Comments on Calibration

The threshold for human discriminability between two test swatches is a deltaE of about 2 or 3. Our mean deltaEcmc values between the two test sets, which we expect to be more accurate than deltaEab, is 7.0652. While significantly above 3, this is a fair judgement of our conversion functions for two reasons:

Our gamma function is an approximation. As seen in its graph, the curve does not fit exactly.

Our rgb2xyz transformation function was only calculated for the 24 swatches on the Macbeth ColorChecker, and thus may not be entirely accurate.

We can confirm the imprecision of our conversion functions by noting the deltaE between our the L*a*b* coordinates calculated from our nonlinear RGB Macbeth values and those from our spectroradiometer values. Since these were the values used to create the conversion functions in the first place, there should be a much smaller deltaE between them. The mean deltaEcmc between them is actually 6.0001. While better than the results for our test swatches, this still indicates that our conversions are not exact.

start back next