Bug 145442

*** Bug 148538 has been marked as a duplicate of this bug. ***

Just a hint: You can manually do 16-bit contrast masking in Linux using Krita. It's really amazing what it can do to your prints.

Another great info page about the subject:

http://www.normankoren.com/PWP_contrast_masking.html

*** Bug 193336 has been marked as a duplicate of this bug. ***

A new "Local contrast" Digikam plugin has been introduced by Julien Pontabry based on code by Nasca Octavian (http://zynaddsubfx.sourceforge.net/other/tonemapping/). This plugins appears in svn version of 1.0.0beta4. Please give us your opinion about the gui and the result you can get.

Warning: this plugin is not yet finished, it does not work with 16bits pictures and progress bar does not work.

Here are some tests results and screenshots:

http://www.flickr.com/photos/julien_narboux/3802975521/
http://www.flickr.com/photos/digikam/3771531595/in/pool-digikam-labs
http://www.flickr.com/photos/mikmach/3782299455/

Julien Narboux

Wow, good to hear that this has gone into DigiKam. It looks like it is exactly the same thing as I'm doing manually using ImageMagick right now. Having this work in 16-bit color is vital for this process, if you want high quality output. Having more than one stage is an interesting idea, I only apply the procedure once.

Unfortunately my production machine runs KDE 3.5, so I cannot easily try this plugin. I'll test as soon as I dare to jump to KDE4. GUI looks nice, good work!

To Julien Pontabry, 

Do you have already started to port Local Contrast plugin to 16 bits color depth ?

Gilles Caulier

To Gilles Caulier,

Not yet. But I start working on it. It should be quick.

Julien

Created attachment 37382 [details]
Local contrast patch to support 16 bits color depth...

Julien,

See this patch against svn trunk to start 16 bits color depth support.

- ToneMappingFloat is used now, for 8 and 16 bits.
- 8 bits work.
- 16 bits give a black hole image. I don't know why.
- Fast mode is disabled. I think this option is not yet available for 16 bits. In fact ToneMappingInt do not have process_rgb_image(). this is wshy float method is used instead.
- in original code d->toneMapping container is not deleted from memory and end.

Gilles Caulier

Julien,

Another problem in code is missing alpha channel support. It's removed completely.

The solution is simple : restore alpha channel from original as well...

Gilles

SVN commit 1031632 by jnarboux:

Modified version of the patch by Gilles Caullier for 16 bits support for local contrast plugin.

CCBUGS: 145442



 M  +30 -0     ToneMappingFloat.cpp  
 M  +1 -0      ToneMappingFloat.h  
 M  +46 -14    localcontrast.cpp  
 M  +1 -0      localcontrasttool.cpp  


WebSVN link: http://websvn.kde.org/?view=rev&revision=1031632

SVN commit 1031836 by cgilles:

implement cancel handlin in LDR tone mapping algorithm


 M  +144 -135  ToneMappingFloat.cpp
 M  +34 -33    ToneMappingFloat.h
 M  +8 -8      ToneMappingInt.h
 M  +14 -0     ToneMappingParameters.cpp
 M  +17 -9     ToneMappingParameters.h
 M  +3 -1      localcontrast.cpp


http://websvn.kde.org/?view=rev&revision=1031836

SVN commit 1031957 by cgilles:

implement tone mapping progress information.
CCBUGS: 145442


 M  +85 -74    ToneMappingFloat.cpp  
 M  +5 -0      ToneMappingParameters.cpp  
 M  +1 -0      ToneMappingParameters.h  
 M  +11 -0     localcontrast.cpp  


WebSVN link: http://websvn.kde.org/?view=rev&revision=1031957

Julien,

Plugin now support 16 bits color depth images, dispatch progress info, and is cancelable...

Still "Fast Mode" option which is not implemented...

If i have understand old code Fast mode is to compute all tonemapping using integer instead float values. Right ?

If yes, what's ToneMappingParameters::info_fast_mode variable ? It's never used in code...

ToneMappingFloat code is fine now. If we want to do the same in ToneMappingInt to implement Fast Mode, just take a look in ToneMappingFloat...

Gilles

Great Gilles !

If you select the fast mode, the ToneMappingInt class is used. If not, the ToneMappingFloat class is used. Both herit from ToneMappingBase. So in the control class, we switch from one to another, depending on the fast mode's value (enable or disable).

Julien

To Julien Pontabry,

Yes, i have seen the difference between ToneMappingInt and ToneMappingFloat to compute faster local contrast effects

My patches only use now Float, especially to support 16 bits color depth. Fast Mode is now disabled. ToneMAppingInt do not have rgb16bits method and a lots of code need to be written. Julien Narboux, if you is motivated, let's go (:=)))...

My test with float give good speed results, especially with Raw images in 16 bits. I think it can be enough.

I will cache this option from gui for the moment, and close this file. If you think that we need to provide this, feel free to show it again...

Gilles

Gilles,

To be honest, I think the code contains too many duplicated functions,  it contains another implementation of unsharp mask, saturation and gaussian blur...  
It is working now and give good results and I thank Julien Pontabry for that !
but I think we should not spend time to optimize but rather try to reuse the functions from digiKam and build a nicer framework for image processing in digikam. I will try to do something if I have some free time...

Julien

SVN commit 1032086 by cgilles:

hide Fast option from gui.
BUGS: 145442


 M  +1 -2      localcontrasttool.cpp  


WebSVN link: http://websvn.kde.org/?view=rev&revision=1032086

I agree with Julien. If I had more time, that's what I did...

Julien