Hi folks,

I know about the general Qimage approach towards the handling of 16-b/ch file input. And Indeed, it is only in rare situations that a difference between 8-b/ch and 16-b/ch input, will lead to visible differences in output. Added to that is the difficulty to actually print through a 16-b/ch pipeline, due to operating system / printer driver limitations.

My question is more about the order of conversions, and probably only Mike is able to answer that but maybe I failed to find the post (I did search) where it was explained before.

Assuming we have an input file, coming from a competent (14-bit ADC or better) camera, which is Raw-converted and additionally worked on in a competent Photo-editing application, in a working-space that's a bit on the large size, say ProPhoto RGB, because a lot of image manipulation might push saturation beyond the limits of Adobe RGB. We now have a 16-bit/channel file with relatively large quantization steps to fill the ProPhoto RGB gamut for some colors, and largely unused space for other colors.

Now, what happens when we open such a file in Qimage?
1. Is it already truncated/rounded to 8-bit/channel upon opening, or does the original precision remain available (or is retrieved again when needed) until after later Color Space conversions?
2. When re-sampling (usually deferred until actual printing, unless the editor is used to interpolate first), is that based on the original 16-bit (possibly truncated/rounded) precision, but the result is in 8-b/ch precision?
3. The conversion to the output profile, is that done prior to up-sampling, or after up-sampling.

Number 1 may cause issues when we interpolate notably smooth gradient transitions, although the interpolation itself may be smooth. Number 2 would probably require re-reading the original data, when the moment arrives that it's needed, but the result could again be held as 8-bit/channel data. Number 3 is potentially the biggest issue, especially with large gamut original image data. Conversion from one colorspace to another can be considered a major image editing operation, and we know that large gamut colorspaces like ProPhoto RGB is best used in 16-bit/channel mode if major editing is yet to take place. That is because the huge quantization steps in a large gamut space, may not map nicely to a smaller quantization step size, especially when only 8-bits/channel are available. That's why perhaps dithering is required to hide the errors (does Qimage/LCMS apply dithering?).

So, I'm basically looking for the moment of 16-b/ch conversion to 8-b/ch, and the moment that the profile conversion is done, before or after 16-8-bit conversion. Any help and profound insight into the inner workings of Qimage would be appreciated.

Cheers,
Bart (A long time user of Qimage, since 2001)

Hi Mike,

Thanks for clarifying that. It's pretty much what I expected, just wanted some confirmation.


All understood, and it makes sense. Unfortunately, my workflow can get very complex (incl. Focus stacking, Exposure fusion, and Panostitching those).


Clear. Unfortunately, when multiple output profiles are needed (for different media), it is just too practical to use Qimage for also that last number crunching operation called colorspace conversion and resampling to the native printer driver resolution, which will then be in 8-bit/channel space. I assume colorspace is converted first, then resampling is done?  As said, rarely an issue, but it might be in some rare smooth gradient situations.


Yes, ProPhoto RGB is humongous, not only does it exceed any current output gamut capability, it is also revealing to know how little actual image data of the full output media's gamut is actually present in most images, except for the most saturated cases (like some flowers). That's why I often use the Beta RGB workingspace (http://www.brucelindbloom.com/index.html?BetaRGB.html) as a slightly more reasonable sized gamut colorspace. It's much less likely to cause quantization related issues, and still offers more color resolution than Adobe RGB does.

Thanks for your help.

Cheers,
Bart

Super. No cutting of corners for speed reasons, but rather quality, first and last.


Indeed, I forgot about that one.


Mike, thanks again.

Cheers,
Bart

Hi Ernst,

It has (16-bit) gamma 2.2 TRCs, just like BetaRGB and AdobeRGB. Oherwise it is somewhat similar in gamut trade-offs to BetaRGB.

Cheers,
Bart

I was thinking of pRGB, not ProPhoto. The last will suffer more in the conversion at 8 bit to the printer profile as already discussed in this thread. Makes no difference on your reply I guess as DCRAW will not allow pRGB either.

--
Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
April 2014, 600+ inkjet media white spectral plots.

Hi David,

Yes, should be no problem. Depending on your OS version, just make sure that Beta RGB (or pRGB, Printer RGB) is installed (right-click the profile and click Install, or copy/paste it) in the required location(s) so that Photoshop also picks it up, and it can also be embedded in a TIFF so it should be picked up by other systems.

Cheers,
Bart

I don't use the dither option for two reasons.  First, it is almost never needed in normal photographic prints and only seldom needed in mathematically derived gradients.  Second, in my own testing, the dithering can cause clotting and other artifacts that become more noticeable, taking an almost imperceptible level of banding on some mathematical test gradient and turning it "dirty" in appearance.  In other words, the dithering can in some cases become more noticeable than the slight banding it is trying to address.  I haven't retested it since version 1x of LCMS so I may retest again in the future to see if there are any benefits.

Mike

What would actually help is if you have a particular image in a particular color space where you notice posterization when converting to another profile.  I've learned over the years that certain images with certain profiles (particularly printer profiles that can vary wildly) can produce effect that are hard to reproduce.  The more samples I can get (image and the "to" profile) the better.

Mike