Handling Fujifilm X-Trans Auto Focus Pixels

Thanks. I just need dark frames of each model only if problem occurs.
In fact I need to measure square coordinates :).

I have shared a link to my source files via a private message.

I’ll start doing the shots. Does ISO matter here?

I took some frames for X-T2 and X-T3 but I cannot upload them to filebin, I get errors all the time. Is there an alternative available?

I finally managed to upload the files here: Filebin | kuslcmtjmg6x2m0o

Please note that as found here: https://www.dpreview.com/forums/thread/4416122#forum-post-62966528 the Fuji X-Trans cameras apply some RAW filtering for exposures longer than 5 seconds. It might be that for these cases a different processing is needed …

The actual locations of the auto focus pixels can be really hard to find. I created this 8-bit tiff that shows their locations well. There’s also noise in this image, but the auto focus pixel pattern is easily visible.

pixel_locations.tif (1.2 MB)

I created this by taking several dark frames and first doing a minimum stack in Siril. Then I did some severe stretching and clipping until the pixel pattern popped into view. I clipped the dark pixels by about 90% or more until most of the low background noise was gone. I also pulled the middle triangle really close to the left triangle in the Histogram Transformation tool. It helps to use the zoom feature in the tool. It gives more granularity to the controls. Anyway, this might be helpful for anybody trying to find the auto focus pixels on their sensor.

It should be possible to get an even better map by doing this on only the green channel, but this worked well enough for me to find the locations.

Thanks to @setnes we have a nice code that works.
So now we can remove this ugly pattern.

So, we need any available informations on other sensors that the Fujifilm_X-T20.

1 Like

I can provide files from X-T2 and X-T3.
What kind of information you need?

I created Fuji X-T3 support - White Frames · Issue #5839 · Beep6581/RawTherapee · GitHub and Fuji X-T2 Support - White Frames · Issue #5824 · Beep6581/RawTherapee · GitHub for the white levels in rawtherapee, but I guess that for Siril you need more the dark frames.

Also, in some threads on dpreview where it was discussed how suitable are the Fuji cameras for astrophotography, the analyze showed that there is some sort of rawfiltering that the cameras make for exposures longer than 6 seconds. I can link it here if you find it helpful.

Hello @Daniel_Catalina .

Indeed I need dark frames. For example taken at an exposure of 60sec with lens cap. And I believe, not sure though, it is better to do it at high ISO.

Many thanks.

In fact, it would be better to shoot 3-4 files, in same conditions, in order I stack them.

I have a few darks available, that @heckflosse used for hot pixels detection for xtrans sensors: https://drive.google.com/drive/folders/1WpMNWeJ5WFIOR_9_3jIt_dqEmI3rB2vz?usp=sharing

Feel free to use them if you think they can help (made with an X-T2)

1 Like

I have some samples from an X-T3 and an X-T4 that I can look at. I can try to look at X-T2 samples from @sguyader as well. It should be the same as the X-T20, but we should verify.

I assume that long exposure noise reduction must be turned off in the camera settings for these samples. I don’t know if ISO will make any difference at all for these cameras (invariant).

I cannot promise I will be able to look at this as soon as tonight. The forecast shows a clear sky. :slight_smile:

1 Like

My tests show it is the same indeed

@lock042
I have a bunch of shots, made with X-T1, X-T3, and X-Pro2.
part …a.zip = all ISO, 1/2 sec, f/5.6
part …b.zip = all ISO, 5 sec, Long exp noise ON
part …c.zip = lenscap on, base ISO, 1/2, 1, 5, 10, 20 sec

Perhaps part c (one set per camera) might be of use to you, even if shot at base ISO?

Have fun!
Claes in Lund, Sweden

2 Likes

I’ve put here Filebin | hvq0by5xmsn62dj3 the X-T3 samples:

  • 6 x 60s, ISO3200 uncompressed raf
  • 6 x 60s, ISO 3200 lossless compressed raf

Let me know if you need the same for X-T2 or the ones already available are enough.

1 Like

Yes I would need part c :slight_smile:
Thank you so much.

Thank you so much. For the X-T2 it is ok, thanks !!

1 Like

@lock042 Three zips pm’d

Have fun!
Claes in Lund, Sweden

Thank you all.

It is a pain in the … to get the accurate square coordinates

1 Like

One question here: with Siril 0.99.4 I get an error for Fuji X-T3:

14:27:57: Preprocessing…
14:27:57: Fix X-Trans: Unknown camera Fujifilm X-T3, trying to read information from preferences.
14:27:57: No information available in preferences.

Is this expected? Are more files than the ones provided above needed? Or is the feature not useful with the X-T3?

Two possible reasons for this message:

  • The camera was tested and no ugly pattern was detected. So the fix is not needed.
  • The camera was untested because we do not have samples to test it.

In your case, I’ve tested some images of T3 without seeing any square pattern.

1 Like

Thanks for the reply!