On Feb 21, 2008, at 11:51 AM, Florian Höch wrote:
This time I will not assert anything at all, just post
a test case and
kindly ask for your (and of course, anybody else's) input on the test
case itself and the results.
That really doesn't work either, because then the test has zero
context and no assertion as to what it is supposed to demonstrate.
Scientists don't just do tests without stating a purpose for the
test, how the test is going to be carried out, potential problems
with the test, etc.
So now it is totally unclear what your point is in presenting this
data. The context is not stated. We just have data. So what?
Testing image: Linear gray ramp, Lab, 256 levels at 8bpc
http://hoech.net/files/Linear_Gray_Ramp_Lab.tif.zip
First you start out with data that is by definition encoded to a TRC
defined by L*. And you're going to end up in a space with TRC defined
by L*. It is obvious that you will have no loss, and it's because the
source data in 8bpc and the ending data in 8bpc have the same
encoding. It's exactly what I've been saying about the importance of
source-destination TRC match when the data is limited to 8bpc.
If you do this test with ECI-RGBv1 going to ColorMatch RGB you will
get the same number of gray levels in as you do out. 256.
Your test does *NOT* prove your assertion that the PCS TRC is
relevant in a high precision context. Case in point, if you uses ACE
for this test, your image data is being converted from LAB to XYZ
since that is the PCS for ECI-RGBv2. And XYZ uses linear encoding.
There is no gamma function, there is no L* function. The fact that it
didn't cause level loss is because the image data is at a very high
level of precision.
What I'd rather see is go back to the color test from previous
postings. A color space defined with NTCS primaries and varying TRC's
defined by gammas 2.3 through 2.6. So just like ECI-RGB, except
varying the TRC, and converting to the ISO Coated (ECI) profile, with
all the same parameters as the previous tests. Again, I'd do this
myself but it's not clear how to use the python script on Mac OS. I
have acquired PIL and installed it, but now when I run the script I get:
im = Image.open(sys.argv[1])
IndexError: list index out of range
Chris Murphy