X10 RAF files having wrong dimensions in darktable

Hi everyone,
I just started working with Darktable and seem to have run into some unexpected behaviour…
Imported RAF files from my Fuji X10, they seem to be read with wrong Image dimensions.
The Information tab tells me a size almost 1000 pixels less in width than it should be, same goes for exporting.
does anybody know how to work around this or set the correct dimensions manually inside DT?

→ DT 2.6.2 on Win 10


jpg or raw files?

what format did you set in the camera settings for capturing?

Camera is set to the largest possible, I shot JPEG and RAW and the JPGs do have their correct size.
Opening the raw in Fujis converter also yields the correct size.
Only the RAF files opened in Darktable are interpreted wrong.

Can you provide a sample image? What version of darktable?

No problem - here we go:
DSCF8004.RAF (18.8 MB)

(example) → this should be 4000x3000 or close by if we keep barrel distortion in mind, but it is displayed and exported as 2944 x 2144.

Camera JPGs are correct and so are any files exported from silkypix .
As stated above, I am on Windows 10 using DT 2.6.2

This is either a setting, my incompetence with the new tool or a bug with these quite old raw files.

Thanks for the help!

[MakerNotes]    RawImageWidth                   : 2848
[MakerNotes]    RawImageHeight                  : 2144
[RAF]           RawImageFullWidth               : 2944
[RAF]           RawImageFullHeight              : 2144

from exiftool output, so this isn’t raw loading problem.
I suppose the question is, as per the camera specs, what is the sensor dimensions?

which is exactly what the JPGs do represent.
Windows Meta info shows this aswell :

The resolution it tells can happen with EXR Mode, where it is increasing the dynamic range by using only 6MP, but this is definitley always off.

Hi - I’m new here but think I can shed some light on this as a former X10 owner.

The X10 is an example of Fujifilm’s fondness for exotic image sensors. The EXR sensor used in the X10, nominally 12MP, actually consists of two 6MP Bayer arrays overlapped with each other and offset by one pixel vertically and horizontally. This needs special processing to generate a 12MP output image. The EXR RAF file actually contains two 6MP RAW bayer encoded images, which need to be processed and combined. Unfortunately I never found an open source RAW editing tool that could detect and combine both of the images to give 12MP, instead they just handle the first 6MP raw image in the file, which is what you are seeing here.

In some experiments I did manage to extract both 6MP images manually using a command-line raw processing tool (I forget the name now, sorry), but I didn’t have any clue how to combine them into the larger 12MP file with a higher resolution.

Sorry I don’t have an answer, but hopefully that explains why you are seeing this behaviour.

The X10 was (is) a lovely camera, but frustration with RAW processing was my main reason for abandoning it. Have fun anyway!

1 Like

Hey Steve,
I did read some similar articles on this matter with the same conclusion.
Looks like I am forced to extract my tiffs with Silkypix then - annoying, but working.

The Fuji is my everyday companion and still doing awesome Images for the non-professional line of this art for me and fun to work with - except for the raw issue.

Do you happen to knwo, if there is anything similar happening to RAW files coming from the GFX 100? (would be a bad no for it then)

Thanks a lot!

I’d like to add this link as reference:

Afraid I don’t know about the GFX100, though I think I read it has a standard Bayer array. I can say that Darktable works very well with the recent X-Trans sensors, as I am getting excellent image quality from a Fuji X-T20 now.

RawTherapee also can extract both 6 MP images and also allows to average them to increase SN ration in case the exposures of both images are equal.

According to tests of people I know, Adobe Lightroom can handle dual-frame RAF files like the camera software does to calculate the JPEG from the raw images.

The Adobe DNG Converter can read Fujifilm EXR-HR (I have an S200EXR), and if you set it up to use a custom format and enable “Linear (mosaic-free)” mode, it would read the diagonally shifted Bayer arrays like a checkerboard image and interpolate the missing pixels to a 24 MPx image; I believe that both the camera software and Adobe Lightroom will then scale down to the marketing resolution of 4000x3000 pixels (yes, the sensor has 12 MPx, but their diagonal interleaving does not allow the output of a 12 MPx image directly).

Are you still using darktable for these images? I have been considering getting a X-T20, but was a bit worried about support in raw processing tools (I tend to bounce between darktable and RawTherapee). Anything special you need to do to get quality from X-T20 images?

I have an X-T20, the image quality from darktable us good.

1 Like

I agree with @steve-d-mottram and @paperdigits. I have an X-T2, the same sensor and image processor as X-T20 and the image quality I get from darktable is great.
No need to do anything special …

1 Like

Yes - I’m using Darktable exclusively. I’ve found the X-T20 images easy to work with, mainly using Filmic, since the “Fuji-like” base curve seems to cause clipping on dark blue pixels. Using the “neutral” base curve also works fine. I found setting colour smoothing to 3 times in the demosaic module is good for reducing luminance noise as well as chrominance noise on the X-Trans images too. Otherwise I haven’t changed how I process much.

1 Like