Ernst Dinkla wrote:
I think that the straight line from the high point in the 650-700 nm through the lowest point between 450-520 nm may stay too high and straight beyond that lowest point at the short waves side, papers without FBA tend to curve more downwards there. It may be a range with less practical value for profile creation though. Where you verified the FBA compensation with different illuminants it could be interesting to measure an FBA free paper with a UV measuring spectrometer and apply the algorithm in that case too and see what is altered where no change should happen. Creating that curve from a third value between 450 and 700 nm will not work, papers vary too much on that range and I see no relation between that range and the lower than 450 nm range.
The current code has been tuned slightly from the algorithm presented in the paper, although the gist of it is the same. There is no doubt that the details are influenced by the characteristics of the paper I was able to test. If you have any examples of papers where you think that the current code isn't doing a good job, by all means send me the spectral .ti3 files so that I can take a look at it.
While reading I got the idea that the UV filtering of colorants could be taken into account.. To refine the estimation of the FBA influence it could be a route to explore further. I guess the UV filtering and fluorescence properties of the most common colorants is a shorter list than all the paper reflectancy data and it can be made more absolute if the software gets the ink brand name so use the data of that ink. Ink layer thickness being the unknown then with varying media presets.
The approach has the advantage that it makes a correction to the measured reflectivity, so that even if the correction isn't marvellously accurate, it tends to move in the right direction, and doesn't do much damage. I suspect that measuring the illuminant using the method in V1.2.0 will also tend to cancel errors out. Graeme Gill.