how to set negative black point without clipping?

[false_dmitrii]

Every time I scan the blank lead-in area of a negative, I get a wide
range of values. Depending on film type and camera used (auto
everything vs. fixed everything ), the following image data often
doesn't start until well above the average base value. But other
times, there's significant overlap. I've been sticking to the PS white
point dropper method for applying the film base color to a linear
negative, but I can't find a consistent way to avoid clipping when
using it. I tried using a fill layer set to "difference" blend; this
can prevent clipping, but it has a solarizing effect on any stray
outlying values, and I don't know if it's even remotely appropriate for
this task. The whole procedure isn't as straightforward as I thought.

It doesn't make sense to me that there's a lot of "white" data above
the average values of the clear negative base. I'm not confident that
there should be so much overlap with well-exposed daylight image data,
either (this is Kodak Royal Supra 400 again in my autoexposing Olympus,
so it has a wider dynamic range than the cheaper film/camera
combinations I've tried, but it still doesn't come close to filling the
histogram). It can't all be pure noise...can it?

Is it possible that the scanner lamp's constant warming could create
such a discrepancy? If not, am I doing something wrong in
postprocessing?

false_dmitrii

 

[Don]

Every time I scan the blank lead-in area of a negative, I get a wide
range of values. Depending on film type and camera used (auto
everything vs. fixed everything ), the following image data often
doesn't start until well above the average base value. But other
times, there's significant overlap. I've been sticking to the PS white
point dropper method for applying the film base color to a linear
negative, but I can't find a consistent way to avoid clipping when
using it. I tried using a fill layer set to "difference" blend; this
can prevent clipping, but it has a solarizing effect on any stray
outlying values, and I don't know if it's even remotely appropriate for
this task. The whole procedure isn't as straightforward as I thought.

It doesn't make sense to me that there's a lot of "white" data above
the average values of the clear negative base. I'm not confident that
there should be so much overlap with well-exposed daylight image data,
either (this is Kodak Royal Supra 400 again in my autoexposing Olympus,
so it has a wider dynamic range than the cheaper film/camera
combinations I've tried, but it still doesn't come close to filling the
histogram). It can't all be pure noise...can it?

A couple of points which may or may no be relevant... ;o)

I'm presuming you're scanning the blank area for each frame separately
using the same exposure you used for the frame itself?

If not, that may account for the difference/overlap.

Also, I don't trust the Photoshop color sampler (even with the 5 pixel
radius) so when I need to sample an area (e.g. to set the gray point)
I select a larger portion and then do a histogram on that selection.
That way I get much more accurate average/median values instead of
blindly clicking with the gray sampler and getting a different readout
each time.

Don.

 

[false_dmitrii]

A couple of points which may or may no be relevant... ;o)

I'm presuming you're scanning the blank area for each frame separately
using the same exposure you used for the frame itself?

If not, that may account for the difference/overlap.

I was using the empty leader frame prior to the first image on the
strip to set global per-channel exposure, then using the same exposure
settings for each following image. Scanned at a couple of notches
below avg. of 255 to avoid unanticipated clipping--not that I'd ever
see much of it, but I'd rather get things right on the first pass.
Then sampled the same empty leader for global black point. Exposure
gain stayed the same, but perhaps the lamp performance shifted too far?

Also, I don't trust the Photoshop color sampler (even with the 5 pixel
radius) so when I need to sample an area (e.g. to set the gray point)
I select a larger portion and then do a histogram on that selection.
That way I get much more accurate average/median values instead of
blindly clicking with the gray sampler and getting a different readout
each time.

Why not run an Average on a clear center crop of the leader frame?
Except for the original question of why this sometimes results in a big
endpoint spike in the image data hump if whitepoint is sampled to pure
white. Clipping is usually solved if sampled to some arbitrary white
in the 220-240 range, but doesn't that throw away half the purpose of
setting the film base black point in the first place?

Standard disclaimer that I don't actually have the foggiest idea what
I'm doing.

false_dmitrii

 

[Don]

I was using the empty leader frame prior to the first image on the
strip to set global per-channel exposure, then using the same exposure
settings for each following image. Scanned at a couple of notches
below avg. of 255 to avoid unanticipated clipping--not that I'd ever
see much of it, but I'd rather get things right on the first pass.
Then sampled the same empty leader for global black point. Exposure
gain stayed the same, but perhaps the lamp performance shifted too far?

That's a good point! Speaking of which, the scanner may also have
recalibrated causing even more shift!

I guess you could add a few empty leader scans sprinkled equidistantly
(say, every 6 frames) to allow for any fluctuations (lamp or
auto-recalibration). I know, it complicates things, but...

I scan my blank area immediately before or after I scan the frame. The
main reason for doing that is that I use a different exposure for each
frame because I experiment a lot. I also recalibrate manually before
each scan so the scanner doesn't do it "randomly" and "surprise" me.

Why not run an Average on a clear center crop of the leader frame?

Seriously, that's exactly what I do!

Except for the original question of why this sometimes results in a big
endpoint spike in the image data hump if whitepoint is sampled to pure
white. Clipping is usually solved if sampled to some arbitrary white
in the 220-240 range, but doesn't that throw away half the purpose of
setting the film base black point in the first place?

Indeed! Is this spike at 255? If it is it may still be clipping.

Another thing - which I found out the hard way - is that you have to
take an 8-bit histogram of a 16-bit scan with a boulder of salt!
Without going into details but - depending on how the 16-bit data is
squeezed into 8-bit bins - the resulting 8-bit histogram may exhibit
non-existent "ghost" spikes! Even recursive "waves" of spikes down the
whole length of the histogram!

That's why I implemented two different methods of calculating 8-bit
histograms from 16-bit data in my histogram program: the conventional
adding up of 256 16-bit bins into one 8-bit bin (this may cause
"ghosting") as well as my own "averaging" which results in a much more
smooth 8-bit histogram. Actually the two shapes can differ quite a bit
depending on source data!

Oh, yeah! One more! ;o) In Photoshop (version 6 here) the Histogram
function (as opposed to composite RGB histogram in Levels) "does
things". The Histogram uses Luminance values of individual channels
(30% R, 59% G, 11% B) to calculate the histogram. (BTW, Threshold does
the same!) This makes the resulting histogram quite different from the
cumulative (i.e. Levels) histogram. In some cases using Luminance
actually "distorts" the histogram quite a bit!

Standard disclaimer that I don't actually have the foggiest idea what
I'm doing.

As I like to say: I'm hopelessly lost, but making good time! ;o)

Don.