Calibration data not applied in Hasselblad 3FR format

[Beep6581]

Originally reported on Google Code with ID 2611

RT 4.2.43

I just got an H4D-50 and noticed really bad results on the 3FR conversion. Haven't
really noticed this before, but it seems my 'blad has some pretty strong calibration
data, and dcraw does not apply any. If the camera does not have strong calibration
curves you may not notice any problem.

This means that there's reverse engineering to make...

Reported by torger@ludd.ltu.se on 2014-12-18 16:52:26

[Beep6581]

Congrats with the new toy:)

Reported by michaelezra000 on 2014-12-18 17:21:30

[Beep6581]

Thanks! We'll see if I keep it :-), I haven't sold my Leaf Aptus 75 yet. I probably
will keep the Hassy and sell the Leaf rather than the other way around, but I really
like it to work well in RawTherapee (like the Aptus does), and it hasn't started well
:-).

I'm also really impressed by Hasselblad's Phocus raw conversion, they've really managed
to extract maximum performance out of the CCD, which in DxOmark score is not that good
compared to modern CMOS.

Hopefully we'll get RT there too...

The camera writes 3FR to the CF-card, which gets converted to FFF when opened in Phocus.
I did not know what the difference was before but now I know, FFF is the same as 3FR
with the difference that the calibration data has been applied. Dcraw does not apply
any calibration data so what we support now is FFF and not really 3FR. My workflow
would be a lot smoother with 3FR support directly so I would not need to pass through
Phocus. So I guess I'll get into reverse engineering again...

Even with FFF the result is not perfect in RT though, there's something with the green
eqiliberation feature that needs to be refined (separate issue filed for that).

Phocus has some impressive noise supression of the shadows, better than what I've seen
in any other software (clean but not really looking like being noise-reduced), but
RT's new powerful noise reduction tools may be able to produce the same result, I just
need to learn them first :-).

Reported by torger@ludd.ltu.se on 2014-12-19 08:12:16

[Beep6581]

I have now comitted a patch which is a sort of "work in progress", but as it's very
isolated I think it was safe enough to commit, and I really need this for myself :-)

I've worked hard for this. Reverse-engineering can be really frustrating and the 3FR
calibration data is no exception... but now it works for my H4D-50 and a bunch of other
Hasselblad cameras. There's more work to do to get 100% complete support for all tags
and all camera models, but this is a really good start.

At some point I will send a patch upstream to dcraw-Dave, but I'll let it live in RT
and be tested by me for a while first.

Reported by torger@ludd.ltu.se on 2014-12-29 19:39:24

• Status changed: Started