Photoshop print soft proofing under the hood

[rasworth]

I don't believe Adobe has published their algorithm for soft proofing, hopefully the expertise in this group can fill in the blanks.  Following is my best shot:

1.  Convert image pixel array of RGB values with embedded profile and white point to XYZ array.

2.  Convert XYZ array to XYZ array, using Bradford transform, from image white point to D50 white point.

3.  Convert XYZ array to standard (D50) Lab array.

4.  Convert Lab array to RGB array, using printer profile and selected rendering intent (B to A table).

5.  Convert RGB array to Lab array, using printer profile back path and absolute rendering intent (A to B table).

6.  Convert standard (D50) Lab array to XYZ array.

7.  Convert XYZ array to XYZ array, using Bradford transform, from D50 white point to display white point.

8.  Convert XYZ array to RGB array with display profile and white point, relative rendering.

9.  Compare print to display with lighting matching display white point.

Please jump in and comment/correct.

Richard Southworth

[digitaldog]

I don't believe Adobe has published their algorithm for anything other than openly documented file formats.
There's this very old blog post with comments by Chris Cox:
http://blogs.adobe.com/jnack/2007/03/adobe_cmm_released_ready_for_download.html

[Doug Gray]

Yes, that's how it should be done when "Show Paper Color" is selected in the soft proof dialog with one exception. Bradford adaptation is not needed for sRGB, Adobe RGB, etc. with D65 profiled displays because their primaries in the profiles are already Bradford adapted to D50 from their actuals for D65.

That said, Adobe CC has slight differences (ie: bugs) in the computational path that can alter soft proofs. I get the best results selecting Settings->DesaturateColors and the amount to desaturate set at 0. In theory this should show the same results as unchecking the DesaturateColors but in practice I see differences. Especially in low key areas near the black point.

Another way to approach it is to roll your own. To do this or cross check the accuracy of the view soft proof process, convert the image to device space using the desired Intent then convert back again to the working colorspace using Abs. Intent.

Oh, make sure you are using Adobe's ACE CMM. Microsoft's is buggy.

[rasworth]

Andrew,
Thanks for the reference, interesting read.

Doug,
Thanks for the correction, gives me something else to dig into and understand.

I did the round trip experiment, ProPhoto to Printer profile via Relative intent, and then back to ProPhoto via Absolute.  It provided a good match to the soft proof with "show paper color", although I'm still confused.  Is the Adobe CMM "smart" enough to use the A to B table for the Absolute Convert back to ProPhoto?  I had assumed it would B to A Absolute Convert intent.  More head scratching.

Richard Southworth

[Doug Gray]

I did the round trip experiment, ProPhoto to Printer profile via Relative intent, and then back to ProPhoto via Absolute.  It provided a good match to the soft proof with "show paper color", although I'm still confused.  Is the Adobe CMM "smart" enough to use the A to B table for the Absolute Convert back to ProPhoto?  I had assumed it would B to A Absolute Convert intent.  More head scratching.

Richard Southworth

You do have to select Abs. Col. when converting back to ProPhoto RGB, but yes, it uses the AtoB table.  BtoA goes from PCS (usually Lab) to device RGB. AtoB goes from device RGB to Lab  or XYZ for those rare printer profiles that use XYZ Luts). From there it's converted to ProPhoto, Lab or whatever space you selected. When you do the second convert the Intent applies to the source (device RGB in this case).

[rasworth]

Doug,

I understand, thanks.

Richard Southworth