Sorry about taking some time to respond to this Rafe, but half way
through writing this message my computer "clicked", the screen went
black and it wouldn't reboot - not even as far as the BIOS or POST
beeps! After some tests I concluded the motherboard had died and,
having now replaced that and about 250 reboots later, with new drivers
being found and installed I found the ADSL driver had been corrupted and
I couldn't get net access! Now generally I keep my system CDs in a safe
place, but the CD for the ADSL modem wasn't filed where it should have
been, which meant searching through dozens of boxes of CDs! Eventually
I tracked it down and I am back online. Hopefully most of the other
peripherals still work, but I will have to try them in sequence -
haven't even checked if the Nikon works on the new firewire port yet!
rafe bustin said:
The more I think on it, the present problem
with my 2500 is more likely to be the white-strip
or black-strip than the lamp. Here's a pair of
samples:
http://www.terrapinphoto.com/2500band/normal.jpg
http://www.terrapinphoto.com/2500band/extreme.jpg
(The first "as scanned", the 2nd pushed with
curves to accentuate the banding. The scan is
from a 4x5 Portra negative.)
The banding is too consistent (along the height
of this image) to be caused by lamp flicker
during the scan.
I have no evidence that the Epson 4870
would be any better in this regard, though
I've heard no similar reports of banding
with the Epson.
If it's a problem with Microtek's calibration
firmware, no "alternative" driver is going to
fix this.
To me it looks like a white calibration point problem, since it appears
more in the highlights than in the shadows. Also, because it is broad
bands with soft edges, rather than discrete sharply defined lines, I
agree it looks like contamination of the white reference area - possible
with dirt or dust. I am no familiar with the scanner you are using, but
perhaps the calibration field has a glass surface area - perhaps
calibration on a glass film holder? For a slide scanner, calibration
should really be performed on an aerial image without any surface near
the focal plane.
Any ideas on how to deal with it through
post-processing? I'm stumped.
You could try a variation of Fernando's approach , but using a white
reference to correct the response characteristics. This is a little
more complicated than dark frame subtraction, because it requires both a
dark and a white (or at least light, unsaturated) frame, again captured
in linear gamma space. You need enough pixels in the frame as to ensure
that any random specs of dust in them (if you have a glass surface) are
negligible. As with Fernando's DFS, average these along the scan axis
retaining as much precision as possible, to create six linear arrays
(rgb black and rgb light). Then subtract the dark arrays from the light
ones to get three new linear arrays representing the response of the
individual cells in the scanner. What you want, is to transform this
into correction coefficients for the cell variations - or at least the
residual variations after the Microtek calibration has done its rather
poor job. This is achieved by calculating the average of each of the
three arrays and then calculating the correction factor from
C(i) = R/r(i) where C(i) is the correction to be applied to the i'th
cell, R is the average of all of the cells of that colour, and r(i) is
the response of the i'th cell that you previously estimated from the
difference between the light and dark arrays.
If you then multiply the image by this correction factor for each cell
it should correct the banding. This can be achieved in PS, but it is
probably easier to implement in a separate response correction program,
such as the DFS procedure that Fernando produced. You will get better
results if you also apply the DFS using the averaged dark frame first,
depending on the amplitude of the dark signal for each cell.
As I say, a little more complex, but trivial compared to programming the
whole thing in machine code (yes, machine code, not even CPU mnemonics!)
as I had to do about 25 years ago! Oh, when computers were simple
little beasts. ;-)
Now, back to testing that the drivers for this new motherboard hasn't
corrupted anything else on my system!
