196 lines
7.5 KiB
HTML
Executable File
196 lines
7.5 KiB
HTML
Executable File
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
|
|
<html>
|
|
<head>
|
|
<title>The i1pro Driver</title>
|
|
<meta http-equiv="content-type" content="text/html;
|
|
charset=windows-1252">
|
|
<meta content="Graeme Gill" name="author">
|
|
</head>
|
|
<body>
|
|
<h2 style="text-decoration: underline; font-weight: bold;">How can I
|
|
have confidence in the i1pro Driver ?<br>
|
|
</h2>
|
|
A question that has been asked is : "<span style="font-weight:
|
|
bold;">You've
|
|
written your own driver for the Eye-One Pro. How can I have
|
|
confidence
|
|
that the measurements are accurate, and will match those made with
|
|
the
|
|
original manufacturers driver </span>?"<br>
|
|
<br>
|
|
This is a quite reasonable question. The following attempts to
|
|
answer
|
|
it.<br>
|
|
<h4 style="text-decoration: underline;">Why does Argyll use it's own
|
|
i1pro driver ?</h4>
|
|
Primarily because the Original Manufacturers Driver (OMD) isn't
|
|
available for all the platforms that ArgyllCMS supports (Linux in
|
|
particular). A side benefit is that it's possible to tweak many of
|
|
the
|
|
driver parameters for slightly better results and more flexibility.
|
|
It
|
|
has also helped in understanding the characteristics and limitations
|
|
of
|
|
such instruments.<br>
|
|
<h4 style="text-decoration: underline;">Does it match the OMD ?</h4>
|
|
In principle the behaviour should be very similar. While the Argyll
|
|
driver has been written from scratch, it does use exactly the same
|
|
calibration values from<br>
|
|
inside the instrument, and attempts to use the calibration values
|
|
and
|
|
process the raw instrument readings in an equivalent manner to that
|
|
of
|
|
the OMD.<br>
|
|
<br>
|
|
But the proof of the pudding is in the measuring, so to actually
|
|
verify
|
|
this, the following experiment was conducted:<br>
|
|
<br>
|
|
The Argyll version used was V1.2.0<br>
|
|
The OMD is the original version prior to the introduction of the
|
|
i1pro2, and hence reporting the native instrument measurements,
|
|
rather than applying a conversion to the XRGA standard<br>
|
|
<br>
|
|
The Macbeth 24 patch ColorChecker was used as a sample target. For
|
|
each
|
|
patch (and the calibration tile), the following steps were
|
|
performed:<br>
|
|
<br>
|
|
1) Place the instrument on the calibration tile.<br>
|
|
<br>
|
|
2) Use Argyll spotread to calibrate the Argyll driver.<br>
|
|
<br>
|
|
3) Change drivers to the OMD.<br>
|
|
<br>
|
|
4) Use the OMD to calibrate the instrument.<br>
|
|
<br>
|
|
5) Move the instrument to the patch on the ColorChecker.<br>
|
|
<br>
|
|
6) Read the color using the OMD.<br>
|
|
<br>
|
|
7) Change the back to the Argyll driver.<br>
|
|
<br>
|
|
8) Using the calibration made in step 2), read the color using
|
|
Argyll.<br>
|
|
<br>
|
|
Each calibration or reading was performed 15 seconds from the
|
|
previous
|
|
one, to put the instrument lamp in a repeatable state.<br>
|
|
The instrument was kept in exactly the same position for calibration
|
|
and patch measurement with the two drivers.<br>
|
|
(The whole idea is to reduce all other sources of error, other than
|
|
the
|
|
driver itself.)<br>
|
|
<br>
|
|
This measurement was repeated just once for each patch + the
|
|
calibration tile. This was done in one run, and the readings were
|
|
not
|
|
specially selected.<br>
|
|
<h4 style="text-decoration: underline;">Results:</h4>
|
|
The following D50 L*a*b* values were recorded for each measurement:<br>
|
|
<br>
|
|
A) The OMD internally calculated L*a*b* value<br>
|
|
B) The L*a*b* value calculated by Argyll from the OMD
|
|
spectral values.<br>
|
|
C) The L*a*b* value calculated from the Argyll measured
|
|
spectral values.<br>
|
|
D) The L*a*b* value calculated from the Argyll
|
|
Hi-Resolution mode measured spectral values.<br>
|
|
<br>
|
|
<span style="text-decoration: underline;">A is compare to B, to
|
|
check
|
|
that the spectral to standard observer calculations are
|
|
equivalent.</span><br>
|
|
<br>
|
|
The result was an average Delta E (CIE76) of
|
|
0.006,
|
|
with a maximum of 0.012.<br>
|
|
<br>
|
|
This shows that there is very close agreement in
|
|
the
|
|
way spectral values are converted to XYZ and L*a*b*.<br>
|
|
<br>
|
|
<span style="text-decoration: underline;">B is compared to C to
|
|
check
|
|
that the Argyll driver behaves the same as the OMD.</span><br>
|
|
<br>
|
|
The result was an average Delta E (CIE76) of
|
|
0.028,
|
|
with a maximum of 0.051.<br>
|
|
<br>
|
|
This shows that the OMD and Argyll driver are in
|
|
close agreement in spectral measurement.<br>
|
|
This error is an order of magnitude smaller than
|
|
uniformity induced errors typical in the media being measured.<br>
|
|
<br>
|
|
<span style="text-decoration: underline;">A is compared to C to
|
|
check
|
|
that the Argyll driver and spectral to XYZ differences don't
|
|
compound.</span><br>
|
|
<br>
|
|
The result was an <span style="font-weight:
|
|
bold;">average</span>
|
|
Delta E (CIE76) of <span style="font-weight: bold;">0.026</span>,
|
|
with
|
|
a <span style="font-weight: bold;">maximum</span> of <span
|
|
style="font-weight: bold;">0.048</span>.<br>
|
|
<br>
|
|
Rather than compounding, any spectral to XYZ
|
|
differences tend to cancel
|
|
out slightly. This is the <span style="font-weight: bold;">bottom
|
|
line</span>
|
|
experimental difference between
|
|
the two drivers. The actual underlying difference may in fact be
|
|
less
|
|
than this, but it would be necessary to do multiple test runs to
|
|
filter out experimental error.<br>
|
|
<br>
|
|
<span style="text-decoration: underline;">C is compare to D to check
|
|
that the Argyll Hi-Resolution mode is behaving reasonably.</span><br>
|
|
<br>
|
|
The result was an average Delta E (CIE76) of
|
|
0.158,
|
|
with a maximum of 0.353.<br>
|
|
<br>
|
|
Because the ColorChecker samples have relatively
|
|
smooth reflectance spectra, it can be expected that<br>
|
|
the normal and Hi-Res mode results should be
|
|
fairly
|
|
similar. And indeed, this is the case. The biggest<br>
|
|
differences are for patches
|
|
with the largest spectral transitions in them, which is to be
|
|
expected
|
|
as the<br>
|
|
Hi-Res measurement more
|
|
closely follows the spectral shape, while the differences for
|
|
spectrally flat<br>
|
|
patches is neglegable, since both can follow the
|
|
spectral shape well.<br>
|
|
<br>
|
|
Example Yellow-Green Patch, Hi-Res & Normal spectrum:<br>
|
|
<img style="width: 709px; height: 259px;" alt="Yellow-Green patch,
|
|
Hi-Res vs. Normal" src="YellowGreen.jpg"><br>
|
|
<br>
|
|
<h4 style="text-decoration: underline;">Conclusions:</h4>
|
|
The experimental average difference of <span style="font-weight:
|
|
bold;">0.026</span>
|
|
Delta E76 shown above provides evidence that despite using a
|
|
completely
|
|
different instrument driver to that supplied with the instrument,
|
|
the
|
|
ArgyllCMS Eye-One pro measurement values have comparable accuracy,
|
|
and
|
|
can be relied upon to match measurements made using the original
|
|
manufactures driver.<br>
|
|
<h4 style="text-decoration: underline;">Raw Data:</h4>
|
|
The raw data is available in this <a href="i1proDriver.xls">spread
|
|
sheet</a>.<br>
|
|
<br>
|
|
<br>
|
|
<br>
|
|
<br>
|
|
<br>
|
|
<br>
|
|
</body>
|
|
</html>
|