifies how ISO is determined and why Fuji may properly use different criteria than others.

The second shows* how at least some of the differences between Fuji and other RAW files may go beyond simple EV adjustment.

If everything Fuji does with its ISO, response curves in RAW files and DR settings may be duplicated with EV settings and converters that allow saved custom settings, then what Fuji provides is neither magic nor cheating, but simply providing convenience for those whose tastes coincide with theirs.

It also strikes me that a hidden part of this discussion may be the "Purity of Essence" feeling that what comes out of official camera settings is somehow more correct or true than what is done by the photographer in manual setting adjustments and in post. If Fuji (or any maker) happens to add controls that duplicate what may be done in post, it may simply make some users more comfortable in that the results are in accordance with some more official source of authority and/or expertise.

I have known photographers who have an almost religious reluctance to make EV adjustments, let alone go further ...

If convenience and/or the comfort of authority make users happy with Fujis, that seems to me like a good thing for both parties.

I have not had a Fuji since my much loved F30 of years ago, and no "dog in this fight". I'm pretty brand agnostic; if the Fuji's had been available when I needed a change of tools, I could well be a Fuji user instead of Oly. So I believe my comments are pretty impartial.

Finally, I have Ctein's sample E-P1 print. I also have Pens with the same sensor system and E-M5 and E-M5 II. While the print is every bit as impressive as advertised, from that old 12 MP sensor system, the 16 MP sensor systems in the OMDs are a significant step up in practical resolution and other IQ factors.

* Among other useful things.

Like many other said, in my experience X-trans files don't suffer from inferior detail (AKA sharpness), quite the contrary: when used with some RAW developers like Photoninja or Iridient, X-trans actually has a slight advantage when compared with bayer.

What the X-trans array do have is a bit less color resolution when compared with bayer. If I may, here's a discussion on Fred Miranda's alt forum on this subject and where I posted some direct comparisons between a X-trans and bayer:

But do I care? Not much, when I want fine detail on a photo (usually landscapes) I just use Photoninja to demosaic the RAW file to a TIFF and then continue my Lightroom workflow.
But just for files from ISO 200 to 400. At higher ISO's I find Lightroom (and ACR) actually do a better job balancing detail, noise and color reproduction.

I don’t understand the notion that the Fuji sensor is no big deal because it’s just an old Sony sensor that’s been rearranged a little differently. ‘Seems to me that’s a little like saying that a Tesla really isn’t anything new, it’s just a Chevy that’s been rearranged a little differently.

As for the rendering of X-Trans files by Lightroom: Adobe just seems to be using the same method Fuji is using with their in-camera JPEG engine. I compared some pictures that I converted to JPEG in-camera with the Lightroom RAW conversions and they essentially show the same artifacts for foliage etc. (I remember reading somewhere that Adobe stated they worked with Fuji to improve the rendering of X-Trans files, so I guess it's possible that Fuji just shared their algorithms with them.)
And then, everybody praises the fabulous JPEGs from the Fuji cameras. Go figure! :-) (To be fair: what people probably refer to, when the say that, are the colors.)

But in the end, what keeps me with Fuji and Lightroom is the exquisite (to my taste) rendering of colors that I get out of the box with Lightroom's Fuji color profiles.

I recently used my old D7000 alongside the Fuji to shoot a family event, since I don't have any Fuji lens wide enough for pictures of large groups. And when I imported the pictures to Lightroom, the Fuji ones were mostly gorgeous from the start and required minimal processing of the colors. But the ones form the Nikon overly just didn't feel right and I had to use some serious tweaking to get them where I was satisfied. Especially skintones of outdoor pictures were something that cost me quite a few hours to get where I liked it. And then it still wasn't all the way there. (The lighting situation might have been a cause here, as I mostly shot the Fuji inside, but overall, even the less problematic files from the Nikon required more work to get them where I was ok with them.)

That experience set an end to the lingering thoughts of switching away from Fuji that I had been having for some weeks. (Maybe not forever. For some use cases these demosaicing problems, that are definitely there, might be a big enough factor to be a deal breaker in the long run. It certainly complicates things sometimes.)

Also another factor: Lenses. I'm totally in love with the rendering of the Fuji 23/1.4.
I did a test recently with the same lens (Nikon 50/1.8D) on the Fuji and the Nikon and that levels the playing field a bit. With the respective cameras standard color profiles selected in Lightroom (not the Adobe ones), the difference is not so big anymore. But still, the combination of the Fuji lenses and color profiles makes a package that makes it hard to beat for me right now.

"And when I imported the pictures to Lightroom, the Fuji ones were mostly gorgeous from the start and required minimal processing of the colors. But the ones form the Nikon overly just didn't feel right and I had to use some serious tweaking to get them where I was satisfied. Especially skintones of outdoor pictures were something that cost me quite a few hours to get where I liked it."

In PhotoShop, Image=>Adjustments=>Match Color will often do an amazing job of converting the color 'look' of one image to that of another.

I can't guarantee it will work for your Fuji/Nikon images, but where it does, it's like magic and saves a great deal of work.

Dear Moose,

All very well said!

I have to say, for me, the ways in which the Olympus EM5 is superior to the EP1 are, in order of importance,

–– Exposure range (which is two stops greater)
–– Image stabilization (at least one stop better and two with many of my lenses)
–– ISO 400+ noise (one stop better)

Yes, the extra bit of resolution is nice, I ain't complaining, but it's fourth on the list.


Dear Andy, et.al.,

Yeah, finding the right raw converter is not usually a problem… But when it is…

I think of camera sensors being like film and the converter as being like developer. 99% of the time a standard developer is going to give you the best all-around results. Or as close as matters. A good example of that in film is D-76. Every so often you're going to hit a film that doesn't want to play that way. You *could* pull-develop Tech Pan film in highly-diluted D-76 to get the contrast down to normal, but you'd be pretty damned unhappy with the results. By far the best developer for that film was Technidol. Even Kodak's all-around-superior-to-D-76 Xtol was inferior to D-76 for a few percent of the films out there.

So it is with raw converters; just as with developers, there's a lot that goes on under the hood that you don't see directly. ACR has a hell of a lot of secret sauce built into its processing. You can see that in the JPEG below, which compares ACR on the left with a well-regarded RAW display program on the right (I am intentionally not identifying which one I'm talking about, because I am not trying to badmouth a product, here; please respect the sensibilities).

These are 200% blowups from the same Olympus OMD E-M5 file. I've turned off all the “enhancements” that I can in ACR––clarity is set to zero, as is sharpening and noise reduction. Still, clearly a lot is going on under the hood!. The noise is much finer in the ACR rendering. Even more important, ACR understands the Bayer filter array offset in ways that allow it to produce clean monochromatic edges at the bright transitions. Finally, the fine detail is much better: look at the braiding in the cables where ACR has managed to reconstruct real image detail that the “straight” display program completely misses.

I'm not trying to say that ACR should be superior in every case. Rather, the contrary. I'm illustrating how much goes on behind the curtain in a good raw converter, and those hidden machinations will be different for different converters. In the context of Kevin's article, you might say ACR is to D-76 as X-Trans files are to Tech Pan.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com

"... I have to say, for me, the ways in which the Olympus EM5 is superior to the EP1 are, in order of importance,

–– Exposure range (which is two stops greater)
–– Image stabilization (at least one stop better and two with many of my lenses)
–– ISO 400+ noise (one stop better)

Yes, the extra bit of resolution is nice, I ain't complaining, but it's fourth on the list."

The E-M5 II adds little to sensor system performance. It does bring significant improvement in IS. I don't know that this is of great significance for 'normal' focal lengths and uses. I have yet to press it much in the dark, although a couple of grab shots (NO photography!!) last night in an auditorium with 75/1.8 @ f9, before I remembered to change aperture, are remarkably free of blur at 1/20 sec.

I do a lot of very long lens and macro work, often shooting the same subject from very close with the 12-50 in macro mode or 60/2.8 macro and from a distance with 75-300 near its long end.

Oly claims improved IS for macro, which I've not really tested. Certainly I know the O-M5 is significantly better at macro IS than any of the Pens through the E-PM2.

At the long end, the improvement is quite noticeable. It's almost as though the 75-300 @ 300 mm suddenly became a better lens.

There is what appears to me to be an exposure anomaly in A Mode, in the range bounded between good and dim light. It's choosing exposures in A Mode with shutter speeds far too low for any but short focal lengths before using Auto ISO to keep shutter speeds reasonable.

The algorithm is clearly different from that of the Mark I body, in side by side tests. I've been in contact with Oly support; they seem concerned, in what may be a continuing dialog, as I now have more info.

Fujifilm or not, there are so many mistakes, misconceptions and misstatements in this article I just flat give up.

Except... please show one piece of objective evidence, you know the kind that would be accepted a factual, Fujifilm uses a SONY sensor chip (which to minimize ambiguity in my question) is only part – albeit an important part - of the sensor assembly.

The issue of the "water color effect" in green foliage is quite clearly discussed in academic literature as a failure of SOFTWARE to properly bias or weigh the mathematic relationship with between chroma and luminosity information contained in any given RAW file. Water color effect is just one of four ways to get it wrong.

This has nothing to do with any inherent quality with the Fujifilm 6x6 color filter array. Instead, the industry complacency generated by a decade of Bayer array programming caught out Adobe. Their software approach was a decade old. Others, such as Capture One, had no such problem because the software they used math of four or five generations newer refinement. End of story.

Fujifilm industrial color dyes used by a very large percentage of other camera makers for their color array filters. We can suppose Fujifilm has reserved their very best formulation for their own camera products. While not magic, the color rendering of the Fujifilm RAW file is subsequently very precise and well controlled in most lighting conditions. Fujifilm is the unquestioned industry leader in chemical dyes for many uses, including medical imaging, cosmetics and photography.

Kevin, you are going to write about photography, please do the homework yourself on how demosaicing works instead of condensing urban myth generated on the internet. Popular information is not proven information, often far from it.

Mike, overall I like your site very much. Keep improving!


To MikeS:

Could you please give the references of academic literature you are mentioning in your comment? I'd be very interested in giving it a read.

X-trans is a solution in search of a problem, as we're all pretty happy with Bayer already, which tends to be a better compromise between low ISO detail and high ISO noise.

The thing is, Fuji already nailed everything with the Bayer sensor in the original X100, which is the only camera I shoot. I tried the X100s with the X-trans, but went back to the X100 after months of frustration with the X-trans files.

I don't remember anyone saying that the X100's IQ was bad when it arrived, but I think Fuji did the X-trans thing in an attempt to differentiate themselves, which was a bad move, because the lenses and camera designs are already a great differentiating factor. As it stands now, there are many shooters who would be shooting Fuji, had it not been for the introduction of the X-trans.

I still hope they go back to Bayer someday, myself.

Fuji being Fuji they are certainly NOT using any Sony sensor. I worked for many years in factories as an EE where we partnered with Fuji for machine control vision systems. Their semiconductor mfg. and RD is the finest available. As to the X-trans geometry being a solution in search of a problem; well just do the math, you'll see literally why it allows finer algorithm design and control.
Anyway, a fun and often unintentionally funny read here!

For S.Sorace


This article requires a strong maths background to fully comprehend, but along about page 52 are observations written out in plain language. The water color effect is exactly described.

For GH,

The notion of a trade off between low ISO detail and high ISO noise has noting to do with the array configuration. This is quite a meaningless statement or as Wolfgang Pauli once remarked, "Not only is it not right, it is not even wrong." ("Das ist nicht nur nicht richtig, es ist nicht einmal falsch!") I am sorry, but Bayer or otherwise, it is strictly a digital signals filtration problem related to dynamic range.

Where are you guys picking up this hearsay? Adobe never did come clean about their problems. How that somehow makes Fujifilm's technical choices a 'bad move' should be a warning to all who use the internet! And, a waring to those who participate in a monopoly.

As you browse thorough the articles I have referenced, you will already see the problems Bayer has with many aspects of demosaicing. It is precisely these trade offs that a larger 6x6 matrix and more complex maths make easier to resolve.

Bayer is simple to manipulate mathematically, that is why it was developed. Personal computers back then could not process anything more complex in a reasonable amount of time. But, it is inferior when compared to the power of modern maths routines and a larger matrix in all aspects of demosaicking. The only problem with larger matrices is the rapidly increasing maths complexity and computing overhead involved. For this reason the 6x6 color filter arrays was chosen as the best compromise between complexity and image quality benefit.

Note that the article referenced above is about Bayer and not about Fujifilm 6x6 pseudo random arrays. For a more complete understanding of larger array demosaicking, it is necessary to examine the more modern maths that have the power to process larger arrays.

For this understanding, the pivotal paper was by David Alleysson, et.al.


This is perhaps the best reference for understanding the advantages of more complex maths and larger arrays and/or why Fujifilm would have gone to the trouble in the first place. Just ignore the maths and read through the explanations and conclusions. David is a very cool guy, currently working on concepts in artificial vision for blind persons. Please respect his privacy.

There is a considerable body of academic literature from 2000-2010 that foretell the appearance of the X-Trans 6x6 matrix, but these two papers I have listed are probably the most important ones.

MikeS: Thank you!

The comments to this entry are closed.



Blog powered by Typepad
Member since 06/2007