Pages: 1 [2]   Go Down

Author Topic: Shadow Improvement stumper?  (Read 3042 times)

Mick Sang

  • Full Member
  • ***
  • Offline Offline
  • Posts: 170
Re: Shadow Improvement stumper?
« Reply #20 on: June 10, 2019, 10:53:49 am »

Quote
Upon closer look, the large b*/a* deviations that occur at low L* on the M3 data are exaggerated by the low L*. When the data is looked at through the M2 data set, the a*/b* deviations become much smaller. OTOH, the L* slope changes are more problematic. Turns out the major difference is a lumpy L* instead. There is a significant slope change in L* between about 5 and 15 on the M3 profile. The profile tracks this and alters the smooth transition BPC affords the M2 profile which doesn't have that slope change. Interesting stuff.

Here's an interesting graph. It's the L* of the M3 and M2 perceptual profiles (BtoA0) that is reported by the M2 RelCol profile (AtoB1), which more closely approximates what we actually see, when plotting requested L* (what would be in an image).

The blue line is the M3 Perc. response, the orange line is the M2 Perc. response. The reason is that the M3 measurements show a large increase in L* of from 3.5 to 8 between rgb 3,3,3 and 10,10,10. This is interpreted as a real change, not an artifact of polarization rejecting specular reflections, which our eyes can't do. The result is a lowering of the actual print response.


Another problem is apparent gamut increase. Also attached is a graph of the a*/b* gamut at L*=30. The outer circle is the M3 profile and it shows that it can print colors about twice as saturated as the M2 gamut. It can't. In fact the gamut boundaries are limited by the device RGB values when they clip at either 255 and or 0 and those clipping points are the same regardless of whether the spectra is measured with polarized filters or not.

Doug,

Thank you very much for taking the time to provide this investigation of the anomaly which I encountered with Hahnemuhle Museum Etching. It is fascinating and bewildering at the same time. I’m not a colour scientist and have much still to learn about the various levels and interactions of "A"s to "B"s. But,  I think I get the gist of your findings. Last thing first, the gamut volume is falsely inflated. In an earlier post, you pointed out that the gamut volume for M3 reading condition is “fiction.” That much I understood from the start although I don't really know exactly how that occurs. Nevertheless, since improvements to shadow rendering has been the goal, I simply accepted and ignored that effect.

Furthermore, the L* of M2 perceptual response is far more linear in the shadows than the L* of the M3 perceptual response which contorts in such a way as to subdue shadow rendering relative to the M2 readings. I hope I’m on track so far. If so, this makes perfect sense. But, I wonder what to put this down to, since all other papers that I have profiled using M3 have presumably similar features in terms of the analysis of the readings, certainly in terms of the gamut volume anyway. However the “wonky” shadow result is probably not present or at least closer to or better than the M2 for those papers since apparent shadow detail is improved in the prints. Is it the moderate texture of the paper throwing the readings for a loop?

On another note, if you don't mind sharing, what software did you use to make your analysis?

Thanks again,

Mick
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Shadow Improvement stumper?
« Reply #21 on: June 12, 2019, 07:55:22 pm »

Doug,

Thank you very much for taking the time to provide this investigation of the anomaly which I encountered with Hahnemuhle Museum Etching. It is fascinating and bewildering at the same time. I’m not a colour scientist and have much still to learn about the various levels and interactions of "A"s to "B"s. But,  I think I get the gist of your findings. Last thing first, the gamut volume is falsely inflated. In an earlier post, you pointed out that the gamut volume for M3 reading condition is “fiction.” That much I understood from the start although I don't really know exactly how that occurs.
The apparent gamut volume increase from M3 occurs because specular reflections are removed by the cross polarization. I posted the plots because the question comes up from time to time whether M3 actually does increase the printable gamut. So its for the benefit of future readers.
Quote


Nevertheless, since improvements to shadow rendering has been the goal, I simply accepted and ignored that effect.

Furthermore, the L* of M2 perceptual response is far more linear in the shadows than the L* of the M3 perceptual response which contorts in such a way as to subdue shadow rendering relative to the M2 readings. I hope I’m on track so far. If so, this makes perfect sense. But, I wonder what to put this down to, since all other papers that I have profiled using M3 have presumably similar features in terms of the analysis of the readings, certainly in terms of the gamut volume anyway. However the “wonky” shadow result is probably not present or at least closer to or better than the M2 for those papers since apparent shadow detail is improved in the prints. Is it the moderate texture of the paper throwing the readings for a loop?

That odd, non-linear effect for that paper with M3 is quite curious. It may be the ink distribution change. Or perhaps the paper's texture is somehoe interacting with the ink load.  You might try altering the ink saturation and evaluating patches of RGB 0,0,0  2,2,2, 4,4,4, … 20,20,20 and set it at what achieves the most linear M3 response. This should give better deep shadow prints.
Quote

On another note, if you don't mind sharing, what software did you use to make your analysis?

I use a bunch of custom and some built-in MATLAB functions. It's very interactive and I can easily pull out whatever stuff is of interest and explore.

For instance the following evaluates both the M3 profile and M2 profile for L*=30

     ProfileConvert(ProfileConvert([30 0 0], prof1, 'f', 0), prof2, 'r', 3)

Produces (L*a*b):  35.5346    0.3633    1.5430

This will be quite close to what is actually measured with a M2 spectro.

It evaluates the M3 profile using the Perc. LUTs to determine what RGB value is sent to the printer when requested to print L=30 in Perc. It then takes that RGB value and uses the M2 profile's AtoB1 tables to determine what the actual L*a*b* color gets printed by that RGB. as seen w/o de-polarization with normal 0/45 illumination.

Mostly I use a variation of this to evaluate what the effect of M1 v M2 is on paper with high OBA content. Evaluate one profile then the other for the actual environment.

I'm willing to share my custom code to anyone that is also using MATLAB with the image processing toolbox. MATLAB is pricey for any sort of formal use but reasonable for home use that is completely non-commercial. I use it because it was a critical tool for the kind of engineering work I once did. But now I'm retired and just do this stuff for fun.
« Last Edit: June 12, 2019, 08:09:49 pm by Doug Gray »
Logged
Pages: 1 [2]   Go Up