While implementing an image-processing plugin, I learned something very interesting (to me) about the Shadows and Highlights develop settings. The other basic tone settings (exposure, contrast, saturation, etc.) always map any one color to another single color uniformly throughout the image. But the Shadows and Highlights may map a given color to multiple colors in a single image, depending on the location of the pixels.
For example, in a test image I made with Shadows = +100, pixel (824,286) had its original 16-bit color value (0, 0, 0) transformed to (1724,1387,852). Pixel (861,287) also had original color (0,0,0), but it was transformed to a much different color (3093,2688,1316).
Abstractly, the other basic tone settings are functions (any input color always has the same output color), while Shadows and Highlights are many-to-many mappings (any input color may have multiple output colors). I knew that Clarity was not a function -- the transformed value of a pixel would depend on neighboring pixels. But it never occurred to me that Shadows and Highlights weren't functions either.
Is this property of Shadows and Highlights well-known by expert users of LR? (I.e. Rob, I assume you knew of this?)
Yup - @Lr4 (PV2012) there are 3 adjustments using Adobe's new "magic" algorithm:
I really don't know enough to fill in much in the way of details, except to say that when you use -highlights or +shadows, it "clarifies" the highlights and/or shadows a little bit, by way of the laplace-transform based algorithm Adobe developed. Jeff Schewe has pointed us to a white-paper about it in the past, but I dunno the link.. To me, it's akin to Topaz Detail and/or Topaz Adjust's "adaptive" enhancements - two Photoshop plugins I used regularly until Lr4 came out (Adobe's algorithm is very well done - in my opinion, better than Topaz labs, or any of the others doing similar things, e.g. PhaseOne, DxO, Nikon, ..).
FWIW - this is one of the reasons some people were raising heck when Lr4 first came out - no way to brighten shadows using the basic sliders without getting the clarification effect.. That said, Eric Chan also has informed us that Lr3's fill light had a similar mathematical algorithm backing it, and so Lr4 was not new in that regard, just improved..
Thanks for confirming with the details. I vaguely recall that discussion now.