Re: LAB as working/interchange space
Re: LAB as working/interchange space
- Subject: Re: LAB as working/interchange space
- From: olaf druemmer <email@hidden>
- Date: Thu, 24 May 2001 12:19:02 +0200
In an ICC based RGB space like AdobeRGB you more or less use the full
range of 8bit values to encode all visible colors. In Lab that same range
of colors will be encoded within a smaller range of 8bit values (because
for certain combinations of 8bit values in Lab no visible color exists,
or said differently: in Lab you can encode colors no human being can
see). The problem that arises is that you lose precision.
You often won't notice this if you go from e.g. AdobeRGB directly to
color space for a print color space (with potentially a smaller gamut) or
from AdobeRGB via Lab to that same space. At least you do not notice it
in terms of color. You may notice it though in regards of how well
details are maintained or e.g. blends/vignettes stay smooth or not.
So using (8bit) Lab the color as such is still OK but you have lost
precision. You may be able to avoid this using 16bit Lab but that is not
yet very well supported in standard tools and also implies a performance
penalty (larger file sizes, longer times to process, etc.) .
Olaf Druemmer
>
--- Chris Murphy wrote 5/24/01 8:49 AM: ---
>
Chris Cox writes:
>
>
>16 million colors in 8 bit/channel sRGB will transform to only 2.47
>
>million colors in 8 bit/channel LAB (without dithering).
>
>
and
>
>
>But if you look into
>
>the details at all, you quickly find that LAB is not a great
>
>representation for your image data. It's a nice mathematical
>
>construct for interchanging data -- but whenever you use LAB you have
>
>to keep around extra bits of precision to maintain image quality.
>
>
Why is it that I can take a wide variety of RGB images defined with
>
Adobe RGB, convert them to Lab, then take the Lab set of images and the
>
RGB set of images - and convert them properly from their respective
>
source to a destination using a suitable output device profile - and end
>
up with visually the same thing? From inkjets to printing presses, I
>
don't see the decrease in image quality.
>
>
I have no doubt what you're saying is true, I'm just wondering to what
>
degree this actually affects the reproducibility of images? In what
>
respects should I be seeing image degradation just by converting a good
>
RGB image into 8-bit/channel Lab?
>
>
>
Chris Murphy
>
Color Remedies (tm)
>
Boulder, CO
>
303-415-9932
>
_______________________________________________
>
colorsync-users mailing list
>
email@hidden
>
http://www.lists.apple.com/mailman/listinfo/colorsync-users
>
-----------------------------------------------------------------------
Olaf Druemmer
-----------------------------------------------------------------------
callas software gmbh Tel +49.30.44390310 Leonardo +49.30.44342390
Schoenhauser Allee 6-7 Fax +49.30.4416402 connect.pdf +49.30.44342391
D 10119 Berlin email@hidden
http://www.callas.de
-----------------------------------------------------------------------
Now hiring - open positions for C++ developers on Mac, Windows & Linux.
-----------------------------------------------------------------------