Re: GaMapICC v0.5
Re: GaMapICC v0.5
- Subject: Re: GaMapICC v0.5
- From: Graeme Gill <email@hidden>
- Date: Thu, 28 Jan 2010 12:36:16 +1100
Marco Ugolini wrote:
Regarding my image, I was showing a difference in results using image-specific gamut
mapping, which in my opinion (a) makes the image overall too light and (b) loses some
detail in the indicated red area.
Right, and I'm indicating that (due to the nature of the image) this is little to do
with image specific gamut mapping, but rather to do with simply different gamut mapping.
It's gamut almost looks like it's already been clipped to the destination space (or
very near),
My inspection of the file in ColorThink Pro reveals that the red areas are up to 8
DeltaE 2000 outside the gamut of the destination profile. Eight DeltaE 2000 is not very
near.
In terms of gamut mapping issues, it is near in the sense that it amounts to
about a 20% compression - which is not very much. Difficult gamut mapping
situations are where the needed compression is 100% or 200% or more.
But I was commenting on the fact that the shape of the image gamut is rather
similar to the destination space - it basically looks like it has been clipped
to the destination space "light" surface, and lies on the source colorspace "dark"
surface, which is is a bit more saturated than, but has a similar shape
to the destination "dark" surface.
Am I to understand that image-specific gamut mapping actually works less well when
colors are better-differentiated to begin with? How does that help?
As I explained, it shows advantage when the image contains colors within
the destination gamut that would otherwise be unnecessarily compressed
if the source colorspace gamut is used to guide the gamut mapping.
Even by the standard of better differentiating colors that are near in hue, I don't
detect an improvement for those in the GaMap image.
It won't differentiate colors near in hue, in fact color differentiation
inevitably gets worse when compression is needed. But what it does do
is not unnecessarily apply compression when it's not needed. The example
image you chose doesn't have such a situation though - it is dominated
by a few major colors (white, black, brown, orange), all of which extend
up to the source gamut surface. So the overall gamut mapping will
be very much the same for image specific or source colorspace specific
gamut mapping, and won't be that different to simply clipping it.
Naturally simply clipping it gives a more contrasty result.
I've done a trial with typical settings I'd use (after fixing a glitch in the image
gamut code that Klaus drew to my attention), and also created a soft proof back in
AdobeRGB space here <http://www.argyllcms.com/Axe.zip>, and while there are some
differences to your result, it all looks rather similar.
I disagree. The proofed AdobeRGB file looks different enough from my source image to be
noticeable. If that is what you meant to say.
Well, it's all a matter of degree. There are differences of course, but they
are minor compared to the differences seen in more difficult gamut mapping
situations.
I also fail to see what proofing the image back to AdobeRGB aims to prove.
It properly takes into account the viewing condition issues. The gamut
mapping I've done allows for the display and print viewing conditions.
If you're doing a soft proof comparison and don't undo this adjustment,
you're not making a valid comparison, and the image might look (say)
too light, where in fact it might better match in real life.
In any case, the CMYK files are the only ones I really care about, because they
represent the results that would go on press, i.e., the results that the client would
have to like enough to pay for. A comparison between the CMYK conversions made with
GaMap and with the "regular" Photoshop conversion (when compared to the "look and feel"
of the original) still shows a visual advantage for the "regular" Photoshop conversion
(better definition, better contrast, better detail, even better saturation).
It's not the conclusion I come too. Comparing the softproof of the
Photoshop conversion and softproof the gamut mapping I created I would
say that in comparison to the original image, the ProfileMaker/Photoshop
conversion has lost saturation and lightness of the orange "blood", and
overly darkened the woodgrain. But the use or not and type of
BPC used in the softproof has a big influence on this assessment.
[The original image looks a bit like it's been mistreated to begin
with. It's kind of dull, and only begins to look natural if you set
the RGB levels to about 220 and correct the white point a bit.]
(I updated <http://www.argyllcms.com/Axe.zip with
a slightly higher quality conversion too.)
cheers,
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden