Scanner Question about Histogram/(Level Graph)

  • Thread starter Thread starter narke
  • Start date Start date
N

narke

I send some slides to a local lab, they said they have a drum scanner.
But after the files came back, I was confused with their histogram.
Pixels are totally not spread soomthly on the whole range. On the two
end, [0-40], [200, 255], there is nearly no any piexl. In the range of
[190-200], the curve is very low, on the other hand, in the range of
[40-64], the curver is very high(density). And, the curve on the
lowest side was likely cut with a knift.

I asked the lab why the picture looks so poor in contrast and so dark
in illumilance, they said, they intent to do that in sake of post
processing.

Any ideal?

-
narke
 
I send some slides to a local lab, they said they have a drum scanner.
But after the files came back, I was confused with their histogram.
Pixels are totally not spread soomthly on the whole range. On the two
end, [0-40], [200, 255], there is nearly no any piexl. In the range of
[190-200], the curve is very low, on the other hand, in the range of
[40-64], the curver is very high(density). And, the curve on the
lowest side was likely cut with a knift.

I asked the lab why the picture looks so poor in contrast and so dark
in illumilance, they said, they intent to do that in sake of post
processing.
Click to expand...


It's not uncommon for negatives to use only
a portion of the total range on any given
channel. Chromes are more likely to use the
full range. (Or in a CCD scanner, to use
too much range.)

It's a wise move to reserve some space at
the ends of the range for post-processing.

For example, sharpening works by increasing
contrast. If your image used the full
range prior to sharpening, then the
sharpening step would push lots of pixels
over the edge (ie., into saturation.)

The shape of the histogram means almost
nothing. Some images will give you a nice
gaussian-looking curve, others will have
humps. Don't worry too much about that.



rafe b.
http://www.terrapinphoto.com
 
Thank you. But ... I'm not scanning negs, I am scanning slides (Kodak
EB, and E100VS).

You explained that only use a portion of the whole range is ok, but you
did not explain the shape. As I said, the curve, beside shows only
using a portion of the total range, it also shows a strange use of the
portion. The shape is not like a ^, it looks like a \.

I like to learn more from you.

-
narke
 
Thank you. But ... I'm not scanning negs, I am scanning slides (Kodak
EB, and E100VS).

You explained that only use a portion of the whole range is ok, but you
did not explain the shape. As I said, the curve, beside shows only
using a portion of the total range, it also shows a strange use of the
portion. The shape is not like a ^, it looks like a \.

I like to learn more from you.
Click to expand...



The shape is *mostly* irrelevant but there
are a few things you do NOT want to see.

It doesn't really matter if the hump is on
the left, the right or in the middle. It
doesn't matter if there is one hump or if
there are three.

If possible, you should check the histograms
for each color channel independently, and
adjust them individually. It's difficult
to set up a scan properly using just a single
"luminosity" view of the histogram.

You don't want to see a large number
of pixels piled up against either the
left or right edge or wall of the
graph.

Ie., you DON'T want to see y >> 0 at x=0,
or y >> 0 at x=255. (">>" means "much
greater than.")

If you see y >> 0 at x=0, your scan
is underexposed. Try again.

If you see y >> 0 at x=255, your scan
is overxposed. Try again.

These are the two most serious mistakes
you can make in scanning. If you have
a very contrasty chrome you may find
that you cannot satisfy both conditions
above. In this case the density range
of the chrome exceeds the dynamic
range of the scanner, and there's no
simple solution. (Or maybe for another
lesson.)

Do you understand the terms of calculus
and derivatives?

If you think of the shape of the histogram
as a curve, then...

You want to see the derivative of the
curve near zero at both the left and
right edges of the graph (ie., at x=0
and x=255.) This may not always be
possible to achieve, so don't sweat it.

You also want to see the amplitude
very close to zero at x=0 and x=255.

(These two qualities are also
characteristic of a true gaussian
curve.)

You do not want to see significant
ranges of x having y=0. This would
indicate wasted input range. (Ie.,
empty "buckets" in the histogram.)



rafe b.
http://www.terrapinphoto.com
 
Rafe,

You gave a very professional and wonderful explain. Thank you. Only
one question left:
If possible, you should check the histograms
for each color channel independently, and
adjust them individually. It's difficult
to set up a scan properly using just a single
"luminosity" view of the histogram.
Click to expand...

How you do that without introduce color cast? Every time I do the
operation on individual color channal, I always get a heavy color
casted result.

-
narke
 
Rafe,

You gave a very professional and wonderful explain. Thank you. Only
one question left:


How you do that without introduce color cast? Every time I do the
operation on individual color channal, I always get a heavy color
casted result.
Click to expand...


Read up on setting white points and black points
and neutrals. I'll try to post some URLs later,
no time right now.


rafe b.
http://www.terrapinphoto.com
 
do you mean this: http://gimp-savvy.com/BOOK/node61.html

in this method, author suggest adjust level for individual channals and
use three gray points to correct color cast. because three points of
gray is not easy to locate in most image, i want to ask why adjusting
on individual channal is so import that someone here does not mind
taking more efforts to do color correction after?

-
narke
 
narke said:
Thank you. But ... I'm not scanning negs, I am scanning slides (Kodak
EB, and E100VS).

You explained that only use a portion of the whole range is ok, but you
did not explain the shape. As I said, the curve, beside shows only
using a portion of the total range, it also shows a strange use of the
portion. The shape is not like a ^, it looks like a \.

I like to learn more from you.

-
narke
Click to expand...

A good example of a full Histogram with all possible colours is at:

<http://www.brucelindbloom.com/index.html?RGB16Million.html>

This Histogram gives a mostly Bell shaped curve, though it helps to view
each Channel. Definitely, the image does not represent anything that could
actually be photographed.

If you think of the Histogram palette, or the Levels palette, like in
PhotoShop, then the dark points and light points can be matched to the
Zone system . . . sort of . . . . . .

The darkest end on the left would be Zone 0, while the lightest point
could represent Zone X. The mid point is about Zone V (note roman
numerals), giving you an idea of how to divide a Histogram into ten
regions.

A darker overall image will show the Histogram larger on the left side. A
lighter overall image, like a high key image, would show more of the
Histogram towards the right. To compare to an actual image, just decide if
it is somewhat light, somewhat dark, or very even in overall exposure.

Of course, I am greatly simplifying this explanation. Hopefully, you will
see the relationship when you change Levels, or Curves, and view the
differences on your computer monitor. There use to be a great book that
explained more of this, called "PhotoShop Artistry", though I am not sure
if a recent print is available.

Ciao!

Gordon Moat
A G Studio
<http://www.allgstudio.com>
 
Back
Top