F
false_dmitrii
After a lot of aggravation and $300 for Photoshop CS2's color
management capabilities, I'm close to understanding the steps necessary
for correcting a linear "raw" scan from the SE5400. I'm still not sure
about the best way to fit them all together, though. The goal is to
convert the negatives accurately with as little tweaking as possible,
avoiding guesswork and preserving image color casts and
under/overexposure until they're the only things left to consider. How
does this sound:
1. Scan as linear positive with per-channel exposure bringing the film
base close to white. In most recent test scans, I stopped at average
base RGB values of 240-244 to avoid clipping image data (see below).
2. Bring into CS2 and assign Minolta's SE5400 posilinear profile.
3. *Convert* to a standard working space, otherwise remaining steps
don't look right (I've been using Ektaspace PS5, but don't really know
if it's an appropriate choice here). Maybe I have a CS2 color
management setting wrong (for instance, "use gamma 1 internally" is
checked in the CM dialog...).
4. Levels adj. layer, using white point dropper to sample a
same-exposure scan of the film leader (also taken through steps 2 and
3).
5. Not sure about this one. Either a levels adj. layer with an RGB
gamma boost, or a levels adj. layer with separate RGB graypoint
adjustments using characteristic curve values I dug up from
http://www.cacreeks.com/films.htm. I don't know if I'm using these
values at all correctly, but they seem to do a very good job of
balancing out the color by themselves.
6. Whatever's left from step 5.
7. Invert adj. layer.
8. Curves adj. layer to fix contrast to taste. The previous steps are
usually good enough that a modest S-curve variant will suffice, which
gives me some confidence in the direction I'm taking.
Naturally, each layer is added above the previous ones.
The biggest points of concern:
1. If I absolutely have to convert to a new color space to get useful
results, is Ektaspace the best general-purpose compromise between
preserving the original film+scanner gamut and not going too wide?
2. On the recent scans I made to test all this, sampling the film base
black point (as white point) causes significant histogram clipping.
The unadjusted linear scan of the blank film leader has what looks like
a fairly wide histogram (around 50 levels of data, though I'll bet some
of that's related to being linear), and only the lightest points
produce black point adjustments without much of a clipping spike. Not
sure why this is, though I wonder if the scanner lamp continuing to
heat up could have contributed. Film is Kodak Royal Supra 400, which
seems to contain a greater dynamic range than regular consumer stuff
but still falls well short of filling the histogram end-to-end.
Setting white point sampler to Gray 240 before sampling avoids the
clipping. Issue was less pronounced but still present on some Max
Versatility 400 scans. What's the most reliable way to sample a
variable film base?
3. I can't get the linear-to-monitor-gamma step to come out
consistently enough using Levels gray point. .45 is sometimes enough,
but some earlier scans (with pure white film base via exposure) needed
much lower (.30-.20). Due to above black point problem? Is CS2 Levels
really the right way to make this adjustment, or do I need to come up
with a true 2.2/.45 gamma Curves preset?
4. Not confident about applying the characteristic curve values as
Levels channel gray points...but the color comes out looking really
good! What's more, when I tried this on an earlier scan that
apparently *wasn't* linear (but I don't know what exactly it *was*),
colors and brightness came out just about perfect with *no* additional
curve or levels adjustments. Can't seem to reproduce this in other
images--see problem #3 above. Usually I need to tweak the channel
levels or white points a little, which is why I'm still trying to
improve steps 5 and 6. On linear scan, applying the characteristic
curve level adjustment without additional gamma compensation squeezes
the histogram to one side, but when it's re-expanded via gamma
adjustment, colors look close to ideal. Is the histogram
compression/expansion a natural side-effect of this method and not a
sign of serious problems in my approach?
5. If I add or even substitute an adj. layer white point sampler to
steps 5-6 (or apply it via the step 8 curve layer), I can sometimes get
equally compelling results. Setting aside subjective color adjustments
to the final image, is a white point adjustment *necessary* to achieve
reasonable fidelity to the negative? In place of or in addition to the
characteristic curve step, and before or after gamma adjustment?
6. Just generally confused at this point by all the photography,
optics, and CS2 oddities flying over my head. Every little change in
the order of the correction steps and the values used has a major
impact on the outcome. The effect of color space conversion on the
adjustment layers was particularly startling--images went from "okay
with work" to the best I've seen from my own hand. I'm hoping that the
above workflow is sound and in the right order, since with recent
results I finally have some slight hope of seeing all this through.
How does it look to the experts?
Thanks for your patience, as mine is getting worn down by all this.
false_dmitrii
management capabilities, I'm close to understanding the steps necessary
for correcting a linear "raw" scan from the SE5400. I'm still not sure
about the best way to fit them all together, though. The goal is to
convert the negatives accurately with as little tweaking as possible,
avoiding guesswork and preserving image color casts and
under/overexposure until they're the only things left to consider. How
does this sound:
1. Scan as linear positive with per-channel exposure bringing the film
base close to white. In most recent test scans, I stopped at average
base RGB values of 240-244 to avoid clipping image data (see below).
2. Bring into CS2 and assign Minolta's SE5400 posilinear profile.
3. *Convert* to a standard working space, otherwise remaining steps
don't look right (I've been using Ektaspace PS5, but don't really know
if it's an appropriate choice here). Maybe I have a CS2 color
management setting wrong (for instance, "use gamma 1 internally" is
checked in the CM dialog...).
4. Levels adj. layer, using white point dropper to sample a
same-exposure scan of the film leader (also taken through steps 2 and
3).
5. Not sure about this one. Either a levels adj. layer with an RGB
gamma boost, or a levels adj. layer with separate RGB graypoint
adjustments using characteristic curve values I dug up from
http://www.cacreeks.com/films.htm. I don't know if I'm using these
values at all correctly, but they seem to do a very good job of
balancing out the color by themselves.
6. Whatever's left from step 5.
7. Invert adj. layer.
8. Curves adj. layer to fix contrast to taste. The previous steps are
usually good enough that a modest S-curve variant will suffice, which
gives me some confidence in the direction I'm taking.
Naturally, each layer is added above the previous ones.
The biggest points of concern:
1. If I absolutely have to convert to a new color space to get useful
results, is Ektaspace the best general-purpose compromise between
preserving the original film+scanner gamut and not going too wide?
2. On the recent scans I made to test all this, sampling the film base
black point (as white point) causes significant histogram clipping.
The unadjusted linear scan of the blank film leader has what looks like
a fairly wide histogram (around 50 levels of data, though I'll bet some
of that's related to being linear), and only the lightest points
produce black point adjustments without much of a clipping spike. Not
sure why this is, though I wonder if the scanner lamp continuing to
heat up could have contributed. Film is Kodak Royal Supra 400, which
seems to contain a greater dynamic range than regular consumer stuff
but still falls well short of filling the histogram end-to-end.
Setting white point sampler to Gray 240 before sampling avoids the
clipping. Issue was less pronounced but still present on some Max
Versatility 400 scans. What's the most reliable way to sample a
variable film base?
3. I can't get the linear-to-monitor-gamma step to come out
consistently enough using Levels gray point. .45 is sometimes enough,
but some earlier scans (with pure white film base via exposure) needed
much lower (.30-.20). Due to above black point problem? Is CS2 Levels
really the right way to make this adjustment, or do I need to come up
with a true 2.2/.45 gamma Curves preset?
4. Not confident about applying the characteristic curve values as
Levels channel gray points...but the color comes out looking really
good! What's more, when I tried this on an earlier scan that
apparently *wasn't* linear (but I don't know what exactly it *was*),
colors and brightness came out just about perfect with *no* additional
curve or levels adjustments. Can't seem to reproduce this in other
images--see problem #3 above. Usually I need to tweak the channel
levels or white points a little, which is why I'm still trying to
improve steps 5 and 6. On linear scan, applying the characteristic
curve level adjustment without additional gamma compensation squeezes
the histogram to one side, but when it's re-expanded via gamma
adjustment, colors look close to ideal. Is the histogram
compression/expansion a natural side-effect of this method and not a
sign of serious problems in my approach?
5. If I add or even substitute an adj. layer white point sampler to
steps 5-6 (or apply it via the step 8 curve layer), I can sometimes get
equally compelling results. Setting aside subjective color adjustments
to the final image, is a white point adjustment *necessary* to achieve
reasonable fidelity to the negative? In place of or in addition to the
characteristic curve step, and before or after gamma adjustment?
6. Just generally confused at this point by all the photography,
optics, and CS2 oddities flying over my head. Every little change in
the order of the correction steps and the values used has a major
impact on the outcome. The effect of color space conversion on the
adjustment layers was particularly startling--images went from "okay
with work" to the best I've seen from my own hand. I'm hoping that the
above workflow is sound and in the right order, since with recent
results I finally have some slight hope of seeing all this through.
How does it look to the experts?
Thanks for your patience, as mine is getting worn down by all this.
false_dmitrii