Scanner Profiling:
(Draft page, further information will be added)
The following SW packages were used in the evaluation:
HW Setup:
- Scanner: Epson 2450
- OS: Windows 2000 Prof
Evaluation procedure:
I made an uncorrected scan with Silverfast and used it as reference for all
profiling packages (beside Vuescan). The scan was done with 300 dpi, gamma set
to 2.2 for the reflection target and 2.4 for the transmission target. All
packages used the same target from Wolf Faust (http://www.targets.coloraid.de/)
beside EZColor and WiziWYG which have proprietary reflection targets.
I generated (tried to generate) a profile with all packages. I assigned the
profile to the original scan of the Faust target and converted the RGB File to
Lab. For this I used ColorLab from GretagMacBeth (Download
of ColorLab). After this the scan (Lab format) was compared to the target
data and deltaE was computed. I used two different tools for this and both
generated similar results. One was the MeasureTool from the ProfileMaker 4.1
from GretagMacBeth. The other two was the profile checker from the Little cms
profiler package and IPhoto (Homepage)
.
I know deltaE is just one parameter to judge a profile, but I started with
this and others may follow.
During the evaluation I had several problems with the different packages.
It's currious that just commercial packages showed problems.
- EZColor generated profiles which contained tags which are not compliant
to the ICC specification. This lead to the situation that the Little cms
profiler package could not work with these profiles. ColorLab and IPhoto
seemed to have no problem.
EZColor could also not be used with absolute rendering intend in ColorLab.
The result was a all blue picture.
- I was not able to generate a transmission profile with WiziWYG because I
found no way to enter the data for the target. Somehow the program
expected always propriatary files (extension wyg). The documentation claims
that you should be able to profile transmission targets with targets from
other vendors.
- Also with Vuescan I was not able to generate a transmission target,
because it always said that the scan is not aligned or so and after 30 min
trying hard to get rid of the error message I stoped.
- ColorLab also rejected the Vuescan profile
The two following diagrams show the results for the Average deltaE and the
Max. deltaE. The results are from IPhoto (which used perceptual rendering
intend). The other evaluation tools showed similar results but did not work
with every profile. Please not that I added perceptual just because EZColor
profile did not work with absolute colorimetric.
Results for reflection target (perceptual rendering intend)
Results for transmission target (perceptual rendering intend)
Results for reflection target (absolute colorimetric)
Results for transmission target (absolute colorimetric)
Contents and attributes of the generated Profiles
The following attributes are not necessarily a measure of quality (large
tables do not mean good precision) but may give some insight in the quality of
the SW (flaws in the profile) and limitations (8bit tables can not meet the precision
of a 16bit table).
| Evaluation of reflection profile |
|
|
Tags
|
|
Size [kB] |
PCS |
CMM |
Ill |
desc |
dmnd |
dmdd |
cprt |
| lcms |
16 |
Lab |
lcms |
0,96/1,00/0,82 |
# |
# |
# |
# |
| AIM |
25 |
XYZ |
MSFT |
0,96/1,00/0,82 |
# |
- |
- |
# |
| ProfilePrism |
205 |
Lab |
MSFT |
0,96/1,00/0,82 |
# |
- |
- |
# |
| SilverFast Ai |
425 |
Lab |
Lino |
0,96/1,00/0,82 |
1) |
- |
- |
# |
| VueScan |
1 |
XYZ |
lcms |
0,96/1,00/0,82 |
# |
- |
- |
- |
| WiziWYG |
196 |
Lab |
MSFT |
0,96/1,00/0,82 |
# |
- |
- |
# |
| EZColor |
192 |
Lab |
- |
0,96/1,00/0,82 |
# |
- |
- |
# |
| BasICColor |
261 |
Lab |
appl |
0,96/1,00/0,82 |
# |
- |
- |
# |
| ProfileMaker 4.1 |
220 |
Lab |
appl |
0,96/1,00/0,82 |
# |
- |
- |
# |
| Epson Org |
193 |
XYZ |
KCMS |
0,96/1,00/0,82 |
# |
# |
# |
# |
|
rXYZ |
gXYZ |
bXYZ |
rTRC |
gTRC |
bTRC |
| lcms |
- |
- |
- |
- |
- |
- |
| AIM |
0,60/0,25/0,02 |
0,22/0,73/-0,18 |
0,12/0,02/0,95 |
4096 Points |
4096 Points |
4096 Points |
| ProfilePrism |
- |
- |
- |
- |
- |
- |
| SilverFast Ai |
0,52/0,19/0,00 |
0,27/0,54/0,08 |
0,25/0,00/1,24 |
G=2,2 |
G=2,2 |
G=2,2 |
| VueScan |
0,71/0,39/0,00 |
0,12/0,61/0,04 |
0,14/0,00/0,80 |
G=2,17 |
G=2,02 |
G=2,32 |
| WiziWYG |
0,60/0,24/0,01 |
0,29/0,76/0,00 |
0,07/0,00/0,81 |
256 Points |
256 Points |
256 Points |
| EZColor |
2) |
2) |
2) |
3) |
3) |
3) |
| BasICColor |
1/0/0 |
0/1/0 |
0/0/1 |
G=1 |
G=1 |
G=1 |
| ProfileMaker 4.1 |
0,62/0,23/0,00 |
0,21/0,76/0,00 |
0,12/0,00/0,82 |
515 Points |
515 Points |
515 Points |
| Epson Org |
4) |
4) |
4) |
G=1 |
G=1 |
G=1 |
|
A2B0 |
A2B1 |
A2B2 |
B2A0 |
B2A1 |
B2A2 |
| lcms |
16bit/20 Entries |
- |
- |
- |
- |
- |
| AIM |
- |
- |
- |
- |
- |
- |
| ProfilePrism |
16bit/21 Entries |
- |
- |
16bit/21 Entries |
- |
- |
| SilverFast Ai |
16bit/33 Entries |
16bit/33 Entries |
- |
- |
- |
- |
| VueScan |
- |
- |
- |
- |
- |
- |
| WiziWYG |
16bit/32 Entries |
- |
- |
- |
- |
- |
| EZColor |
16bit/25Entries |
16bit/25Entries |
16bit/25Entries |
- |
- |
- |
| BasICColor |
16bit/33 Entries |
- |
- |
- |
- |
- |
| ProfileMaker 4.1 |
16bit/25 Entries |
16bit/25 Entries |
copy of A2B0 |
-. |
- |
- |
| Epson Org |
16bit/32 Entries |
16bit/32 Entries |
16bit/32 Entries |
- |
- |
- |
Legend:
| 1) |
ICC Profile Inspector gives error message
for this tag |
| 2) |
Value 0,99/0,99/0,99 |
| 3) |
ICC Profile Inspector gives error message |
| 4) |
Value 0,00/0,00/0,00 |
| - |
not supported |
| # |
supported |
Features of Profiling packages
|
Monitor |
Scanner |
Printer |
16bit Raw data |
Parameters can be
influenced |
Profile Editor |
| lcms |
yes |
yes |
- |
- |
yes 2) |
no |
| AIM |
- |
yes |
- |
1) |
no |
no |
| ProfilePrism |
- |
yes |
yes |
- |
yes 3) |
no |
| SilverFast Ai |
- |
yes |
- |
internal |
no |
no |
| VueScan |
- |
yes |
yes |
internal |
no |
no |
| WiziWYG |
yes |
yes |
yes |
- |
no |
no |
| BasICColor |
4) |
yes |
4) |
yes |
no |
no |
| EZColor |
yes |
yes |
yes |
yes |
no |
for Printer Profile |
| ProfileMaker 4.1 |
yes |
yes |
yes |
yes |
yes 5) |
yes |
|
|
|
|
|
|
|
| 1) |
may work, but I did not get it to
work easily |
|
|
| 2) |
Mainly generation algorythm |
|
|
|
| 3) |
Gamma, Contrast, Saturation,
Color Bias, others |
|
|
| 4) |
Must be bought seperate |
|
|
| 5) |
Size and grey algorythm |
|
|
Discussion of the Results
EZColor and WiziWYG had a little drawback with the reflection target because
their profile was generated with a different target as they were evaluated
afterwards. But the transmission target shows similar results at least for
EZColor.
The first finding was that I was surprised that there seem to be still flaws
with the compliance with the ICC standard and that the support of the profile
header was quite weak. Special EZColor showed a number of problems.
I was also surprised by the wide range of results and that the freeware tool
somehow showed by far the best result (at least in the deltaE) evaluation.
But in the end at least all packages gave better results than the profile
supplied by Epson with the scanner.
Side remark: It is interesting that nearly all packages (beside the freeware
lcms) put their company name into the copyright tag of the profile. This would
mean also that the company which build the printing machine has the copyright on
a book?
For questions and comments please send an email to: info@tkupfer.de