Gamma differential

[Desmis]

I want to push a new branch of course with changes to the master branch. I'll call "gammadiff".

"gammadif" will make changes to almost the whole or part of the RT process.

To avoid doing stupid things ... (I used to do with Git) I propose:

Currently my new code is in the master branch.

Do the following approach is correct?

On my computer (master branch) 1) create a new branch \ git branch gammadiff

2) add the files * git add rtgui \ gamma.cc * git add rtgui \ gamma.h

3) commit changes \ git commit -m "init gamma" -a

4) updating the remote server \ git push

[heckflosse]

Jacques, after * git branch gammadiff you have to * git checkout gammadiff to switch to the new branch

[heckflosse]

and to update remote * git push --set-upstream origin gammadiff instead of * git push

[Desmis]

Ingo

Thank you !

[Desmis]

Ingo I do exactly..what you write...and :

• error message : src refspec gammadif does not match any
• fail to push some refs to 'git@github.com:Beep6581/Rawtherapee.git'

[Beep6581]

J, "gammadif" or "gammadiff" (two "f")?

[Desmis]

thank you, but I am very head in the air...

Now it works :)

[Desmis]

I made a topic in "Rawpedia" in french..."Gamma differential"

Also I changed the TIFF / JPEG output. You can now make an "output" without "ICM" with the RGB data

:)

[iliasg]

Jacques, in this tiff/jpeg output without icm, which is the RGB color space of the exported data ?. Same with the working space I guess .. correct ? Is this a solution for less posterization by overcoming LCMS inaccuracies ?. Like in https://github.com/Beep6581/RawTherapee/issues/2735 .. there I think the posterization occurs due to reduced density of 8bit linear data (sRGB at the darks) which is followed by RGB to Y'CbCr conversion (again 8bit ..)

[Desmis]

I don't realy understand what you mean..but an example is good :

• open an image an choose "Working profile" = Prophoto
• then : 1) choose "Output profile" = Prophoto ==> TIFF and "RGB instead ICM " unchecked
• 2) choose "Output Profile" = RT_Large_gBT709 ==> TIFF and "RGB instead ICM " unchecked 1) and 2) seems identical : Lab values are practically identical, but RGB values are different - look histogram : ICM does his work
• 3) as 2) but with "RGB instead ICM " checked==>TIFF. Image is different, RGB values are same as 2) but Lab values are different - ICM is out :)

[Desmis]

I have update "manually" branch "gammadif" with last commit "retinex" (Ingo) to avoid difficulties..I hope no error.. :)

[iliasg]

Jacques, EDIT "I see no "RGB_without_ICC" option in the GUI nor in preferences" is invalid .. temporary-blindness ??

• Looks like the last commit for retinex has not passed in gammadif branch .. the highlight restoration (this for both color_propagation and blend methods) gets disabled when I change retinex settings and I have to disable and re-enable it to make it active again

[Desmis]

I have update this branch with all (I hope) last chnages Retinex (Color propagation, events...)

Ilias :

• the 2 new are 1) for "gamma-differential" : "Exposure"==> second heading 2) for "RGB instead of ICM" in output : "Color"==>Color Management, last line :)

[iliasg]

Hi Jacques, I found the the new GUI elements .. the blindness was temporary :) .. I will test the new commit later. May I ask what is the objective of "RGB instead of ICM" ?

[Desmis]

Ilias Same objective as in Prophoto with proof setup, RGB internet, printing...

Here no soft-proofing, but "support" for:

• software that does not support color management (Web..),
• printing in RGB or CMJN (in english CMYK ?)
• ...

[Desmis]

I push 2 fixed for 2 bugs 1) GUI 2) process when change Standrad gamma absolute to relative

I hope no more error :)

[Beep6581]

Testing this branch I noticed two bugs where the preview image tiles weren't all updated correctly or sometimes had artifacts: 1- Only the top part of the image is drawn and its red, the rest is black http://i.imgur.com/r3MMLxi.png http://i.imgur.com/g76D6Ah.png 2- I can't reproduce now for a screenshot, but the image looked generally correct except for a few square areas which had correct fine but the "background" of this detail was wrong, had a color that didn't match the rest of the preview. To help you imagine the effect, it looked as if you used the Wavelet tool but some parts of the image didn't use the residual, so the fine detail was correct but the general color was wrong.

Google Translate does a bad job with http://rawpedia.rawtherapee.com/Gamma_-_Differential/fr I don't understand the purpose of this tool, what it can accomplish or how to use it.

[Desmis]

I push a change for bug "NOICM"!! I add a new method : Standard absolute ICC g=1 + end (see Rawpedia in french) I update Rawedia..for the purpose of this tool (it is similar to gamma Retinex, but with ICC instead of gamma only) :)

[iliasg]

Jacques, I hope I will manage to test in the weekend. What setting would you suggest if, for example, we want to decrease the action of deconvolution sharpening on highlights ?

[Desmis]

Hello Ilias Probably, use "RGB + L_a_b* differential ICC" and choose a gamma < 1

:)