Luminous Landscape Forum
Equipment & Techniques => Mirrorless Cameras => Topic started by: Chris Kern on October 14, 2014, 06:46:04 pm
-
I've been using Iridient Developer to convert raw files from my FujiFilm X-E2 for critical work until now because no matter how much I fiddled with the sliders in Lightroom, I just didn't feel I could capture as much fine detail without introducing artifacts.
I'd prefer to avoid the extra overhead of having Iridient produce a TIFF, however, and the other day I stumbled on an interesting article by Pete Bridgwood (http://petebridgwood.com/wp/2014/10/x-trans-sharpening/) which suggests, among other things, that moving the LR sharpening-detail slider all the way to the right — something it had never occurred to me to try — works very well with files from the Fuji X-Trans sensor.
The results of my initial tests of Bridgwood's technique are quite promising. With very little effort, I'm now able to perform capture sharpening with Lightroom that is essentially indistinguishable from what I was seeing with Iridient. At least that seems to be true with respect to detail; inevitably, the default settings of both programs produce differences in color and perhaps other characteristics of the respective images.
I really don't understand what goes on under the hood during a raw conversion, but apparently the use of non-Bayer patterns such as X-Trans may require changing some entrenched post-processing habits.
-
My understanding is that Irident uses some deconvolution, which is what happens when you slide the detail slider to the right, from 50-100 a form of deconvolution is being used in Lightroom.
Alan
-
My understanding is that Irident uses some deconvolution, which is what happens when you slide the detail slider to the right, from 50-100 a form of deconvolution is being used in Lightroom.
That's what Bridgwood says in the essay I cited, and Jeff Schewe explains what happens with that slider in even greater ... ummm ... detail in The Digital Negative (p. 89 in the print version). At the time I read Jeff's book, I wasn't using a camera with an X-Trans sensor; I rarely had occasion to stray from LR's default sharpening-detail setting of 25 when working with Bayer-sensor files as long as the images were properly focused. It didn't occur to me that a different sensor pattern would demand a different sharpening algorithm, which is what Bridgwood claims.
-
It didn't occur to me that a different sensor pattern would demand a different sharpening algorithm, which is what Bridgwood claims.
rather different demosaicking (which in xtrans case involves blurring to avoid artefacts) requires more deblurring... that may or may not be incorporated into demosaick itself, in Adobe's case it was apparently not... try NoiseNinja - their approach is apparently different - they intentionally produce artefacts (false details) creating an impression of better resolving demosaick
-
I've just finished a quick test of the public Beta of Iridient Developer version 3 (3.0b3). Using a maxed-out sharpening detail slider in Lightroom and leaving all the other settings at their default values, I'm still seeing capture-sharpening results form LR that are competitive with those from Iridient at its default settings. And the results remain comparable after I have made whatever additional sharpening adjustments I want for a given image. Again, there are other differences between respective images — default colors, for example — but the two products seem to have equal ability to represent the details of the original capture.
-
I find that the detail=100 method is indeed less prone to introducing haloing in my X-E2 files than keeping it around 25-35 as I do for my Bayer sensor cameras.
However, the problem for me is that in a fair number of files, Lightroom's demosaicing introduces halos or similar artifacts even when there is no sharpening at all applied in LR. So I'm using Photo Ninja for initial exposure adjustments, sharpening and perspective control before refining and printing in LR. But you need a really light hand on the Detail slider in PN because it's easy to produce files that look horribly "digital" in the print, although they may look OK on the monitor.
I'm in the process of exploring how the Sharpening Radius and Detail sliders interact in Photo Ninja, will post crops when I have the results.
-
i, too, found the detail slider tip in LR5 very useful. although i found that i stop around 80 instead going all the way too 100. at 80, it looks abit more natural to me. i'm still exploring how to sharpen X-E1 files. i always end up with photos that looks like oil painting.... some times it works, some time just looks like cartoon....
i do mostly landscape shots, i will do amount:40, radius: 0.7, detail: 80, then mask will depends on the picture, but generally make sure the sky is not sharpened. then i use 10-15 on noise reduction.
any more tips?
-
I find that the detail=100 method is indeed less prone to introducing haloing in my X-E2 files than keeping it around 25-35 as I do for my Bayer sensor cameras.
However, the problem for me is that in a fair number of files, Lightroom's demosaicing introduces halos or similar artifacts even when there is no sharpening at all applied in LR.
I've not seen this, myself, although my experience has been that raw files produced by the X-Trans cameras neither require nor can tolerate much sharpening beyond the amount I automatically apply during import (i.e., the sharpening-amount slider set to 25 and the sharpening-detail slider to 100). The files from cameras with Bayer-pattern sensors seem to be more accepting of aggressive sharpening. On the other hand, I've been very favorably impressed with the focus accuracy of the two Fujifilm cameras I have used—the X-E2 and the X-T1—as well as the quality of the Fuji lenses. So I don't often feel the need to sharpen the Fujifilm images aggressively.
What version of Lightroom are you using? From what I've read, here and elsewhere, Fuji reps have been helping Adobe refine the X-Trans demosaicing algorithms in LR and Camera Raw. If you're not using the latest rev of Lightroom, perhaps that's the reason for the artifacts you're seeing.
I've attached three test shots I made this afternoon to illustrate the results I get from my default import capture sharpening:
(1) the approximate coverage of the full frame in the following two examples;
(2) a full-resolution crop from an X-T1 with the Fuji 10-24mm zoom at 10mm;
(3) a full-resolution crop from a Nikon D800E with the Nikon 18-35mm zoom at 18mm.
I applied the Lightroom lens-correction module to the D800E image; it's my understanding that LR automagically performs lens corrections on files made with all Fuji lenses based on manufacturer's makernotes included in the metadata. The D800E image also had the sharpening-amount slider set at 25, but the sharpening detail value was 25. Both crops represent my typical starting point for post-processing images made with the respective cameras. (Obviously, the dimensions of the D800E sample are larger because of the higher resolution of the sensor.) Both look equally clean to my eye.
-
i, too, found the detail slider tip in LR5 very useful. although i found that i stop around 80 instead going all the way too 100. at 80, it looks abit more natural to me. i'm still exploring how to sharpen X-E1 files. i always end up with photos that looks like oil painting.... some times it works, some time just looks like cartoon....
i do mostly landscape shots, i will do amount:40, radius: 0.7, detail: 80, then mask will depends on the picture, but generally make sure the sky is not sharpened. then i use 10-15 on noise reduction.
any more tips?
Nope. I think every image pretty much needs to stand on its own. The gimmick I reported in my original post is an exception—and even it probably shouldn't be applied mechanically.
-
I'm interested in settings that work in Lightroom. I have been shooting with x-trans since they were first released and never have been happy with the files for landscape work where there are a lot of fine details such as leaves, or bare branches. I'm on a PC so can't use Iridient, and my trials with Noise ninja haven't been much better than Lightroom.
Sometimes adding grain will break up the smooth halo/watercolor effect which can surround fine detail, but that adds grain to the entire image, which isn't always desirable.
Any specific setting that you see working best, or as a starting point?
I tend to start at about 50-80 on Amount, .7 or .8 on Radius, and vary the detail to suit the image.
Using Lightroom 5.6
-
Any specific setting that you see working best, or as a starting point?
I tend to start at about 50-80 on Amount, .7 or .8 on Radius, and vary the detail to suit the image.
Using Lightroom 5.6
Since I read the article that prompted me to start this thread, I've been setting the sharpening detail to 100 on import for my X-Trans files and the sharpening amount to 25, and leaving the radius at 1 until I begin working manually on a particular image. These import settings appear to produce the same amount of detail as Iridient at its default (i.e., out-of-the-box) settings. I really don't have any way of knowing whether they allow me to exploit every bit of detail a particular sensor-lens combination is able to capture, but cranking up the detail value (i.e., using as much deconvolution as Lightroom 5.6 offers) allowed me to remove Iridient from my standard workflow. The files at this point always look clean to me: I can't detect any halos or other objectionable artifacts. As I mentioned in an earlier post, I find I have to use a very light hand on the sharpening-amount slider after this initial capture sharpening. The X-Trans files don't seem to tolerate aggressive sharpening as well as files from cameras with Bayer-pattern sensors.
Having said that, I tend to think of the Fujifilm cameras as better suited to street photography or cityscapes than to landscapes with a lot of fine, irregular detail.
-
Hi Chris,
I read your comments with much interest and also relief at the thought that there is finally an acceptable alternative to messing with my LR workflow. In the end I finally bit the bullet and spent several hours testing out these suggestions on 6 of my own files, including both 2 portraits (for skin tones) and some tricky landscapes (one extreme high contrast situation and others with fine, regular tree details) to see if I could replicate the talked of issues and then get a satisfactory fix in LR 5.6. I compared outputs form the same 6 images using Capture One's trial version and also Iridient Developer's latest release. Put simply, the objective was to see what all the fuss is about and whether the other converters make a big enough difference to be bothered about.
Long story short is that I was blown away by just how much extra detail I can get from ID than LR (and to a lesser extent Capture One, which had the nicest "look" aesthetically but still less detail). Bearing in mind I am a competent LR user and hadn't used either of the others until today, and I spent many hours trying to prove myself wrong as I so desperately would like to stick with LR only. No contest (still) I'm afraid and I couldn't get rid of those awful "paint effect" trees in LR whatever I tried :(. Capture One had the same effect, but more muted. The extra detail in ID is truly remarkable as much as I hate to admit it! No contest.
Steve
-
Long story short is that I was blown away by just how much extra detail I can get from ID than LR (and to a lesser extent Capture One, which had the nicest "look" aesthetically but still less detail).
If I were to take a stab at summarizing the comments from this and a related thread, I think it's accurate to say that none of the contributors felt Lightroom provided the best conversion of raw X-Trans files of the four major contenders: LR, Iridient, Capture One and Photo Ninja. We didn't agree on which of the four was best, but we all agreed it wasn't Lightroom.
My conclusion from my own experiments was that Photo Ninja seemed to provide the best detail when comparing three of the software products at their default settings. (I didn't test C1.) However, I need to qualify that conclusion carefully. I didn't attempt to tweak any of my test images in any of the products. I assumed it would be possible to get better results for any particular image with any of the products by skillfully using the tools they provide, but I didn't make that kind of best-case comparison. My purpose was simply to determine whether the "trick" of moving Lightroom's sharpening-detail slider to the right sufficiently improved the level of detail in X-Trans files, as Pete Bridgwood claimed (http://petebridgwood.com/wp/2014/10/x-trans-sharpening/), that I could stop routinely using Iridient to demosaic all my Fuji files. I decided the improvement was sufficient that I can now make a decision on an image-by-image basis. But I'm not entirely satisfied with any of the products I tested.
My hope is that Adobe will do a better job with X-Trans files in Lightroom 6. I presume we'll find out before too much longer.
-
My hope is that Adobe will do a better job with X-Trans files in Lightroom 6. I presume we'll find out before too much longer.
I'm generally satisfied with Pete Bridgwood's guidance as well. When it comes to demosaicing non-Bayer sensors though I wonder if Adobe would really undertake that kind of effort. Have the other RAW converters really done that or have they just developed ways of dealing with X-Trans that doesn't require them to write new demosaicing routines?
-
Hey Chris,
Thanks for the quick response. As it happens, I reluctantly spent a bit more time playing around with these results earlier today. I say "reluctantly" as, being a bit lazy, what I'm really after is a solution to this issue that will work without having to spend forever tweaking images in front of a computer - I'd rather be taking images and get it almost right in camera, and just have minimal routines for processing (or at least ones to give me a reasonable starting point, that can be played with if necessary from there).
As it happens, I think I now need to retract pretty much most of what I said late last night! By then I had lost patience and felt defeated at having to accept an outcome that was not what I was hoping for. Now, on further analysis I think I HAVE successfully managed to get versions of all 6 test images that are every bit as detailed in LR (in each different scenario) as I previously got in ID. The thing that was bugging me (and lead to further experimentation) was that (a) I preferred the interpretations of the colours I was getting in LR to ID, and nothing whatsoever I tried before or since could get them to match adequately, and (b) I've recently invested in using the amazing VSCO Film simulation products for LR, which are designed to work with their unique and carefully mapped Camera Profiles (i.e. from RAW or DNG files) to give the best results (though they do have generic "Standard" profile presets for when you have to use other formats including TIFF, which is my only real output file option when using ID as it doesn't support DNG or other outputs with embedded camera profiles to the best of my knowledge). In fact, if you're impressed with the look of the Fuji in camera simulations (but Jpeg isn't sufficient for your needs), then I'd highly recommend a look at the VSCO Film packs, which bring back gorgeous film like looks to your images (with or without the optional film grain simulations too). These guys really know there stuff and have invested a lot of time doing amazing profiles of processed films.
Here's how I cracked the problem: First, I was using output settings for ID suggested by Olaf Stzaba, who's work I love (see -http:////olafphotoblog.com/?s=iridient+developer (http:////olafphotoblog.com/?s=iridient+developer)). His simplified approach (after input from the author of ID) is "R-L deconvolusion sharpening with the following inputs 0.5 & 30. Then we put saturation at 7 or 10 and adjust the exposure to our liking". To start with, with my DNG files in LR I used Pete Bridgwood's settings and ended up going with his sharpest "Tack " settings to give me anything like the detail I was getting in the ID files. However, the detail in the LR files was nowhere near as good, especially in the very high contrast landscape test shot I was using and, worse still, the example I used to test out the painterly effect in LR with repeated fine tree details showed up really strongly in one particular file (though we're talking pixel peeping at 200%). Here the comparison in detail of the ID files was huge.
Today however, whilst trying in vain to get the ID files to better match the colour rendition of the LR files, I started playing more with the Noise Reduction settings in the Detail panel in LR. I had noticed that, although the ID files had better detail, there were some observable artefacts being introduced (using ID's default noise reduction settings) though we're talking pretty subtle when scrutinised at 200%. In fact, if you look at the full sized files in Olaf's examples (link above), you can see them appearing there too (the best/worst example is the beach scene (bottom left corner), in the wet sand just before it meets the water to the right of the guy in the red t-shirt). They look like faint worm like patterns. So, in my LR files, I backed off the Noise Reduction settings to where you could see a tiny bit of noise zoomed at 200% (comparable to the ID files). First thing I noticed was that "voila" the painterly effect on my LR file with the tree detail was suddenly gone! So I pushed the Sharpening settings some more (leaving Detail set at 100, to still get RL Deconvolution), increasing the Amount slider (and in some cases pushing Radius a tad too) until comparable artefacts started appearing, and then backed them off a bit and/or played around with Masking.
In the end, I would say that some of the originally worst areas of the LR files (compared to ID) became at least as detailed, and here and there even MORE detailed than I had got with ID - unbelievable! For the life of me, I have no idea how or why I couldn't get this with my experimenting yesterday, but I guess learning the finer nuances of the LR settings takes time (and I'm not talking about "wanging" around the sliders as Guy Gowan likes to say, just tiny adjustments, a bit at a time). End result is that I'm finally every bit as happy with my LR files (and in some cases/respects more so) than those from ID. However, I confess I haven't experimented with the ID settings to see if I can improve output detail further still (and I don't think I can be bothered to). I did initially try some more subtle settings in ID for the portrait examples (as per http://www.marksoon.com/blog/2014/fuji-x-trans-raw-processing - note the Iterations level is only 7, not 30 as used by Olaf Stzaba) but based on the preview in ID it looked ok when I pushed it higher - So I never checked actual output TIFF's at these lower settings.
The only down side (if you could call it that) was that EVERY image had to be tweaked differently in LR (just a bit) to get these results, whereas ID was originally giving me better results off the bat using the same settings. However, it has given me a better starting point to use for my own files from now on (and I have learnt a great deal), and in reality I will only be fine tuning my very best images, not every single one - so it's not a big issue.
Well, I think that's more than enough rambling from me - I apologise for going into such detail, but I thought this may help those of you pondering the same issues. Feel free to message me if you have any questions, or if you would like to see any of the files for side by side comparison.
Steve
-
If you did not see it I also played here: http://www.luminous-landscape.com/forum/index.php?topic=94681.0 with some files.
I'll try some of your suggestions and see if there is any improvement.
-
Armand, thanks for the heads up on the other discussion - I wasn't aware of that and in particular the halo issue until you pointed it out. Don't know whether to thank you or to cry! Just when I thought I'd finally got what I needed all sorted eh? (ha, ha!)
I see that Chris has also posted details of halo issues he's been having, though I struggled to see it in the images attached to his post (all I could see looked like a bit of colour CA fringing). By the way Chris, love that image. Do I take it that's the Charles Bridge tower in Prague? Amazing place, though last time I went there was many years ago when I was still shooting film!
Anyhow, after reading all your posts (and seeing the halos in Armand's LR shots) I found it too in one the 6 X-E2 test files I'd been using (damn!). I haven't had time to investigate this fully within LR yet, but here's what I found so far. In my own image it's again in a high contrast boundary between sky and the dark edge of a tree's structure. However, this had been processed using VSCO Film's Fuji Velvia 50 from Pack 04, which as they explain in their support pages make some pretty extreme changes and can bring about artefacts in boundaries between colours sometimes (see their suggested fix here - )https://vsco.zendesk.com/hc/en-us/articles/200324310-VSCO-Film-04-I-m-seeing-some-unnatural-looking-lines-and-artifacts-on-my-photo-with-a-VSCO-Film-04-preset-applied-what-s-wrong- (https://vsco.zendesk.com/hc/en-us/articles/200324310-VSCO-Film-04-I-m-seeing-some-unnatural-looking-lines-and-artifacts-on-my-photo-with-a-VSCO-Film-04-preset-applied-what-s-wrong-)
Note, when I examined this same image in the out of camera Jpeg there was NO halo/fringing at the same boundary, nor in the original RAW/DNG before processing with VSCO Film, and it also appeared in the ID TIFF file only after this was processed the same (so perhaps this is not a good example to illustrate the point you guys are making). Having said that, it may be worth you guys trying the fix suggested by VSCO using the HSL panel all the same. So far, I successfully managed to remove the fringe, but only by reducing the sharpening Amount and/or Radius sliders but to such a point that I was no longer happy with the level of detail in the image (i.e. it fell back below the standard in the ID file), so I'm not happy going down this route to solve it.
In the meantime, I'll have another look for a different image where there's halos even with the sharpening turned off (if I can find one like Chris mentioned). Would be interested to know what you both think on the other changes I mentioned before to get rid of the plastic/painted details and to improve overall detail. I could clearly see this in Chris' shot of the trees over the lake (which were fine in the Nikon D800e version) - I'm sure you can fix that trying what I did.
Also Armand, I note you're particularly keen on the profiling to give you something like the Fuji Provia (and others perhaps?) look without having to spend ages getting there - In which case, definitely check out the VSCO Film packs, which will do exactly that (in a massive list of flavours) with just one click (and they don't generally give you problems like the exception detailed above).
Finally, if it helps to share and compare full sized images properly (or if you want to try some VSCO presets and profiles), it would make life a lot easier to do this using a shared Dropbox folder (or similar). If you like I could set one up for you guys?
Cheers,
Steve
-
I see that Chris has also posted details of halo issues he's been having, though I struggled to see it in the images attached to his post (all I could see looked like a bit of colour CA fringing).
I didn't even notice it until you pointed it out, Steve. There is indeed some green fringing, although probably not visible at resolutions less than 1:1.
By the way Chris, love that image. Do I take it that's the Charles Bridge tower in Prague?
Thanks, and yes: that's the Charles Bridge tower in the Nové Město, shot at the end of a warm summer day in August, just minutes before the skies opened up and my wife and I got drenched walking back to our hotel.
-
Steve, thanks for your detailed posts! Please continue to keep us informed of what you discover.
-
Hey Chris & Armand,
I think I've found the Holly Grail of solutions to these perplexing issues (the detail recovery and the halo issues too) .... However I'm not yet finished exploring this enough to give you the low down, but I'll let you know as soon as I have.
Don't know if either of you have heard of Guy Gowan? I came across his details whilst looking into this subject over at Fujix-Forum.com. He has VERY frank views on a lot of things, but he really, really knows a lot of amazing techniques based on what he learnt about Colour Theory and techniques they used back in the days of Lithographic Darkrooms before Desktop Publishing came along. He's been adapting these for digital since the very first version of Photoshop and the results are nothing short of jaw dropping compared to what we're all used to.
After much deliberating (principally due to some negative, but largely uninformed, opinions about him caused by his brash views and style) I recently took the plunge and subscribed to membership of his site GuyGowan.com which has massive resources of really useful information and techniques, if you have the time and determination to get your head around it all and navigate through the many, many video tutorials and webcasts. I do not regret this one bit. I have learnt sooooo much great stuff already.
When I get to the point I need to, I will do you guys a detailed post with some comparison examples. The sharpening techniques alone are worth the cost of membership, and I'm confident you will be blown away by the detail that can be extracted from ALL your files (not just the Fuji's) without the slightest trace of halos or artefacts even at 500% on the Amount slider!
Be back soon ...
-
Iridient Developer 3.0 Beta 4 - The 'Real Deal' For Fuji X (http://www.soundimageplus.com/soundimageplus/2014/12/12/iridient-developer-30-beta-4-the-real-deal-for-fuji-x#)
-
Iridient Developer 3.0 Beta 4 - The 'Real Deal' For Fuji X (http://www.soundimageplus.com/soundimageplus/2014/12/12/iridient-developer-30-beta-4-the-real-deal-for-fuji-x#)
Those colors are way more saturated than the Fuji ooc jpeg. And there is some red-orange weird stuff going on (look at the oranges for example).
-
Iridient Developer 3.0 Beta 4 - The 'Real Deal' For Fuji X (http://www.soundimageplus.com/soundimageplus/2014/12/12/iridient-developer-30-beta-4-the-real-deal-for-fuji-x#)
Brian Griffith, Iridient's developer, has posted a change history here (http://www.iridientdigital.com/products/rawdeveloper_history.html).
-
Those colors are way more saturated than the Fuji ooc jpeg. And there is some red-orange weird stuff going on (look at the oranges for example).
I'd be wary of making any judgement about colour from this site's on-line jpegs. I'm not sure if its a browser issue or his method of exporting images to web (profile issue) but a ton of his work appears 'bizarre' and not just these Fuji files ( note: there was a thread on this recently on LuLa involving Hans Kruse and some others IIRC).
I've been using ID since version 2, haven't downloaded 3beta yet, and it's long been my only converter for x-trans files, only now challenged by the much improved CaptureOne v8. Colour rendition has never been an issue, though I use both Fuji and some custom profiles.
As far as sharpening goes, the jpegs speak for themselves. From what I see, using ID is close to a no-brainer - if for nothing else other than just sharpening ! I'm assuming the colour differences are an aberration - the only way to know is to run your own comparative tests on your own setup before forming a definitive opinion.
-
interested if Lemondixon had any more comments on the processing of X Trans files.
It sounds like you were getting good result, did you find the holy grail settings in lightroom?
-
Many apologies for not responding sooner - Life's just been too hectic to make time to do this, and I want to do this properly as I think I will have something really useful to share with you all. Not there just yet though, but have promised myself will do this for you in next couple of weeks.
-
I just took a look at the latest version of Iridient and boy they have come a long way, and in the positive.
I of course wish they had a windows version, but they really can pull out the details from the Fuji xTrans files.
With their latest version there are a lot more features to help on shadow/highlight corrections, but still no dedicated color controls. Maybe in the future.
Paul
-
I just took a look at the latest version of Iridient and boy they have come a long way, and in the positive.
I've also been playing around with Iridient v.3. But I'm not persuaded its conversion of X-Trans files, even with the new demosaicing options, is that significantly better than in earlier revs. It's certainly acceptable, but I still am able to achieve equivalent results with Lightroom 5.
I'm waiting now to see what if anything new LR 6 has to offer Fujifilm shooters.
-
I've also been playing around with Iridient v.3. But I'm not persuaded its conversion of X-Trans files, even with the new demosaicing options, is that significantly better than in earlier revs. It's certainly acceptable, but I still am able to achieve equivalent results with Lightroom 5.
I'm waiting now to see what if anything new LR 6 has to offer Fujifilm shooters.
HI Chris:
I agree that LR can at times to a great job, but there are times where it just seems to really damage the files.
Landscape shooting where I have faint tree branches against a blue sky, LR tends to add a reddish hue in these parts and if you attempt to sharpen the image at all, haloing seems to come into play. I also see this same issue with green leaves against blue sky, sky being polarized or not problem is the same. I have tried various settings including taking the detail slider up 100% which supposedly brings on deconvolution sharpening, and that helps on certain file.
I have not had any real success with exporting to PhotoNinja for the raw conversion and then working on the tif in LR. PhotoNinja does seem to put in a lot of false details, something I had read elsewhere and on this site, but never really paid attention to.
C1, which I love as a tool, for Phase One and Nikon D800e and D810 files, just can't begin to get the finer details, however it does a great job on colors so I use it when I can.
Iridient pulls out extremely fine details which is what I am looking for and does it without a lot of false detail. Their weakness is a tool set, but with each new version it's getting better. With 3 there is a lot of great new control for shadows and highlights, albeit 100% image wide, as there is no brush/masking style. But if he would just add a color editor it would be a great starting place.
Paul
-
I've not seen this, myself, although my experience has been that raw files produced by the X-Trans cameras neither require nor can tolerate much sharpening beyond the amount I automatically apply during import (i.e., the sharpening-amount slider set to 25 and the sharpening-detail slider to 100). The files from cameras with Bayer-pattern sensors seem to be more accepting of aggressive sharpening. On the other hand, I've been very favorably impressed with the focus accuracy of the two Fujifilm cameras I have used—the X-E2 and the X-T1—as well as the quality of the Fuji lenses. So I don't often feel the need to sharpen the Fujifilm images aggressively.
What version of Lightroom are you using? From what I've read, here and elsewhere, Fuji reps have been helping Adobe refine the X-Trans demosaicing algorithms in LR and Camera Raw. If you're not using the latest rev of Lightroom, perhaps that's the reason for the artifacts you're seeing.
I've attached three test shots I made this afternoon to illustrate the results I get from my default import capture sharpening:
(1) the approximate coverage of the full frame in the following two examples;
(2) a full-resolution crop from an X-T1 with the Fuji 10-24mm zoom at 10mm;
(3) a full-resolution crop from a Nikon D800E with the Nikon 18-35mm zoom at 18mm.
I applied the Lightroom lens-correction module to the D800E image; it's my understanding that LR automagically performs lens corrections on files made with all Fuji lenses based on manufacturer's makernotes included in the metadata. The D800E image also had the sharpening-amount slider set at 25, but the sharpening detail value was 25. Both crops represent my typical starting point for post-processing images made with the respective cameras. (Obviously, the dimensions of the D800E sample are larger because of the higher resolution of the sensor.) Both look equally clean to my eye.
I have used the following sharpening parameters to my Nikon D800E, D810 and also Canon 1Ds III and 5D III and it works very well I think and brings out visible more detail than closer to default. They are amount=50, radius=0.8, detail=70 and masking 30. This is for apertures up to f/11. For f/16 and beyond I use detail=100 and the other parameters the same.
I have downloaded a number of Fuji X-T1RAW files and the same parameters worked well. Here is one example that included the sky and where I had brought exposure down by -1 and highlights to -100 and it looks good to me in 1:1. 2nd attachment full picture and 3rd with no adjustment except sharpening.
-
interested if Lemondixon had any more comments on the processing of X Trans files.
It sounds like you were getting good result, did you find the holy grail settings in lightroom?
Apologies for the delay, but (as you will see) this is not a quick or easy explanation to give, nor something you're likely to easily accept (at least until you've done the Pepsi-challenge on some of your own images and seen the difference).
Basically, there IS a Holy-grail of extracting detail from all your images (not only the Fuji X files) and, in part, this comes down to better sharpening technique and tools (as well as other techniques), however what I have learnt is that (as much as I would like it to be so) you ain't gonna get this from ANY of the RAW converters out there at the moment (though there ARE differences between them, even WITH these other techniques I will point out for you).
I guess, right now (like I did) you're groaning as this isn't what you want to hear and at the thought of having to buy more software, or not being able to do everything all in the one place. Let me make it clear - I LOVE Lightroom, and so wished this wasn't the case, but don't fret; It's not all bad news. Let me first set out some important points (before you shoot me down or dismiss this all):
1. You can STILL use LR (or your preferred RAW converter) and have a slick workflow that won't be time consuming;
2. You're likely to already have all the software you’ll need, assuming you have any version of Photoshop from CS1 or upwards (it may still work on older versions, but I'm not 100% sure);
3. You can download all the comparison images here: https://www.dropbox.com/sh/175lh66sfkeplvx/AABhdyUZejJO0pUY3aBHn9DGa?dl=0 (https://www.dropbox.com/sh/175lh66sfkeplvx/AABhdyUZejJO0pUY3aBHn9DGa?dl=0)
4. Yes, I KNOW the foreground in the image I used is out of focus (so please don't criticise), nor is it an aesthetically pleasing image - However, it was used purely as it demonstrates the cursed LR "painterly" effect with large groups of trees / foliage / repeating detail;
5. I have tried these same techniques on lots of other images and the results were the same, including improvements in skin tones/details in portraits and landscape shots with fine detail in gravel etc in the foreground;
6. You also MUST, MUST, MUST remember that images can only truly be compared in the context in which they are finally intended for (print, web etc) and that, if some images seem over sharpened when zoomed in to 200-300% this is NOT how they look printed at 240-300dpi from normal viewing distances (they look fantastic), but more importantly they still do NOT contain any extra visible noise or generated artefacts if you use the best techniques (i.e. not a RAW converter);
7. The author of these techniques (Guy Gowan - see his website www.guygowan.com (http://www.guygowan.com)) has persuaded me to abandon the whole input and output sharpening route - Just do it once, but do it really well;
8. No, I'm not doing this as publicity for GG, nor am I paid or induced to promote his stuff - I simply became a member of his site last November (2014) and it has COMPLETELY changed my attitude to what is achievable with sharpening, contrast, colour correction etc. In fact, it was only as a result of searching for clues on how to extract the best detail from Fuji X files that I came across his site;
9. Finally, remember that you WON'T NEED to do these techniques on EVERY image - just the ones that are your final, best picks for display etc - So you can still make brilliant use of LR (or whatever) for managing your library, editing a shoot, output to print (or other media), and even some final tweaks to the image (if you really want/need to). The workflow using LR and GG's techniques if particularly slick I must say.
Let me first explain what can be seen on the various images in the above link - Aside from the full, original RAW / .RAF file for the image used, in the subfolder there are Jpegs made carefully (with like for like settings where possible) for direct comparison:
a) Original Out of Camera Fuji X-E2 produced Jpeg (+1 Sharpening, -1 Highlights/Shaddows)
b) The RAF Raw file processed with LR Default settings (sharpening / noise reduction etc), then converted to Jpeg.
c) The same file converted to DNG in LR , same settings (to check no difference to (b) above
d) The DNG file at my optimum LR settings (as per previous posts) and the RAF file with same settings (again for comparison of DNG conversion process effects, if any), then to Jpeg.
e) The RAF file output (with LR settings for sharpening and noise etc at ZERO) sent to Photoshop from LR, and then “processed” using Guy Gowan’s techniques, then converted to Jpeg
f) The RAF file output as above to PS but from Iridient Developer (ZERO sharpening etc), “processed” GG style as above, then to Jpeg
g) RAF file processed in ID only, but at optimum settings (per previous posts), output to Jpeg from there
h) RAF file output same as (e) but from Capture One to PS, “processed” GG style as above, then output to Jpeg; and (finally!)
i) RAF file processed in Capture only but at optimum settings (per previous posts), output to Jpeg from there
Given the length of this post already (and the limits of your patience), I won’t go into great detail on how GG’s processing works (though I will in another post if requested) but the basics are:
You can push files far, far further in terms of sharpening, contrast and dynamic range expansion without ANY banding, halos or introducing artefacts IF you use properly constructed masks (appropriate to the subject matter - e.g. protecting skin tones). You can get true HDR without it looking like some weird caricature of reality, as in Sin City movie posters! Buy hey, if that’s your thing then no problem. His objective is to give you what your camera is capable of producing, instead of the heavily compressed narrow dynamic range all manufacturers assume we want as “consumers”.
These masks (unfortunately) cannot be created in ANY RAW converter, and are based on knowledge of colour theory and many traditional techniques used in film darkrooms (which I used to do, and can easily relate to) and lithographic printing. Their beauty is that they are calculated or generated automatically, not by manually drawing selections, and so there are no harsh transitions. They can also be cleverly targeted to focus the processing effects in specific areas only and restrict/prevent other areas being influenced (such as shadows, highlights, skin tones, skies, etc). The learning curve can be pretty steep at first, but it is well worth the time invested.
Better still, these “processing” techniques can be applied in an automated batch method and still give consistent results without having to manually adjust sliders (“wanging” as GG calls it!) for every individual image. This will save you hours, and hours of time, leaving you free to concentrate on final re-touching and adjustments which you will find much more satisfying.
In terms of detail and producing realistic looking shape, contrast and texture in your images, these techniques far, far exceed what any of the RAW converters can achieve in terms of highlight and shadow recovery - a big claim I know, but I have seen it proven and done the same comparisons myself. You can also put on more contrast than you ever thought imaginable (without colour shifts) and an incredible amount of sharpening (his standard is 500%). In retrospect, the example image I have used maybe isn't the best but I'll happily demonstrate this on an image of your choosing.
GG does produce the “Process Action” that does all of the above and allows you to modify the results both image by image, as well as setting up batch processing tailored to suit a particular set of images and your own tastes. This is updated regularly (weekly/monthly) with ongoing improvements. However, if you’re just after a ‘quick fix’ solution without truly understanding the mechanics of what is going on, this may not be for you and you won’t get the best out of it. GG does not sell these Actions as a ‘product’ - he only teaches the techniques to members, who pay an annual subscription, so that they can also do this for themselves as well as make the best use of the Actions he creates and gives out to members (without charge).
If you’re a member over at Fuji-X Forums, he does a 25% subscription discount. Personally, having paid for lots of professional tuition, what I paid was less than most half day courses and about 500 times more informative and useful. I’d got my money’s worth by end of week one. Be warned however, Guy’s style is sometimes a little abrasive and he’s not everyone’s ‘cup of tea’. If you get passed that, you won’t care at all even though he does have great contempt for a lot of what software producers have sold to us as being the only or best way to do things since the digital age took over!
Anyhow, this is not meant to be a GG promo. Look at the images (even if it’s not the best example I could have used). As you will see, ALL of the ones run through his “processing” method are massively more detailed and sharper (and I could have got a lot more out of the shadow detail if needed). All the Jpegs from the camera and other 3 RAW converters (even at optimum settings) look like they’ve had a diffusion / blur filter on by comparison. What you will note is that there is still some difference between each of the RAW converters’ images even after GG processing, so clearly there are differences still in the demosaicing algorithms and you may argue that ID is still best? At the end of the day, there are limits on my needs and, all in all, I am content with the LR results once all sharpening and noise reduction has been switched off and then processed properly using GG’s Action.
Finally, if you would like, I would be happy to produce and publish versions of any of your own images for comparison if you contact me and supply the RAW file. Sorry for such a long post, but hopefully you will find it useful.
-
Finally, if you would like, I would be happy to produce and publish versions of any of your own images for comparison if you contact me and supply the RAW file. Sorry for such a long post, but hopefully you will find it useful.
So, basically, what all that wall of text of yours says is, you have to pay that GG guy to share his magic workflow with you? Sorry, not interesting in esoteric elitism.
You keep writing about his "processing techniques" ("wanging", and whatnot), but fail to explain what these are. Are shameful promotions like this even allowed on the LL boards?
-
Hang on a minute phl0wtography. Please don't take that tone. Try to be civil and polite. It is really not appropriate or helpful.
If you care to read the whole thread, you will see that I was involved in this discussion before I'd ever heard of Guy Gowan and secondly, after some productive dialogue with other members about their findings I was actually asked to share what I later discovered by several members of this thread. It was a real struggle to eventually find the time, but after spending a lot of time putting together what I had been asked for (and I even apologised for the length) I really don't appreciate it being described as a "wall" or "esoteric elitism", nor is it a "shameful promotion".
I made it perfectly clear that I would elaborate on the details if requested, and even offered to show the results on other peoples images. If I did explain the details in that posting it would be even longer! But hey, if you're "not interested" then why even attempt to read what others have taken considerable time to share or add your unhelpful comments. Either go do your homework, actually read and take on board these details and make informed comment, or keep those sorts of negative remarks to yourself. People come to forums like this to learn and share information and are only put off doing so by rude and inappropriate comments like yours.
-
So, basically, what all that wall of text of yours says is, you have to pay that GG guy to share his magic workflow with you? Sorry, not interesting in esoteric elitism.
You keep writing about his "processing techniques" ("wanging", and whatnot), but fail to explain what these are. Are shameful promotions like this even allowed on the LL boards?
This is an uncalled for response.
Imaging processing can be simple or complex. Different strokes..., etc.
It isn't necessary though to be rude when someone has spent a lot of time explaining one of the possible approaches. If it isn't to your taste, fine. But unnecessary rudeness isn't.
Michael
-
Although I am not going to spend the coin to get GG's elixir, I for one welcome all discussions about how to improve output from the X cameras, and therefore give my thanks to Lemondixon.
Truth be told I hope that LR 6 has some special sauce for X-Trans files since I have gotten lazier and lazier about taking images into other programs as LR has improved....
-
Thank you Lemondixon for taking the time for a detailed reply.
As I understand it you are processing the RAF file in the converter of your choice to minimise artefacts and maximise highlight detail. Then sending the tiff to Photoshop and the Gowan actions to recover shadows, improve highlight detail and restore sharpness.
I had a look at one of his presentations. I think he's an annoying man and goes on a bit but I also think he's correct on a number of fronts, particularly that process version 2012 in LR can compress highlight detail. As a result for images where highlight detail is critical I use the exposure slider in process version 2010 to retain highlight detail and use an existing toolset in Photoshop to recover shadows and sharpen (namely LightMachine for shadow recovery and Topaz Detail for, well, detail).
So I wasn't convinced enough to sign up, but it's good to be reminded that there is more than one way to skin a Felis catus
David
-
Thank you Lemondixon for taking the time for a detailed reply.
As I understand it you are processing the RAF file in the converter of your choice to minimise artefacts and maximise highlight detail. Then sending the tiff to Photoshop and the Gowan actions to recover shadows, improve highlight detail and restore sharpness.
I had a look at one of his presentations. I think he's an annoying man and goes on a bit but I also think he's correct on a number of fronts, particularly that process version 2012 in LR can compress highlight detail.
I agree, and on that aspect Guy Gowan is correct, but then many others have commented on that Process 2012 behavior/trade-off, and learned how to correct it if the scene requires that. Reducing the Highlight slider control (even to -100) solves a lot of that loss of highlight detail from compression and highlight recovery.
But apart from his annoying attitude (in an attempt to portrait himself as superior), he is also very much stuck in his old concepts of sharpening, dating back to his pre-press days where film scans were used as source material to produce screening rasters. All that edge preservation masking stuff is mostly needed when using old technology, like USM related sharpening and blurring.
Welcome to the 21 century, where we now can use digitally captured images, very effective noise reduction (even to the point where we need to reintroduce some noise), and use halo free spatial frequency based tools, such as Topaz Clarity for contrast adjustments, and Topaz Detail for detail enhancement, both with built-in edge-aware masking if needed, and color technology that prevents saturation issues when changing brightness and contrast.
Cheers,
Bart
-
Thank you all for the reassurance I hadn't completely wasted my time supplying such a lengthy post.
Yes, I agree about how Guy comes across - which is such a shame, as it puts so many people off what is otherwise a really valuable source of information. I deliberated for quite a while before stumping up the cash to join (because of similar comments on Fuji X Forums) but I'm glad I did. I don't have to like someone to find their ideas valuable. Fortunately, I've always been the sort who needs to understand the how and the why of everything, so getting a much deeper insight into how contrast, sharpening and range expansion can be done (without just trusting someone else's algorithm) was really, really useful for me, and many of the same masking principles will improve a lot of other stuff that I found unsatisfactory before. There is no secret "elixir" however - everything is explained in great detail, it's unfortunate he refuses to share any of what is going on under the hood in any of his free to view videos, like the 'Focus' episodes he has on his website. To be honest, this only made me sceptical about what he really had to offer when I first looked at them and reluctant to commit to joining in case it was just hot air! However I don't object to paying for useful professional training - That's what he makes his living at and I wouldn't have found it otherwise.
David Sutton: That's kind of right - It involves no real processing of any sort in your RAW converter of choice, just conversion to TIFF or PSD in the right colour space with sharpening, noise reduction etc turned off and, where you have highlight clipping (first removing auto-recovery in LR by switching back to process version 2010) these are recovered by reducing exposure only.
The PS Action set then brings back shadow detail (amazingly well, and with much better shape and detail than using any RAW converter) and expands the dynamic range, de-moire, cleans and boosts the colour channels (giving even more detail), reintroduces and boosts contrast to levels you never thought possible, and finally sharpening. All of this can be automated to give you a fantastic starting point on all your chosen images without having to do them each by eye.
It is also really interesting for those who either shoot for the highlights or shadows. His method actively encourages over exposing your images (depending on the camera) by around 2 stops, before bringing back highlights with exposure in your RAW converter. This way, whilst the shadows become really dark before using the process action, because they were correctly exposed (or nearly) at the point of capture, the image imported to PS has the much less or no noise in the shadow regions before you process the image and bring back those areas. The result is what we all want - perfectly exposed, detailed highlight and shadow regions with little or no noise (and without having to splice multiple exposures or versions of the same image).
Don't get me wrong - I wish I could do this all in LR, but once you have export presets set up in LR using PS droplets you can send off images to PS and have them automatically imported back into LR (either as flattened files, or as PSD's with all the layers so they can be re-tweaked later if required). This makes for a really slick workflow.
It would be interesting to do some comparisons (if any of you would like to) between ID, LR, Topaz and Lightmachine etc to see just how much each of these can squeeze out of an image without introducing halos, noise or other artefacts. Anyone up for that?
-
By using the 2010 process you would lose the film simulations, right?
For me colors are as important as the sharpness and I love not having to work much or at all on the Fuji files color vs the Nikon thanks to those simulations.
That's why I keep using the LR vs others.
-
Never mind, I see you can keep the color profile even with the old process.
-
Lemondixon, thanks for the write ups. I too was not pleased with the way GG comes across in his videos and the ones he did for the Fuji X forum, it really put me off from looking further. Which is a shame as I would like to see how it works and if it works well it would be worth it to me. I will download your sample images and take a look. I might be interested in seeing what kind of comparisons we can do. Thanks for making the time for the lengthy post and for posting the samples. Your time was not wasted but I still have this feeling of it is just "magic" from the description but that is probably from the bias that has formed from watching is "advertising" intros which really do not explain anything. So again thanks for being patient with us.
Alan
-
Thank you, Lemondixon for sharing. I truly appreciate it.
I've kept an open mind and I've been going through the free "FOCUS" videos on his website to understand where he's coming from. It all seems to be a rational approach if you take the time to listen and learn where he's coming from.
For those interested in learning more about his approach without paying anything start with his first video from "FOCUS" on RAW (01)
http://www.guygowan.com/focus/video.php?&userid=170
I haven't bought anything from him yet but I'm learning. I too have a long history in the B&W darkroom with a smattering of color. I'm never going back to film as I'm addicted to what I can do with digital cameras and digital processing. 15 years of digital is now ingrained in me.
-
Not sure which film simulations you meant Armand (VSCO or other Presets made for Lightroom?) but either way there is a really simple solution that doesn't involve having to use the older LR3 versions of VSCO Film. I see you found an answer, but just so everyone else understands:
GGowan recommends using Process Ver. 2010, so that LR doesn't apply 'auto-recovery' on your images (which actually does make a huge difference on some images). However, this is only about extracting the maximum info/detail out of your images during the RAW conversion process for export to Photoshop - What you do with it after you've "processed" it using GG's techniques is up to you.
So for those that are your 'final' images for printing etc (not everything in your library):
1. In LR switch back to Process V. 2010, switch off sharpening, noise reduction etc and then adjust Exposure until clipping gone;
2. Export image to Photoshop and process using GG's method for de-moire, auto-levels, range expansion, contrast and sharpening;
3. Flatten image and close, opening back in LR5 (usually as a TIFF or PSD);
4. If you then wanted to apply any film emulation effects (VSCO or other), you can do this on your processed TIFF/PSD using whichever Process Version you like, as by this point you've already got all the highlights and shadow areas and sharpness to where you want it
One point to bear in mind however - For some of the more extreme Film Effects which can really close down the shadow areas or highlights with heavy contrast, you may want to go easy with adding contrast at the GG processing stage, and leave the image slightly flatter than normal so that you've still got plenty of room to play with when applying the film effect
-
These comparisons are stunning!
The (e) file has an incredible amount of detail and beautiful tones.
Thanks for doing this comparison so thoroughly. I know it's a lot of work, but, like you, I'm fascinated to directly compare these variables in photography and really get a grip on the best technical approach.
Very interesting.
-
I don't have a Fuji, was just passing thru, but thought I'd give RawTherapee a try against Lemondixon's samples. I think it compares quite well from a detail/resolution perspective, and I'm quite new to RT, so I'm sure those more experienced could do better.
It's the 2nd image from the left in the top row.
-
BTW, something is really wrong with lemondixon's raw conversions (see his download file and my screen grab above). No one should have to suffer with such soft/blurry results.
On the screen grab below, top left is Raw Therapee, bottom left is Lightroom, bottom right is PhotoNinja. The other three are the "Holy Grail" GG results.
And again, this is my first time processing any Fuji file and I'm a newb to Raw Therapee and PhotoNinja. Further practice should yield better results.
My point is, while those GG results are "fine" as is, they're nowhere close to what I think one could achieve with better staring points out of their raw converters before going to PS.
-
I think its useful to point out that G Gowan's base premise for sharpening is that contrast is required to create three dimensional images.
His approach is based on using channels as the key element for finding the areas in an image for a specific sharpening formulation.
He also uses blending modes to apply one sharpening formulation for highlights and another for shadows.
If one is willing to tune out his presentation style and focus on the technical approach there is much useful information.
Is it required for every image, no. Is it a useful technique, yes.
He is convinced that Photoshop, because it has channels, is much more capable than any raw converter, for getting maximum sharpness, shadow detail, dynamic range and tonality from an image.
-
As some have mentioned not liking the color that is produced by Iridient Developer, I wonder if anyone has had any experience with XT-1 color after using the Fuji camera profiles that come with Lightroom or that were created by the Xrite Color Passport. It is my understanding that Iridient can import those profiles and use them instead of the default profile.
Thanks
Jed
-
Provia it's very close to the real thing, I posted a comparison here some time ago. It deals very well with the more difficult red to magenta spectrum. It's not identical to the in-camera but very close.
Velvia it's a little farther away but still very close, not as good as Provia though. Don't use much Velvia to be more specific.
-
. . . I wonder if anyone has had any experience with XT-1 color after using the Fuji camera profiles that come with Lightroom . . .
I haven't any experience with the in-camera emulations since I don't shoot JPEGs, but at least on casual inspection the Velvia profile looks very much like Velvia 100 film to me.
-
I was pleased to see the following item in the June Lightroom CC release notes (http://blogs.adobe.com/lightroomjournal/2015/06/june-updates-to-cc-photography.html):
- Reduced “color blur” artifacts when processing Fujifilm XTrans raw images (http://feedback.photoshop.com/photoshop_family/topics/lightroom_and_fujifilm_x_trans_cameras_halo_around_blue_areas) – In collaboration with Fujifilm, we are still investigating methods to improve fine detail rendering and overall edge definition.
It's helpful finally to have an authoritative statement that Adobe is working with Fujifilm on this.
-
I was pleased to see the following item in the June Lightroom CC release notes (http://blogs.adobe.com/lightroomjournal/2015/06/june-updates-to-cc-photography.html):
It's helpful finally to have an authoritative statement that Adobe is working with Fujifilm on this.
Thanks for the information. I had noticed a difference, or what I felt was a difference in LR CC conversions on XT-1 files. I do hope they work on the fine details rendering, that would be great.
Paul Caldwell
-
Thanks for the information. I had noticed a difference, or what I felt was a difference in LR CC conversions on XT-1 files. I do hope they work on the fine details rendering, that would be great.
Paul Caldwell
One person on the Fuji rumors forum is reporting that the bug fix has fixed the high contrast edge problem (my bugaboo), which if true for all Fuji images would be good news indeed.
UPDATE:
A recent post on Fuji Rumors with 100% crops shows noticeable reduction in haloing at black branch against blue sky, so that's promising.
-
That alone would be huge.
Glad to see Adobe and Fuji working on this together.
Paul
-
Took me a while to find this thread.
For a long time now my default x-trans sharpening was starting with detail 100%.
These days I realized I get more natural results if I keep the detail around 60% and work with the amount more. I'm not sure if it's related to the new LR or the new x-trans or I was doing wrong all along. There is a point for higher detail and convoluted sharpening but rarely.
-
As I went back to reading this article previously linked here: http://petebridgwood.com/wp/2014/10/x-trans-sharpening/ I noticed there is a small addendum stating the details works better at 60 for the 24MP x-trans. I guess great minds think alike :D
-
When I first got the X-T2 I couldn’t get reasonable sharpening from Lightroom like I could with my X-T1 and Peter Bridgewood’s suggestions.
I started using Iridient Developer and then Iridient X-Transformer. The demosaicing seemed far superior to Lightroom. I still use Lightroom but Iridient is now part of my workflow. Others have said that Lightroom has improved in the last 2 years but I still think I get superior results with Iridient.
-
While playing with the new enhance details in LR I realized the processing in LR must have gotten significnatly better at some point because I don't see much artifacts these days, I have to look for them to occasionally find them.
For raf files in particularly the enhance details occasionally, as in less then 50%, will increase the apparent resolution but on some of those it also comes with increased saturation and I wonder if the better separation between different colors (sec to increased saturation) accounts for some of that perceived sharpness. I didn't run any objective tests for sharpness nor do I intent too.
I will start posting some comparison crops from time to time, maybe we'll figure it when it's worth trying because it does come at the cost of much bigger files.
In this one the old artifact is apparent at 100%, but I doubt it will see it in a normal to large print. You can see the difference a little better at 200%, enhance on the left. If I didn't have the comparison probably I wouldn't have been bothered at all by the original appearance. Printscreens attached.
-
I would agree, the "enhance details" option on the X-T2 and X-H1 files is most impressive. The base raw conversion I feel shows a bit of improvement, still see a bit of wormy or plastic look in the rocks, but when you run the enhance details on the file most of this goes away. Worth the extra effort.
Paul C
-
I've tried the new "enhance details" but in the end I preferred Iridient X-Transformer for sharpness and detail on images that require it. Many images of mine do not require that level of detail though.
-
For [Fuji X-Trans] raf files in particularly the enhance details occasionally, as in less then 50%, will increase the apparent resolution but on some of those it also comes with increased saturation and I wonder if the better separation between different colors (sec to increased saturation) accounts for some of that perceived sharpness. I didn't run any objective tests for sharpness nor do I intent too.
As I understand it, one of the more counterintuitive features of machine-learning is that the authors of the software aren't always able to determine what attributes in the source image the neural network has isolated and used to produce the result in the target image.
We do know that Adobe used a special training set for X-Trans files:
Enhance Details uses an extensively trained convolutional neural net (CNN) to optimize for maximum image quality. We trained a neural network to demosaic raw images using problematic examples, then leveraged a new machine learning frameworks built into the latest Mac OS and Win10 operating systems to run this network. The deep neural network for Enhance Details was trained with over a billion examples.
Each of these billion examples contained one or more of the major issues listed above that give standard demosaicing methods serious trouble. We trained two models: one for the Bayer sensors, and another for the Fujifilm X-Trans sensors.*
So presumably the enhance details feature is especially sensitive to the objectionable artifacts that are sometimes produced when Lightroom's generic sharpening algorithms operate on X-Trans files.
I haven't performed any rigorous tests, either, but my plan is to keep Iridient X-Transformer handy and try it as well as enhance details on Fuji images that don't respond well to standard Lightroom sharpening, then select whichever looks best. Somewhat cumbersome, but without knowing exactly what is happening under the hood in Lightroom and X-Transformer, I think it's going to be impossible to predict in advance which of the products will produce the better result.
———
*Adobe Blog, Enhance Details, by Sharad Mangalick, Feb. 12, 2019 (https://theblog.adobe.com/enhance-details/)