Dual ISO - massive dynamic range improvement (dual_iso.mo)

Started by a1ex, July 16, 2013, 06:33:50 PM

Previous topic - Next topic

0 Members and 5 Guests are viewing this topic.


akumiszcza

After taking quite a big series of dual iso photos on my Canon 50D I've noticed that each and every one DNG created has white balance values off by about -500K, +30 tint (in ACR) compared to non-dual-iso photos taken in the same circumstances. As this value is quite constant, is it cr2hdr error or some ACR bug in reading "As shot" value from DNG?

Additionally, here's one conversion that gives strange results:
CR2: https://mega.co.nz/#!Yc9VFC6T!cidlsw8J2YOGGTVnsmTzlVjuPK8v8MtM4oAZTK_Sgrs
DNG: https://mega.co.nz/#!oQs0kYob!NLGjazsBAfPAuhBT4KuUN0vwUw8kbPlp2kIK3oqgS8Q

a1ex


akumiszcza

Quote from: a1ex on October 06, 2013, 11:48:31 PM
Is exiftool running?

I think so – cr2hdr gives no error while running, exiftool.exe is in the same dir and runs fine when called. It's in the path too.

Here's an example of what I'm talking about:
https://mega.co.nz/#!gYcwSYLJ!cQuj1zx4Q2k1IphjgMZIOBHjOoJpwEC4HUHsTPDM2jI
In the zip you get non-dual-iso CR2, dual-iso CR2 and generated DNG. I need to change WB of the DNG by about +500 K, -30 tint to have it similar to non-dual-iso CR2.

engardeknave

I actually tested this today.





The .dng seemed to be about right judging from both the photo and histogram at +26 tint (+34 change) and no change in temperature.

budafilms

(I see greener .dng)

Do you know if DNG extracted from raw2dng and Son of the batch are the same file?

Marsu42

Quote from: akumiszcza on October 06, 2013, 11:32:31 PM
As this value is quite constant, is it cr2hdr error or some ACR bug in reading "As shot" value from DNG?

I wouldn't say the tint/wb change is constant across the board, but maybe consistent enough to get some "standard" correction for different iso spacings and bv values so less needs to be done in post.

a1ex

cr2hdr does not touch WB coefficients and it's not meant to do so.

akumiszcza

Quote from: a1ex on October 07, 2013, 11:21:22 AM
cr2hdr does not touch WB coefficients and it's not meant to do so.

How come then that two photos — one normal, one dual iso — taken seconds apart in the same conditions have completely different WB? I don't have a single DNG file with proper WB set. Maybe it's ACR issue? See the example files: https://mega.co.nz/#!gYcwSYLJ!cQuj1zx4Q2k1IphjgMZIOBHjOoJpwEC4HUHsTPDM2jI

a1ex

As I said earlier, the issue is in EXIF. So, just find the exiftool command that copies WB properly from the CR2 and I'll integrate it.


Onoxyne

Hello, thanks for your job a1ex !
I've got some problem with my cr2hdr in Ubuntu :

Quote./cr2hdr _MG_5881.CR2

Input file     : _MG_5881.CR2
Canon EOS 60D detected
Full size      : 1336 x 3516
Active area    : 1336 x 3516
Unknown option "-E".
Error: dcraw output is not a valid PGM file

I've download the last Hudson repository this day.
Canon 60D.

a1ex


Onoxyne

But I do the compilation in this day and in the source code, the dcraw_bridge date is 2013_09_08 ! What's wrong ?
How can I take the last build ? Thank you much.

a1ex


Audionut

Quote from: a1ex on October 07, 2013, 03:59:29 PM
A little big usability breakthrough:

Looks to be working good.
Extracted JPG previews.

Highlights


Lowlights


Sharplights


Right click | view image for full res.

fotojohni



Not possible without dual iso.  This is a truly incredible development.  Now if only it didn't have to go line by line and instead could be a dual bayer pattern with 6 instead of 3. 

Also, any chance for an adaptive algorithm that measures scene contrast in live view then auto selects?   

Just putting out random ideas I bet you are already cooking up something even better.  can't wait.

Audionut


Marsu42

Bug report: I just had a look at the exifdata cr2hdr and acr generate - and ml screws up the lens information, this is part of a diff between ml (wrong) and acr (correct).


-Circle Of Confusion             : 0.008 mm
-Field Of View                   : 30.0 deg
-Focal Length                    : 17.0 mm (35 mm equivalent: 67.1 mm)
-Hyperfocal Distance             : 2.37 m
-Lens                            : 17.0 - 40.0 mm (35 mm equivalent: 67.1 - 158.0 mm)
+Circle Of Confusion             : 0.030 mm
+Field Of View                   : 93.3 deg
+Focal Length                    : 17.0 mm (35 mm equivalent: 17.0 mm)
+Hyperfocal Distance             : 0.60 m
+Lens                            : 17.0 - 40.0 mm (35 mm equivalent: 17.0 - 40.0 mm)


In the cr2 the data is also correct, but cr2hdr adds a wrong crop factor:


-Scale Factor To 35 mm Equivalent: 1.0
+Scale Factor To 35 mm Equivalent: 3.9


Bug ticket here: https://bitbucket.org/hudson/magic-lantern/issue/1697/cr2hdr-messes-up-the-lens-meta-information

a1ex

Quote from: Audionut on October 07, 2013, 05:36:48 PM
Extracted JPG previews.

What about enabling HTP? It should make the preview brighter by one stop, right?

fotojohni

Quote from: Audionut on October 07, 2013, 06:08:39 PM
http://www.magiclantern.fm/forum/index.php?topic=8322.0

Okay this pretty much solves that.  I wonder how you get the metering data so that there is no need to go into live view for every shot.

Audionut

Quote from: a1ex on October 07, 2013, 06:22:48 PM
What about enabling HTP? It should make the preview brighter by one stop, right?

Yes it does. 

Audionut

Quote from: fotojohni on October 07, 2013, 06:28:46 PM
Okay this pretty much solves that.  I wonder how you get the metering data so that there is no need to go into live view for every shot.

Raw data is available in image quick review.

Onoxyne

Quote from: a1ex on October 07, 2013, 04:49:47 PM
Type "dcraw" into google.

OK, sorry for all the idiots questions !  ::)
cr2hdr is working ! That normal but my way to succed was very long.

Marsu42

Quote from: a1ex on October 07, 2013, 04:49:47 PM
Type "dcraw" into google.

or, preferably, into some search engine that doesn't track your every movement like ixquick or duckduckgo :-o