It does not seem reasonable that colors would be expanded when converting from AdobeRGB to ProPhoto, since the purpose of color management is to keep colors the same. To test this phenomenon of expansion, I downloaded Bruce Lindbloom's synthetic color checker (which is in L*a*b). I converted to AdobeRGB in Windows Photoshop CC using the Adobe ACE color engine (the other option is the Microsoft CMM. I don't what options are available on the Mac, but colorsynch is likely an option).
I then converted the AdobeRGB image to ProPhotoRGB (relative colorimetric with dithering off) and used ColorthinkPro to assemble color lists for the two images and calculated Delta Es. A representative range of red values is shown. The RGB values are of course different, since we are using different color spaces, but the L*a*b values are virtually identical with Delta Es of less than one. This test fails to show any expansion. For information on colorlists, I would suggest referring to the Digitaldog's tutorial.
Hi Bill, your test using the synthetic colorchecker will not show the "colors expanding" phenomenon, since it does not contain colors outside of Adobe RGB, whether it is generated in Lab or not.
I downloaded another of Bruce Lindbloom's test images, An RGB Image Containing All Possible Colors
. Then I did the following:
1. Assigned ProPhoto RGB to it.
2. Convert that to Adobe RGB.
3. Re-convert that back to ProPhoto RGB.
It exhibits the exact same "colors expanding" phenomenon. I graphed the results in 3D in ColorThink, and also in the Color Worksheet. Unfortunately, the number of unique colors is too large (always more than 60, 000, since the target is made up of gradients) but the pictorial dE map shows a significant number of colors over 4 dE2k.
Please try it yourself.
Lundberg02's explanation by way of inverse transform is not very accurate, I believe, as the difference compared to the original ProPhoto RGB colors is too huge also. I have never tried this experiment before! These are unexpected results, and I hope to know why this behavior is observed.