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 1 Guest are viewing this topic.


Quote from: a1ex on October 19, 2013, 08:13:43 AM
That's where I'm applying chroma smoothing. To desaturate, I think I need to know the white balance first.

Um, sorry for being daft again, but the wb value is supposed to be in the raw file, isn't it? If the problem is calculating awb, the camera could shoot dual_iso with a fixed/known wb as it doesn't matter in raw and you could try fixing the color artifacts - I also noticed them in some of my shots on tree leaves in front of the sun.

Quote from: a1ex on October 18, 2013, 09:52:40 PM
Top: mean23
Bottom: AMaZE interpolation
Left: no chroma smoothing
Right: 5x5 chroma smoothing

Before chroma smoothing, AMaZE has more aliasing in the shirt pattern, but afterwards there are less color artifacts left - that's why I prefer bottom left (AMaZE + smoothing).


The WB value is what you set in post-processing, after running cr2hdr. For desaturation, it does matter, because if you use some other WB, you will get hue shift.


Quote from: a1ex on October 19, 2013, 09:09:38 AM
The WB value is what you set in post-processing, after running cr2hdr.

But.. but ... the camera has a wb setting (either awb, situation (cloudy, ...) or explicit kelvin) that is written to the cr2, in 90% of the cases I don't change that in postprocessing. Or is the in-camera wb different from the postprocessing wb (tint, ...)? Thanks a lot for explaining  anyone, I really feel a bit dumb I cannot tell the difference :-o


Yes, but I don't want to rely on that, because I change WB in post pretty often.


Hi, a1ex,

I found today this black leaking into the picture:

here is original cr2:

I processed it with the last cr2hdr from 15.10.
Could this be cured?


I'm not sure what to look for, can you be a little more specific?


Quote from: a1ex on October 15, 2013, 06:36:36 PM
Would be nice if you can find a before/after example where the difference is clearly visible.
This is all I have for banding.  Badly underexposed image there is improvement in the banding  but not removed.  I see also resolution improved.

http://www.pululahuahostal.com  |  EF 300 f/4, EF 100-400 L, EF 180 L, EF-S 10-22, Samyang 14mm, Sigma 28mm EX DG, Sigma 8mm 1:3.5 EX DG, EF 50mm 1:1.8 II, EF 1.4X II, Kenko C-AF 2X



There is black halo around the person, black color from his clothes is spreading/leaking
into surrounding area.

boosted effect: https://www.dropbox.com/s/d4q4n08ukggjoy8/2013-10-19_3286%20more%20visible.jpg 

LR: exposure 0, highl. -100, shadows -100, clarity +100

Can you see it now?


Ah, that one is from your postprocessing software, not from cr2hdr.

Here's how I would render it (with my HDR script, highlight_level = 10000, midtone_level = 10000, shadow_level = 5000, overall_bias = 1, ufraw_options = "--temperature=5500 --green=1 --saturation=1.3 ")

(click for full image)


Thanks, a1ex,

I do not know what to think, I have never had such a result from Lightroom,
so I asked you about dualISO.
But I see your result, it is really OK...
Hm. ... Thanks anyway, I will look further into it. Perhaps. :-)


Another interpolation method (added at the end). It's from the same data set (a half-resolution image).

Another crop (this one is from the final output). DNG from Africashot. First image is from the current version (a7492e4, 2013-10-15).

If I'm reinventing the wheel with all these interpolation methods, please point me in the right direction.


are those really the same pictures.? the last one looks better in every aspect. great.



whow.. great alex.

can i try some of my pics? is there a cr2hdr version?


Here, just uploaded. I didn't try it on my test collection yet, so you may have surprises.


First side effect seems a good one: it removes some of the hot pixels ("star eater").



The last ones look amazingly good. Wonder how this will apply to dual-iso movies


Hi, a1ex,

I tried your new cr2hdr-amaze-edge.exe - it is really much better. Real improvement. Thank you.
However, I have a cr2 which crashes that new exe. (at least for me)
Previous version of cr2hdr did process it without problem.

Here is that cr2: https://www.dropbox.com/s/egdau7pki8620t1/2013-10-17_3125.CR2


Quote from: sletts02 on October 18, 2013, 12:25:58 AM
I've been out of the Dual ISO game for a while now and was hoping somebody could enlighten me on some topics.

Video Processing - is this the process?
1) RAW file on to PC
3) DNG's on to CR2HDR?

Does "RAWanizer" automatically do this process? It produced some non-banded DNGs that have less aliasing and comparable range so I am unsure. When the files were put straight in to RAW2DNG they came out alternated banding.

Are these the expected results?
Before cr2hdr: https://www.dropbox.com/s/mhgqe2eqkva67kq/beforecr2hdr.dng
After cr2hdr: https://www.dropbox.com/s/j22onsqq70yoq1e/aftercr2hdr.DNG

Should there be more shadow recovery in the processed version? Or am I doing something wrong?

Here is the resulting processed JPG, am I meant to raise the shadows and blacks 100%?



I don't quite understand the question, but you are free to color grade it as you wish.


Quote from: a1ex on October 20, 2013, 10:46:41 PM
Here, just uploaded. I didn't try it on my test collection yet, so you may have surprises.

The surprise over there is that this cr2hdr-amaze-edge.exe crashes in the "Interpolation" step on nearly every cr2 I test it on - from the 6d, that is. The current/old cr2hdr trunk version works fine. I really hope it's not again only me this time :-p

Here's a random cr2 to crash it: https://bitbucket.org/Marsu42/ml-aiso/downloads/DUAL8021.CR2

Here's one of the few cr2 that happen to work, maybe it helps you to find the bug: https://bitbucket.org/Marsu42/ml-aiso/downloads/DUAL7537.CR2


Yea, some are crashing it. At edge detection interpolation. The amze part is fine.

Pretty please can it come to raw2dng when the cause of the crash is found. What it processed looks very clean.


I'm even thinking to try this algorithm for upsampling 720p videos (since they are full-res horizontally, but have skipped lines, so it should be pretty similar). Though, here I'm exploiting the pattern of 2 lines sampled and 2 missing (so I have pretty good confirmation when edges are diagonal), but on 720p videos it's something like one line sampled and 5 missing.


There are a couple more good algos in raw therapee.. I think IGV, don't have a dng on here to test with.


Cool. I need to upgrade my ubuntu box to try the latest RT.

I guess any of them should work, since I copied AMaZE code from RT and did some minor changes (mostly C++ to C conversion).