Re: Accuracy of the Munsell Neutral Value Scale
Re: Accuracy of the Munsell Neutral Value Scale
- Subject: Re: Accuracy of the Munsell Neutral Value Scale
- From: "dpascale" <email@hidden>
- Date: Sat, 17 Apr 2010 17:50:13 -0400
Ben,
With CT&A you can get Munsell equivalents to RGB or L*a*b* values with
fractional precision, and you can select one of 4200 Munsell patches, for
which you can get the XYZ, etc. equivalents for many Iluminants and RGB
(Note: access to Munsell data in the CT&A Color Decks requires the
registered version). However, you cannot assign fractional Munsell
coordinates as input.
There is an equation that gives the Munsell Value vs Y. You can find it in
the CT&A manual, p 172 of the pdf version (Note: the exponents, (1/3 , 2
, 2.3 , and 1/3) written with a smaller font size, are not shown at their
proper upperscript positions; they are too low; I will correct this in the
next release).
You can also select RGB=0,0,0 and increase Y in CT&A while looking at the
Munsell equivalent; this will give you the equivalent Y (and L*) for every
0.1 Munsell Value Step.
#3 certainly seems to be at least a marginal factor. Here's a sequence of
readings of the reference tile, each taken a second or so apart:
This sequence shows a very slight tendency to increased Y (and L*) with
successive measurements, but nothing to be concerned with.
My original reading was 92.528793 -0.681620 0.018600, which would seem to
be well within rounding error.
Agreed.
I then made several measurements of my Watch Your White target,
re-positioning the i1 each time:
(...)
(That reminds me -- while I've got you on the line, is it safe to clean
the target? If so, how?)
Yes, it can be cleaned with liquid soap and lukewarm water, but carefully!
It is preferable to extract the target from the case. I will send you a pdf,
off-list, with instuctions.
Considering that the (PTFE?) target is within the margin of error (if
anything, a bit dim) yet the N 9/ patch is a bit bright, should the gray
scale patches be considered suspect?
As I mentioned in my previous post, the N9 measurement is just a tad too
bright relative to what we expect, but this does not make the gray scale
bad. The closer you get to white, the more difficult it gets. For instance,
from your measurement we see that the i1 calibration plate has L* = 96.1, a
standard value for the i1 tile (Note: perfect neutrality is not expected or
required for this tile since the measured value is compared to a reference
in the instrument firmware, which is different for each unit, and
compensated for when calibrating). Now, any surface contaminant or dirt will
lower L*, which will result in the instrument over-compensating all
subsequent measurements. For instance, I have seen slighly yellow and
slighly dirty calibration tiles resulting in too blue and too reflective
measurements of white surfaces, and then saw the results go back to normal
once the calibration tile was cleaned. Of course, it may be time for a
factory re-cal. Also, the i1 should be parallel to the measured surface and
in the same orientation as the one used for calibration (i.e. parallel to
the ground).
Danny
----- Original Message -----
From: "Ben Goren" <email@hidden>
To: <email@hidden>
Cc: "dpascale" <email@hidden>
Sent: Saturday, April 17, 2010 1:58 PM
Subject: Re: Accuracy of the Munsell Neutral Value Scale
Danny,
Thanks much for the clarifications!
Note 3: You should not assume that N5 corresponds to L*=50. L* is
generally around 10 * MunsellValue, but this is not an exact relation.
N 4.75/ is in fact very close to L*=49, as you measure.
I *was* going to ask where one can find the means to convert between
Munsell and Lab, as my Google skills weren't helping out much. Then I
finally stumbled across BabelColor CT&A...but I still don't see in it any
direct way of specifying HVC values...just RGB in various color spaces or
Lab. Am I missing something? Is that only available in the non-trial
version?
The error between your measument and the nominal value for N9 is:
1.54 CIEDE94.
Which is not as good, but the error is mostly in L*. This could be
explained by :
1-The backing used for measurement.
2- A dirty i1 calibration plate. (clean with pure alcohol).
3- The i1 used many times in a row (some models will show slightly
increased luminance when repeating a measurement, especially if ststing
with a "cold" instrument).
4- And the manufacturing tolerance...
I had just used the swatch book itself for backing. The calibration plate
looked fine and it hasn't been all that long since I last cleaned it, but
I gave it a good cleaning just to be sure.
#3 certainly seems to be at least a marginal factor. Here's a sequence of
readings of the reference tile, each taken a second or so apart:
Result is XYZ: 86.712801 90.127600 71.358082, D50 Lab: 96.049672 -0.348964
2.623734
Result is XYZ: 86.768327 90.186923 71.418420, D50 Lab: 96.074251 -0.351915
2.612415
Result is XYZ: 86.805921 90.225544 71.454580, D50 Lab: 96.090247 -0.351154
2.607828
Result is XYZ: 86.824961 90.246301 71.476349, D50 Lab: 96.098842 -0.352905
2.603288
Result is XYZ: 86.837370 90.259370 71.487486, D50 Lab: 96.104253 -0.353227
2.602715
Result is XYZ: 86.843630 90.265999 71.495016, D50 Lab: 96.106997 -0.353454
2.600751
Result is XYZ: 86.846647 90.267717 71.496233, D50 Lab: 96.107709 -0.350929
2.600896
Result is XYZ: 86.845584 90.267963 71.497499, D50 Lab: 96.107810 -0.353339
2.599946
So, here's the result of the N 9/ patch on top of a block of white
polystyrene (which itself reads 92.179090 -0.282941 -0.564617) after
making several readings of the tile and doing a re-calibration (by
quitting spotread and re-launching):
Result is XYZ: 78.489035 81.736977 67.267302, D50 Lab: 92.458646 -0.637077
0.145746
My original reading was 92.528793 -0.681620 0.018600, which would seem to
be well within rounding error.
I then made several measurements of my Watch Your White target,
re-positioning the i1 each time:
Result is XYZ: 93.959538 97.478963 80.449838, D50 Lab:
99.016891 -0.052186 -0.032418
Result is XYZ: 96.472871 100.151009 83.018264, D50 Lab:
100.058361 -0.160183 -0.325402
Result is XYZ: 94.262579 97.819508 80.956733, D50 Lab:
99.150673 -0.096481 -0.217445
Result is XYZ: 94.378732 97.945708 81.105167, D50 Lab:
99.200172 -0.106092 -0.253500
Result is XYZ: 94.333062 97.896360 81.015258, D50 Lab:
99.180821 -0.102774 -0.213348
(That reminds me -- while I've got you on the line, is it safe to clean
the target? If so, how?)
Considering that the (PTFE?) target is within the margin of error (if
anything, a bit dim) yet the N 9/ patch is a bit bright, should the gray
scale patches be considered suspect?
Thanks,
b&
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden