Color balancing using Analog Gain

  • Thread starter Thread starter Don
  • Start date Start date
D

Don

How do I color balance using Analog Gain (AG)?

What I mean is this:

Given an area of color, say R=12, G=23, B=45, what AG setting I have
to apply to get, say, R=56, G=78, B=23? (These are totally arbitrary
numbers just as an example.)

I can come up with values by experimentation and trial and error, but
that's both too time consuming and inexact.

AG ranges (in my case) between -2 and +2 while an individual color (in
my case) is 8-bit so it ranges between 0 and 255. Is this a linear
conversion? Again, I can experiment, but why re-invent the wheel...

So, what is the conversion formula?

Finally, any hidden gotchas with this approach?

Thanks as always!

Don.
 
How do I color balance using Analog Gain (AG)?

What I mean is this:

Given an area of color, say R=12, G=23, B=45, what AG setting I have
to apply to get, say, R=56, G=78, B=23? (These are totally arbitrary
numbers just as an example.)

I can come up with values by experimentation and trial and error, but
that's both too time consuming and inexact.

AG ranges (in my case) between -2 and +2 while an individual color (in
my case) is 8-bit so it ranges between 0 and 255. Is this a linear
conversion? Again, I can experiment, but why re-invent the wheel...

So, what is the conversion formula?

Finally, any hidden gotchas with this approach?

Thanks as always!



I use the AG controls in NikonScan, while scanning
negatives as positives, but not at all with the approach
you're suggesting.

What I'm looking for in my case is to have the three
histograms (for the three color channels) as wide
as possible, right-justified, and not clipped. By
right-justified, I mean that I want "high" values in
general, but with the very lightest tones comfortably
below 255. Eg., I'd prefer a histogram that ranges
from 120 to 240 over one that ranges from 20 to 140.

(When you scan negs as positive, these sorts of
numbers are not uncommon.)

I do that by trial and error, though I don't take my
experiments beyond 1 or 2 decimal places. Eg,
when I scan negatives as positives in NikonScan,
the following works 90% of the time:


Overall: +0.3
Red: -0.4
Green +0.8
Blue: +1.5


Hope that helps...


rafe b.
http://www.terrapinphoto.com
 
What I'm looking for in my case is to have the three
histograms (for the three color channels) as wide
as possible, right-justified, and not clipped. By
right-justified, I mean that I want "high" values in
general, but with the very lightest tones comfortably
below 255. Eg., I'd prefer a histogram that ranges
from 120 to 240 over one that ranges from 20 to 140.

Although I do this with slides, the real reason is similar. Since
histogram settings in NikonScan are applied *post*-scan I want to,
both, extend the dynamic range and, while I'm at it, color balance at
the earliest possible stage - and that means use Analog Gain. In that
way I start post-processing with the best image the hardware can
deliver.
(When you scan negs as positive, these sorts of
numbers are not uncommon.)

I do that by trial and error, though I don't take my
experiments beyond 1 or 2 decimal places. Eg,
when I scan negatives as positives in NikonScan,
the following works 90% of the time:


Overall: +0.3
Red: -0.4
Green +0.8
Blue: +1.5

I'm wrestling with the bane of Nikon scanners (fx: echo on):
Kodachrome...

That means my values are *very* extreme. Currently, using the test
slide I'm experimenting with, I have to boost AG by around 3.5 EV
(Master at +2, individual channels all at 1.5!!!) to get the blue
channel to 238 while red and green are still loitering around 120!
Hope that helps...

Thanks very much!

Don.
 
I'm wrestling with the bane of Nikon scanners (fx: echo on):
Kodachrome...

That means my values are *very* extreme. Currently, using the test
slide I'm experimenting with, I have to boost AG by around 3.5 EV
(Master at +2, individual channels all at 1.5!!!) to get the blue
channel to 238 while red and green are still loitering around 120!


I think you're on the right track, in any case.

As long as none of the individual histograms are
clipped, you should end up with good scans.
(By clipped, I mean smashed against either the
left or right wall.)

The scanner is working internally with lots of bits,
(bit depth) so even after processing you should
get smooth comb-free histograms.

The one thing maybe left in your workflow is to
set the black and white points individually for
each channel, followed (optionally) by gamma
adjustments, per channel, as required. (I'm
talking about tweaks in NikonScan here...)

The only time this won't work is when the film
itself is too thin or too dense for the scanner to
register. This doesn't happen often for negatives
but I suspect is more likely to happen on chromes.

If memory isn't an issue, bring the images into
Photoshop with 16 bits per channel, at least until
you can routinely get near-perfect scans from
the scanner driver. The extra bit depth will give
you the headroom you need for further tweaks
in Photoshop if necessary.


rafe b.
http://www.terrapinphoto.com
 
| How do I color balance using Analog Gain (AG)?
|
| What I mean is this:
|
| Given an area of color, say R=12, G=23, B=45, what AG setting I have
| to apply to get, say, R=56, G=78, B=23? (These are totally arbitrary
| numbers just as an example.)
|
| I can come up with values by experimentation and trial and error, but
| that's both too time consuming and inexact.
|
| AG ranges (in my case) between -2 and +2 while an individual color (in
| my case) is 8-bit so it ranges between 0 and 255. Is this a linear
| conversion? Again, I can experiment, but why re-invent the wheel...
|
| So, what is the conversion formula?
|
| Finally, any hidden gotchas with this approach?
|
| Thanks as always!
|
| Don.


Why use "Analog Gain" to do color balancing? In the Curves section of the
Tool Palette, NikonScan provides a control for "Auto Levels" (the button
looks like a half moon). Based on my experience, selecting "Auto Levels"
increases contrast and corrects the color well enough that it only takes a
little tweak (if any) to get the colors adjusted right. I think you can
somewhat customize the effect of this control by varying the "% Excluded"
(set in the Preferences-Advanced Color section controls) on the white and
black ends of the curve.

Jean
 
Jean <[email protected]> said:
Why use "Analog Gain" to do color balancing?

To make optimum use of the scanner's analogue to digital convertor's
dynamic range.
In the Curves section of the
Tool Palette, NikonScan provides a control for "Auto Levels" (the button
looks like a half moon).

But that is adjusting the data after it has been captured by the
scanner. Analogue Gain is a slight misnomer because it doesn't adjust
the gain at all, it adjusts the CCD exposure. Either way, this
adjustment is implemented before the signal is digitised. Since Don is
struggling with an LS-30, which only has 10-bits per channel, that is
important.
Based on my experience, selecting "Auto Levels"
increases contrast and corrects the color well enough that it only takes a
little tweak (if any) to get the colors adjusted right. I think you can
somewhat customize the effect of this control by varying the "% Excluded"
(set in the Preferences-Advanced Color section controls) on the white and
black ends of the curve.
But it is important to have a full histogram to begin with, otherwise
those adjustments can result is missing codes - especially with an older
scanner with only a few hi bits on the ADC.
 
On Wed, 11 Feb 2004 11:31:14 GMT, (e-mail address removed) (Don) wrote:

The scanner is working internally with lots of bits,
(bit depth) so even after processing you should
get smooth comb-free histograms.

I'm have an ancient LS30 here so there are only 10 bits per channel
which is not good. I also fear it may be starting to fail as I noticed
some weird things in the histograms (see message "Histogram
Puzzle...") but that's another story...
The one thing maybe left in your workflow is to
set the black and white points individually for
each channel, followed (optionally) by gamma
adjustments, per channel, as required. (I'm
talking about tweaks in NikonScan here...)

The problem with any adjustments in NikonScan (other than AG) is that
they are all applied post-scan. (I might just as well do them in
Photoshop.) So I'm trying to do the most I can at the scanning stage.
Normally, any minor color adjustments would indeed be done with black
and white points, but since I'm getting such extreme data from the
scanner, I'm trying to do some rough color balance at the scanning
stage.
The only time this won't work is when the film
itself is too thin or too dense for the scanner to
register. This doesn't happen often for negatives
but I suspect is more likely to happen on chromes.

Bingo!

Thanks for all your help, Rafe!

Don.
 
| How do I color balance using Analog Gain (AG)?
|
| What I mean is this:
|
| Given an area of color, say R=12, G=23, B=45, what AG setting I have
| to apply to get, say, R=56, G=78, B=23? (These are totally arbitrary
| numbers just as an example.)
|
| I can come up with values by experimentation and trial and error, but
| that's both too time consuming and inexact.
|
| AG ranges (in my case) between -2 and +2 while an individual color (in
| my case) is 8-bit so it ranges between 0 and 255. Is this a linear
| conversion? Again, I can experiment, but why re-invent the wheel...
|
| So, what is the conversion formula?
|
| Finally, any hidden gotchas with this approach?
|
| Thanks as always!
|
| Don.


Why use "Analog Gain" to do color balancing? In the Curves section of the
Tool Palette, NikonScan provides a control for "Auto Levels" (the button
looks like a half moon). Based on my experience, selecting "Auto Levels"
increases contrast and corrects the color well enough that it only takes a
little tweak (if any) to get the colors adjusted right. I think you can
somewhat customize the effect of this control by varying the "% Excluded"
(set in the Preferences-Advanced Color section controls) on the white and
black ends of the curve.

Jean

The problem is that any NikonScan adjustments (other than Analog Gain)
are done post-scan. In my case the data is so extreme (LS30's infamous
inability to handle Kodachrome) that leaving color balancing for
post-processing results in very pronounced artifacts. There just isn't
enough dynamic range. I mean, a neutral scan histogram stops at below
120!! So, I'm forced to do some rough color balancing with Analog Gain
to at least get a little more dynamic range before doing further
adjustments in Photoshop.

Don.
 
I'm have an ancient LS30 here so there are only 10 bits per channel
which is not good. I also fear it may be starting to fail as I noticed
some weird things in the histograms (see message "Histogram
Puzzle...") but that's another story...


It's getting 10 bits from the A/D, but I suspect it's
using 16 bit math (or more) to do its post processing.


rafe b.
http://www.terrapinphoto.com
 
Back
Top