LAB -> RGB and adjustment layers

[jani]

I don't know of any way to avoid this if you feel that you have to save your PSD file in RGB format.

The easiest way around it is to save the PSD file in LAB.

When you need to convert to RGB, you create a copy of the LAB file, convert, and if you think you might use the RGB file more than once, save that under a different name, for instance "rubberandsteel_RGB.psd".

[PeterLange]

Every tone curve in any RGB space has a side effect on color saturation - period. Severity depends on many factors and sometimes we may even appreciate it for artistic reasons.  For example, blackpoint setting (to the right) increases color saturation, mainly for the shadows with decreasing significance towards the lights.

Luminosity blending or changing to Lab-lightness (harmless at 16 bit/ch) can typically reduce such effects – but not necessarily.  For example, increasing midtone brightness by means of the Levels’-“gamma” slider induces a broad de-saturation which isn’t much buffered by those measures.

Even, pure whitepoint settings are better done in normal RGB mode. In fact, this is the only exception with no side effect on color at all. Whereas Luminosity blending or Lab-lightness would lead to a de-saturation.

HSB could be an interesting option, however, color scientists don’t like it and PS doesn’t fully support it on an active basis.

Cheers! Peter

---

[omwoods]

Ok,
now that you've all shown me to be the amatuer here:

Ignoring LAB for my A-RGB files; I should perform curves as an adjustment layer, set to luminosity, but what about hue/sat? Normal blending mode? Sharpening last before print/ save as...

No detailed explanations please, just links or suggested reading; no point you sitting in front of a screen for five minutes typing out something written somewhere else... ::

Thanks again,
Oliver

p.s. digitaldog, currently browsing over your 'tips' page, another nights reading...

[Schewe]

On those images where I tested one round trip between Lab and RGB no image degradation popped out at me, but that of course doesn't mean very subtle changes didn't happen.

The first trip from RGB to Lab is the one that introduces the data drop-after the first trip, additional moves to and back from RGB>Lab>RGB don't do much if anything to a file.

As for exacly what happens to an RGB image that is converted to Lab, it all depends upon the colors in the image itself. The quantization errors are subtle...and as such may not show up as a problem until steps are taken to adjust the image _AFTER_ the RGB>Lab>RGB round trip.

If you are doing the conversions in 16 bit space, the quantization errors are less likely to introduce data loss that can lead to problems. That said, like anything else in Photoshop, the less erros you introduce into an image, the better the image integrety will be maintained. Thus, converting from RGB to Lab for the purpose of allpying a luminance based curves adjustment is foolish because essentially the same adjustment can be made in RGB using the curves layer in luminosity bland mode.

[bjanes]

The amount of data loss depends on the original gamut size and gamma of the working space. For example, if the working space is Adobe RGB, which has 256 values available, converting to 8-bit LAB reduces the data down to 234 values. The net result is a loss of 22 levels. Doing the same conversions from ProPhoto RGB reduces the data to only 225 values, producing a loss of 31 levels.

Rodney, those are interesting figures but could you explain the math behind the loss of exactly 22 and 31 levels? It might be instructive.

[digitaldog]

There’s a Levels calculator provided by Bruce Lindbloom on his site:
http://www.brucelindbloom.com/

[digitaldog]

My understanding is that in Photoshop we have 2^15 levels, which equals 32,768. The standard calculation for deriving the number of luminosity levels is 2 bits raised to the power of the bit depth.

Correct.

The high-bit representation in Photoshop has always been "15   1" bits
(32767 (which is the total number of values that can be represented by 15 bits of precision)   1).  This requires 16 bits of data to represent is
called "16 bit".