2002.03.05 02:20 "16 bit LAB in TIFF - documentation update", by Chris Cox

2002.03.07 13:55 "Re: 16 bit LAB in TIFF - documentation update", by Martí Maria


Taking the chance than kindly Adobe and Chris Cox gave to us to review the TN about Lab TIFF, I would like to make a proposal for the white point stuff. I must admit this is a subtle change. All but the most sophisticated, color-aware TIFF readers could ignore completely this stuff. However, Lab TIFF is being used for some others as a basis for color measurement and manipulation, so, for these this stuff does matter.

The actual TIFF 6.0 states that White point does not apply on Lab TIFF. My proposal is to change spec and make whitepoint tag required for all Lab TIFF using other that D50. This will keep Adobe TIFF in the spec as well.

Lab is encoded by doing first a normalization (X/Xn, Y/Yn, Z/Zn). This has several effects. The first and most important one is dynamic range is greatly expanded, second, there is a embrionary chromatic adaptation that gives a surprising property of Lab:

You can transport a Lab value between two white points by simply changing the destination white point. Gray balance is preserved.

Unfortunately, this chromatic adaptation is done in XYZ space instead of cone space, and this is not completely correct. Some folks call this a "Wrong Von Kries" regarding it as a perversion of Von Kries work.

Modern color applications does use more sophisticated appearance models for dealing this, like Bradford transform or some types of linear ones, with modified cone matrices.

In such way, the white point of Lab TIFF *does* matter, since a very smart reader needs to recover absolute colorimetry to then apply appearance model. Again this is not a normal case. A "advanced" reader capable to handle Lab could completely ignore all white point stuff without obtaining too bad results.

Martí Maria.