Saturday, March 2, 2013

Lightroom 4.4RC and Capture One versus the X-Trans sensor

Over the past few weeks, support for Fuji's X-Trans sensor, as found in the X-Pro, X-E1 and the new X100s has gone from minimal to quite extensive. Phase One have for the the first time shipped a product with support (Capture One V7.1) and Adobe (with Lightroom 4.4RC and Adobe Camera Raw 7.4) have improved their X-Trans support. In addition, AccuRaw is close to release as a commercial product.

In previous blog posts (herehereherehere and here) I've compared the then-current raw developer options, and discussed why demosaicing the X-Trans sensor is difficult. On various photo forums, Capture One has had a largely enthusiastic reception, while Adobe's offering has had a more mixed reception, some praising it, but other complaining of softness, blown blue channels and various artifacts.

How do they really stack up? Let's see:

Important Disclaimer: For those that don't know, AccuRaw is my product. So I'm biased.

Lightroom 4.4RC versus X-Trans

The previous blog posts have extensive detail about the older versions of Lightroom and Adobe Camera Raw's problems with the X-Trans sensor. Using the same section of the image I used in the previous blog posts, here's LR 4.4RC versus the previous version (all images processed with default settings unless otherwise specified):

The latest: Adobe Camera Raw V7.4RC, 400% crop

The old version: Adobe Camera Raw V7.1 beta, 400% crop

So, an interesting result. Compared to the previous generation, the new Adobe algorithm has much less obvious chroma smearing, so it certainly is much improved. Taking a closer look, where previously the smearing was really bright and intrusive, in the new version the smearing is a lot less bright. However, there's actually more smeared pixels - in effect, the smearing now has a wider radius. In addition, the image is noticeably softer than the previous version.

Capture One V7.1 versus X-Trans

Phase One's Capture One is a new player on the X-Trans block. Here's what it delivers:

Capture One V7.1, 400% crop

Another interesting result. Still some smearing, notably inside the letter "A", but not nearly the level seen in the Adobe results. Saturation is good, expect at the tips of the letters - e.g., look at the tips of the "F", and compare to the AccuRaw "maximum resolution" image below. Image sharpness is "interesting"; the way the image looks to me is as if what Phase One have done is to add some extra sharpening on top of an image that probably wasn't too sharp to start with.

SILKYPIX versus X-Trans

We saw SILKYPIX in the previous posts:

SILKYPIX conversion, 400% crop

Previously the best of the breed, some chroma smearing, saturation down, resolution appears slightly reduced

AccuRaw versus X-Trans

AccuRaw, now in its Release Candidate form, delivers the following:

AccuRaw RC1: Maximum resolution settings, 400% crop

AccuRaw RC1: 60% luma and chroma artifact suppression, 
20% post demosaic filtering, 400% crop

In the first crop, set for maximum resolution, AccuRaw gives very good results on the red letters, but has some artifacts. However, with AccuRaw, you can tune the result to what you want. The second crop shows moderate artifact suppression settings - still not as much chroma smearing as the other raw developers, but much reduced artifacts.

Conclusion - How do they stack up?

Firstly, Adobe's products, even in the new LR 4.4RC/ACR7.4 form, still don't stack up. Although much improved over the previous generation, they still have excessive chroma smearing relative to image resolution.  If you were to select a raw processor purely on the basis of getting the maximum out of your X-Trans based camera, Lightroom wouldn't be it.

The other products are much more evenly matched - in my view, technically they're at the same level, just making slightly different choices as to the trade-off between chroma smearing and resolution that the X-Trans sensor brings with it. AccuRaw has the advantage that you can adjust that trade-off to suit yourself and the nature of the image - e.g., for landscapes you can generally use the "maximum resolution" setting because artifacts won't show. However, AccuRaw doesn't have the features that either Capture One or SILKYPIX have. 

So really, it's a choice, and that's a big win for users, and a massive step forward from just a few weeks ago. Nine months ago, my comment on the X-Trans was that with a good raw developer, it was almost as good as a conventional sensor. And "almost as good" is actually mostly enough - practically, with the new raw developers, the difference between a conventional sensor and a X-Trans sensor is small enough to get lost in differences in lens performance, etc. There are now enough good raw developers that most users will be able to find one that works for them.

If the users are the winners here, who are the losers? Adobe certainly haven't covered themselves with glory - they have huge reserves of money and probably the best engineering talent in the business, but don't seem to have been able to apply it. Fuji is also a loser. It's ten months since I first blogged about the X-Trans processor, and so far it's delivered nothing to justify the "greater resolution than conventional sensors" hype. Finally, the really big losers are the many camera "reviewers" out there that uncritically repeated Fuji's claims about the X-Trans sensor's greater resolution. To their credit, some reviewers did raise warning flags - Sean Reid and Thom Hogan to mention two, but they were the exceptions. So next time you read a camera review, here's a suggestion - take look at what they wrote about the X-Pro when it was introduced, and judge accordingly.


fatboyslimmed said...

Well, I just got my XE1 last Wednesday. Today was the FIRST true test -- I went to shoot my 2 year old grandson. I had my settings down after having read through just about every blog on the xpro site. I shot all the way up to 3200 ISO and the images were, well ---- my D800 is up for sale as of tonight --

I am keeping my D700 and assorted lenses, but this camera and the 35mm F1.4 were truly amazing. Crisp, clear, sharp images even at high ISO. I use LR, version 4.3 and I agree that LR just doesn't have the oomph to handle the raw files, BUT the jpegs are so freakin' amazing that I honestly can't compete!
So bravo Fuji!! They have done an amazing job.

Brad Horn said...

I downloaded AccuRaw after reading this. There are two links below to screen shots of one of my shots. This started out as a X-Trans RAF file and was converted to TIFF in AccuRaw.

These screen clips are at 400%. Although the shots look fine at normal resolution, the RAW converter shows artifacts on the edges of high contrast areas. It also appears to have a repeating pattern of color artifact in these high contrast areas.

I'm so glad that manufacturers are working on solutions to the challenges provided by the X-Trans sensor. In my mind, it is the best sensor on the market for what it is.

Eric Kelly said...

I think it would be useful to see comparisons of the various conversions with the out of camera JPG's. People are raving about the out of camera JPG's but I would like to see if there is any advantage to shooting RAW...

70de0bb6-b68b-11e2-92f0-000bcdcb5194 said...

Hi Sandy,

I purchased Accuraw, and I've been enjoying it. Thanks!

I was wondering if you've tried Aperture, yet? I've been using it lately, and while it does have some chroma artifacts, I'm finding that it could be best overall, in terms of trade offs. Any experience with that yet?

Sandy said...

I'm glad AccuRaw is working well for you.

I haven't yet tried the new version of Aperture; hopefully I'll be able to do so in the next few days. I've been very busy with a new version of AccuRaw that should improve speed and general responsiveness. But it's taken a lot more work that I'd hoped :(

Sterndal said...

I read a lot of literature on spectral demosaicing of random CFAs giving similar peak S/N ratio with less artefacts. Artefacts mostly shift towards chrominance noise, which is more pleasant to look. Typically all scientific paper look at the Kodak lighthouse picture and its fence to spot CAs, for example. Did you implement similar algorithms in your software, or is it done using spatial interpolation?

Sandy said...

The Kodak images are indeed very commonly used in the scientific literature. Unfortunately, they're also a very a poor match to the characteristics of modern CFA sensors, so I don't use them. The exact details of the X-Trans algorithm is confidential, I'm afraid.

John Leslie said...

I mentioned your series of X-Trans posts to Thom Hogan as he was complaining about X-Trans colour bleeding. He replied that he prefers Iridient Developer for X-Trans. Might be worth a look?

John Leslie said...

Allegedly ( ACR 9.1 has some X-Trans improvements.

Michael A Shapiro said...

I've been digging around your site, since reading Sean Reid's newest post. I'm trying to find information on the degree of correction with an x100T. All I find is that the demosaicing process is difficult. Have you accomplished it? Can you comment or show examples?

Sandy said...

If your question is how much lens correction an X-100T has, then there's an easy way to see for yourself. Just compare the rendering from a raw processor that doesn't implement lens correction (e.g., PhotoRaw or AccuRaw), with one that does (e.g., Lightroom)

Michael A Shapiro said...

Does ACR apply it without being asked? How can I try out AccuRaw?

Michael A Shapiro said...

I found the answer. There is some distortion, but ACR has a checkbox to enable/disable correction. So, does that mean that the optional correction is from Fuji or from Adobe?

Sandy said...

Well, only Adobe knows that for sure. But I'd guess that the optional correction is more from Adobe than Fuji.