[Digikam-devel] Colormanagement

Peter Heckert lists at hphsite.de
Sun Apr 2 03:35:46 BST 2006


Hello,

have some problems with colormanagement.

I am trying to use linear RGB as a working space.
For this purpose I am using this (homemade) colorprofile as workingspace:
<http://home.arcor.de/peter.heckert/raw/RawSharpen/sRGBG1.icc>

This is a profile that uses sRGB primary colors, but instead of 
Gamma=2.2 it uses Gamma = 1.0.

My imaging software has full colormanagement and is fully 16-Bit 
capable. (Photoline32 for Windows, http://www.pl32.com)

In this application I have two choices: I can /assign/ a colorspace to 
an image or I can /convert/ an image into another colorspace.

In the first case,"assign",  the image data is not modified, only its 
interpretation is modified and so the visual look will change.
In the second case, "convert", the image data is converted into another 
colorspace, the data is modified, but the display stays unchanged, 
because also  the interpretation of image data changes.

Normally I do this for my RAW Images: I load the image with Gamma=1.0.
The image will look dark then.

Then I /assign/ sRGBG1.icc to the image. Then the image looks rather 
fine. (Some contrast improvements and else are necessary anyway)

The image is not converted in any way. The image processor will adjust 
image's *displaybuffer* accordingly, it doesnt modify the image itself.

That means, all image manipulations in follow will be done in linear RGB 
space. All filters and all blending modes work in linear RGB.

That is how I understand the word "workingspace".

Ok, here is another trick:
Sharpening works much better in linear RGB, because sharpening halos are 
reduced. Denoising works better, because local noise amplitude is 
independent from local brightness in linear RGB space.

Therefore I often convert my JPG's to 16 Bit and then I /convert/ to 
sRGBG1.icc before I continue with image manipulations.

(I convert to 16 bit, because linear RGB cannot be used with 8 bit 
resolution, there would be too much loss of information)

I tried to do the same trick in digikam.
In digicam I cannot convert colorspaces. I can choose another working 
space. However, when I choose another working space, then the image's 
visual onscreen appearence is changed and so far I understand it, this 
should not happen.

Ok, now I am unsure: Is this a bug or do I missunderstand 
colormanagement philosophy or is colormanagegement implementation still 
unfinished?

greetings,

Peter



More information about the Digikam-devel mailing list