FujiFilm RAF/RAW Process Interactive Comparison.


Just a small follow up on all the RAW experimentation I've been doing with the FujiFilm X-Pro1 RAF files.

Here's an interactive mouseover table to see the differences. Areas too look for are in the wood grain (you'll see a lot of colour or chroma noise in the different examples), Smearing of details (look at the texture of the walls to see how it gets smeared and creates the water colour effect), Aliasing issues (look on hard contrast edges to see a zipper aliasing effect), and lastly loss of details (in the red brick you'll see the various levels of details in the pores of the brick).


Mouse over the links below to swap. Click on the links to view 100% (opens into new window)

Some quick comments:

DCRAW 9.16: Gives the highest details, however has aliasing artifacts. Some chroma noise (even after filtering). Command line prompt only. Requires some technical know how. I use VNG interpolation and 15 pass median filtering.

1/2 Median: Is the filtering I apply to DCRAW to combat the aliasing artifacts. Requires a program that has Median filtering, and very process intensive.

Raw Photo Processor (RPP 4.7): Utilizes a proprietary method that seems similar to DCRAW but I haven't confirmed what extraction library it uses (I early assumed that it was DCRAW because every function that I've done in the program can be done with precisely the same results in woring with DCRAW directly). Works very well for details, but aliasing and chroma noise is high. The technique of 15 passes of median averaging in DCRAW clears up more chroma noise than what is produced by this program.

Graphic Converter (Patched): Patched with DCRAW 9.16 gives very similar results to RPP however this program is much friendlier to use than RPP and has some real nice post processing options. Could almost be used as an all in one solution.

In Camera JPG: The default standard. Still exhibits some detail smearing compared to something like DCRAW output, but has no aliasing or chroma noise artifacts.

SilkPix Developer Pro 5: The software that is part of Fuji's RAW processing uses an older SilkyPix engine. This current version is very clean however does show some chroma smearing (look at the green colour that appears under the window ledge and the loss of other colours). Shows very little chroma noise and no real aliasing errors. Very clean output and slightly softer than raw DCRAW output, but lot less aliasing even over the 1/2 Pixel Median filter. I would highly recommend this option if it weren't for the fact that the cost of the software is pretty high. 

FujiFilm Raw File Converter: This software bundled with the camera uses an older SilkyPix engine. It does a pretty good job, but the interface is very difficult to use and understand. Compared to the latests version of SilkPix it exhibits chroma noise and some detail smearing.

Lightroom 4.3: Has low amounts of chroma noise, but very heavy detail smearing. Loss of details but no aliasing artifacts. Even at 100% it's hard to see the 'Watercolour' effect if you don't know what to look for. I suspect a very heavy pre-demosaic median filter and bicubic interpolation algorithm is the cause of the issues. However it's still one of the easiest to use, and if you can handle some of the image quality loss, most likely won't notice the detail loss.


Comments

  1. Nice overview of our current options of RAF RAW-conversion. Much appreciated!

    Unfortunately it leaves me in a 'damned-if-You-do' & 'damned-if-You-don't' - kind of state;o)

    Really would like to (VERY soon Fuji!) see something happen between Fuji & just about any software-house in terms of developing some sort of complete solution that takes full advantage of the sensors great potential. I'm one of the early adopters of the system and have been struggling my a.. off with these conversions... Its frustrating to say the least.

    Right now I'm jumping between most of these programs You mention (except the dedicated mac-ones) depending on specific image but it would be sooo much easier to be able to enjoy a simple and homogeneous workflow.

    Best regards
    klehmann

    ReplyDelete
    Replies
    1. There's a rumour that Fuji is going to be releasing a firmware that will output to TIFF file format. For me and I think most would be perfectly sufficient.

      I too don't like jumping between so many programs, and of the major 3 packages only 1 will do that job.

      Ironically for me, I find that most of my images are rarely razor sharp, so most times, just using the default Lightroom settings seem to work for me. However when I do get razor sharp images, I will usually resort to one of these.

      Thanks for your comments... and hopefully a solution from Fuji soon.

      Delete
    2. Hi Terrance,

      The rumoured TIFF-output would indeed be a very welcome 'solution'. I'd still prefer taking full advantage of the RAF-file myself but at least initially it would make my daily workflow considerably more manageable. And that can only be a good thing.

      The biggest problems I face with current conversions are the 'watercolor-effect'/smearing especially in regard to skintones/complexion.

      Thanks for Your fast update/reply + all the hard labour You have put into these investigations which I have followed quite closely online since early beginning.

      Best
      klehmann

      Delete
  2. The Fuji X pro 1 raw processing is uncomfortable. The problem about several of the software you tested is that they are impractical. Is impossible to use all that in hurry during professional assignments. Silkypix or Fuji raw, the software that comes with the camera, are ok in the details but is very bad to recovery highlights, one of the main reasons that one use a raw file. The interphase is uncomfortable too and yes, is not very fast. Hope the rumor you commented will be true, but how much bits will have this file?

    ReplyDelete
    Replies
    1. RAF files are a pain, and it is saddening that there isn't a better solution out there that incorporates a smilier workflow for professional assignments (which is what my dSLR is still primarily used for).

      I can't say for certain what bit level it will be, but likely it will be 16 bit converted from the 12 bit RAF file.

      I don't know why they wouldn't just go to DNG, but I guess that would require licensing.

      I've heard a few sources now about the TIFF rumour or suggestion. Lets hope it comes true.

      Delete
  3. Hi Terrance,
    Very good overview to become crazy with raf-file from fuji!!!

    we all wait impatiently for the best solution of Fuji, and if it is even Tiff.

    many thanks for yours hugh work and time you spent for it!

    Best regards
    Jean Pierre

    ReplyDelete
  4. Thanks for the superb comparison.

    So, it's not supposed to be green below the 'window' ? Because to me that looked like the correct one, as it seemed to get some color out of there that the other ones just smudged away.

    ReplyDelete
    Replies
    1. Love to say that this was the way it was supposed to be. I actually went back there to see how much of that green was there, and actually it was more the Graphic Converter rendition of colours. You can actually see a slight green cast over the whole image from the SilkyPix version so I'm not sure what is up there, but it's not right.

      Delete
    2. Hi Terence, I've had an ongoing support ticket with ISL regarding the slight green cast present in Silkypix DS5 rendering of X-Pro 1 rafs, so its good to read that someones else is experiencing the same problem. Hopefully they will address it at some point, but trying to get them to acknowledge its presence has been very difficult.

      Delete
    3. Yeah John, that is good to hear another confirm that as well. Strange that they won't acknowledge it.

      I am not totally surprised of a green cast. I'm no software engineer, but I believe everyone is trying to use a conventional method of interpolation that does not take into account that the 6x6 array isn't just a neighbouring pixel to deal with. I do hope they acknowledge it and fix it. If you want to point them to this page with my samples, feel free to do so.

      Delete
  5. Terry I read this the other day at:
    http://mikemander.blogspot.fr/2012/11/distraction-new-fujifilm-x-e1.html

    "My basic Lightroom settings that impact X-Trans image quality are for sharpening and noise reduction. Generally for sharpening I use (33, 1.2, 100, 0) and NR I use (5, 50, 0 / 3, 50). Those are the slider settings I use for ISO 200 by default. If I crank up a ton of shadow detail, as in the night shots, I might increase the luminance or colour NR to taste. Normally, for a sensor without an AA filter and sharp lenses, I would tend to use something like (25, 0.7, 100, 0) as baseline sharpening, but the more aggressive sharpening seems to help minimize the "watercolour effect" to some extent. I'll also play with the radius and the amount a little to taste, but I always leave detail at 100 and masking at 0 unless the shot is at very high ISO."

    Steven

    ReplyDelete
    Replies
    1. Thanks Steven. I am not sure where he got his info in regards to this quote: "It is said that Fujifilm engineers are currently assisting Adobe to improve their raw conversion processing of X-Trans files and I do hope that is the case, however even now, with my own custom conversion settings in Lightroom, it is very rare that I am faced with a shot that has enough issues to make it unusable. "

      I have no inside information on that, but I can't imagine that they can do anything about it easily. Part of the problem is the baseline RAW processor in ACR is set up in such way as to work with traditional Bayer patterns. If Fuji was able to convince Adobe to customize their RAW processor, then good for them, but being a minority in numbers would certainly rule out them changing fundamentally how it works (bilinear interpolation is what I believe Adobe uses).

      Delete
  6. Any thoughts if issues of getting software players to use their RAW format is in part a reason for the shift to the new sensors in the X20, etc. (I'd chase up on this myself, but simply haven't time.)

    ReplyDelete
    Replies
    1. I don't know if that would the case. The X20 is also using an X-Trans colour filter.

      The complexities of decoding such a pattern is part in part being a 6x6 array vs a traditional 2x2 array. All raw processors are based on a 2x2 array of interpolating and decoding (with exception to foveon).

      It's actually good that Fuji is sticking to their guns. What I'd like to see if more companies license the filter array from them, which would mean that a new RAW processor would have to be engineered instead of using countermeasures to use the old Bayer based algorithms on the X-Trans filters.

      Delete

Post a Comment

Popular Posts