Beginner's attempts with Fuji X-Trans sensor

[Admiral]

Hi all, I would regard myself as a casual astrophotgrapher but very much an astrophotography virgin! I've dabbled around with DSS and stacked some frames from my mirrorless camera, and I've started looking at post-processing with Star Tools. So, without doubt, I'm on steep trajectory when it comes to getting to grips with these two applications.

Here's my problem. My camera uses the Fuji X-Trans sensor, which as you probably know is not a strict Bayer array, and has rather fewer red/blue pixels in relation to green pixels than the conventional Bayer array. The latest version of DSS doesn't seem to process the RAF raw files correctly, even though the demosaicing engine, DCRaw, is supposed to support the camera. So I can either convert the RAFs to TIFFs using my raw converter (Capture 1), using a linear response curve, or convert to DNG using the Adobe DNG converter, and import those.

Using the DNG files, with the default RAW colors adjustment settings of 1, 1, 1 (brightness, red scale, blue scale respectively) I get a very dark image and DSS can only find a handful of stars, even with a low threshold setting. When the autosaved FITS image is imported into ST, the image, upon Autodev, has essentially white stars/galaxy core, on a bright yellow to red background (the data is not of good quality!). With the DSS RAW colors adjustment settings of 5, 5, 5 (brightness, red scale, blue scale respectively), I get a much brighter image but there is a strong red bias to it. Both "Use per channel background calibration" and "Use RGB background calibration" are set to off.

Given that Star Tools ideally wants virgin data, I'm not at all sure how best to proceed, or indeed if I've even understood properly how everything works. Sorry for the long post, but please advise me on the best way to proceed with DSS in order to obtain files suitable for Star Tools to work on, and whether there are any settings in ST that will specifically require addressing for this sensor.

Ian

[admin]

Hi Ian,

Rest assured, the learning never stops for anyone in this hobby, there's always something new!
Today is the first post I ever encountered about Fuji's X-Trans sensor for example.

Here's my problem. My camera uses the Fuji X-Trans sensor, which as you probably know is not a strict Bayer array, and has rather fewer red/blue pixels in relation to green pixels than the conventional Bayer array. The latest version of DSS doesn't seem to process the RAF raw files correctly, even though the demosaicing engine, DCRaw, is supposed to support the camera. So I can either convert the RAFs to TIFFs using my raw converter (Capture 1), using a linear response curve, or convert to DNG using the Adobe DNG converter, and import those.

You could also try the actual standalone command line version of DCRAW. But anything that can do conversion with a linear response curve, no noise reduction and (preferably) no color balancing should in theory work just fine.

Using the DNG files, with the default RAW colors adjustment settings of 1, 1, 1 (brightness, red scale, blue scale respectively) I get a very dark image and DSS can only find a handful of stars, even with a low threshold setting. When the autosaved FITS image is imported into ST, the image, upon Autodev, has essentially white stars/galaxy core, on a bright yellow to red background (the data is not of good quality!). With the DSS RAW colors adjustment settings of 5, 5, 5 (brightness, red scale, blue scale respectively), I get a much brighter image but there is a strong red bias to it. Both "Use per channel background calibration" and "Use RGB background calibration" are set to off.

Any chance you could post some jpegs (just add .jpg when saving) of the two different outputs after AutoDev?

If the data is otherwise ok, it may indeed look pretty terrible, until you perform a Wipe to rid the image of bias, gradients and light pollution (use the Temporary AutoDev feature in Wipe to see what a sample result would look like). After that, you should be able to get detail coming through, though with unbalanced colors - the latter are ideally fixed using the Color module at the very end of your workflow, just before switching Tracking off.

Provided the data is correctly interpreted and stacked in DSS, nothing further is needed to be able to start using it in StarTools.

If you think you have ok data, but are unsure how to process it, I'm more than happy to have a look if you can share it through, for example, Dropbox, Google Drive or similar services.

Hope this helps in the meantime!

[Admiral]

Hi Ivo, and many thanks for your detailed reply.

As to using DCRAW from the command line, well, that's a wormhole I'd rather not go down Perhaps I'll stick to converting to DNG. I have no idea what processing goes on behind the scenes with the Adobe DNG converter, but I would guess (hope?) it's pretty minimal. Certainly, when the DNGs are imported into DSS they are shown as RGB, not Bayer.

Below I've posted a number of screen dumps which will show you what I've been getting. Let me say at the outset that the files are pretty poor, being my first foray into AP and the process of stacking. I've since got a proper astronomical 'scope but until my camera adapter arrives I'm unable to get any decent frames, not to mention the abysmal weather here in the UK! I'm also still running only the demo version of Star Tools whilst I establish my workflow, so I've had to screen-dump rather than save directly. I hope that is OK.

This set of images relate to using the default Brightness/Red scale/Blue scale RAW settings (i.e. 1, 1, 1). With these settings even with a 4% star threshold I only get 4 stars, and I know that there are a lot more. The DSS autosave applies these settings to the files.

This is the DSS result

DSS result1.jpg (177.78 KiB) Viewed 6880 times

And then in ST with Autodev

ST autodev.jpg (161.98 KiB) Viewed 6880 times

With cropping, wiping, and then autodev, I get this

ST crop wipe autodev.jpg (186.65 KiB) Viewed 6880 times

The wipe module doesn't seem to make a lot of difference, but then again, maybe it is just poor data. I am not able to get a decently dark background, at least not without lowering the gamma and losing a lot of the fainter detail in M31.

If I use (5,5,5) for the Brightness/Red scale/Blue scale RAW settings, I get a clearly red-biased image that results in ST producing a shockingly red image! (I haven't included these as I'd need to put them in another post, but can do if you wish). Clearly the colour balance is totally out, but at least I get 110 stars with a star threshold of 30%. So, I need greater brightness yet with the red and blue proportions changed to more closely reflect the Fuji X-Trans sensor (40% red, 40% blue, rest green).

I suppose really I am looking for some reassurance that converting to DNG before importing into DSS is a reasonable approach, together with making appropriate changes to the RAW setting parameters in DSS. At what point might one need to change the red, green and blue luminance contribution sliders in the develop module?

Ian

[admin]

On the face of it, that all looks quite normal.
Regarding Wipe, can you try upping the Dark Anomaly Filter setting, and perhaps binning the data prior to launching WIpe?
It seems to me Wipe is trying its best but may be thrown off by some spurious very dark pixels (Wipe tries to never clip your data).

If you want to upload DSS' output we can maybe have a look at the data?

[Admiral]

Thanks Ivo. I've now been able to generate a lot more subs and I think I'm beginning to tame the beast. Binning and upping the dark anomaly filter certainly help.

You'll be pleased to know I've now purchased the package!

Cheers
Ian