Kenneth, this is my first post to this forum - but this is a subject which I have had to deal with on more than one occasion.
You mention data loss, but this is a rather general term and can mean many things to many people (original or digitization). My post only addresses the computer side of things and not the original.
This method I mention can be used to help reduce these destructive edits, by using two important principles - 'by the numbers' colour correction and separating colour from tone.
The first step to colour correction is global edits, such as setting endpoints and neutrals. If these regular methods do not fix things, or are too extreme - then other approaches might be necessary before these basic global corrections take place (some images contain too much of a cast and extreme edits which are needed to correct colour may impact on tone).
I have uploaded a ZIP file of the Photoshop layered file, so that APS users can see for themself the correction steps. I will outline these below, so that users of other software can apply the method using similar tools in their other software:
http://members.ozemail.com.au/~samar..._colourfix.zip
The file starts in RGB and does not leave this mode - this method exploits LAB/HSB etc methods without changing modes to these alternate spaces which separate colour from tone.
i) Remove low level colour noise and jpeg compression artifacts -
Dupe background layer, change layer blend mode to color mode - so that only hue/saturation values are blended into the underlying colour and tonal values.
Corrective filters such as despeckle, median, smart blur (edge protected) and gaussian blur are all used at minimal settings.
The separate R, G and B channels can then be viewed, while the correction layer is toggled on/off to 'animate' the before after results. This is much better than just blurring the B channel or whatever.
http://members.ozemail.com.au/~binar...ctremoval.html
ii) Rough neutralization of cast, white balance -
Regular endpoints and neturals hosed the image, as this is an extreme correction. A cast removal method which simulates the edits of LAB mode (without the extreme gamut problems and rounding errors) was then perfromed.
The image is currently in working space RGB, so equal values of R=G=B as in 128R 128G 128B indicate a neutral tone. The shirt was presumed to be white - so a fixed colour sampler was placed in this highlight endpoint so that the info palette would provide permanent interactive feedback on the cast correction.
The eyedropper tool was used to sample the yellow cast colour, then a new layer was made and filled with this solid light yellow colour, over the original image. This layer is then inverted, so that yellow becomes blue. The layer mode is then changed from normal to color (hue/sat only). This will reveal the image tones with a strong blue cast.
The opacity of this inverted colour blend layer is then adjusted, while the values quoted by the fixed sampler in the info palette were evaluated. The most neutral result was when the layer was around 50% opacity - but this will obviously vary on different images...but the numbers don't lie - so strike a happy middle ground between your monitor and the numbers.
iii) Final Curves -
Fixed colour samplers were also placed on the shadow endpoint and on a 'memory colour' as in a skintone.
http://www.execpc.com/~rbean/color/color4.txt http://www.execpc.com/~rbean/color/
A curves adjustment layer was then added, and each individual R, G and B channel was edited so that the image had a full print range in critical image detail and was neutral. The highlight was expanded to around 245RGB and the shadow endpoint was adjusted to around 50RGB. This is a more conservative shadow density but it still provides good range for print in the critical image areas.
Adjusting the skintone is best left for another post - but briefly, there are very rough aimpoints which use ratios of cyan to magenta to yellow. For example a deep caucasian tan might have a high magenta value, with the cyan being around 1/3 to 1/2 the magenta value, and with yellow being at least equal to magenta and probably running higher than magenta, perhaps similar to the cyan ratio. Even though the correction is done in RGB mode, the fixed colour sampler for the skintone only can be changed to read out CMYK values. I personally do not find it easy working in RGB values for non white/black/grays.
I have uploaded an example JPG to the forum with this post, for quick visual reference.
This is just the prep work before the retouching is done - which my example and this post do not address (they are rather mundane when compraed to the cast removal).
More on the theory behind 'scientific' colour correction can be found at this link (scroll down to curves, levels & colour correction):
http://members.ozemail.com.au/~binar...V_links.html#C
Hope this helps.
Stephen Marsh.