Digital BW, The Print

* -----Original Message-----
* From: Mark Hahn [mailto:markhahn2000@yahoo.com] 
* Sent: Monday, September 29, 2003 11:07 PM
* To: DigitalBlackandWhiteThePrint@yahoogroups.com
* Subject: [Digital BW] Re: New Photoshop CS -- aka Photoshop 8
* 
* 
* Well, Levels *is* just a handy interface for a Curve Adjustment... 

Mark,

Yes but it is not obvious to most that when you change the gamma in Levels
you are changing the slope of the curve. If you do not change the gamma
setting in Levels the curve remains linear. You can set your B&W points in
Curves leaving a straight line but it is much more difficult to see what you
are doing since you do not have a histogram to see exactly where you are
clipping.

If you change the gamma at the same time as setting the B&W points or apply
a non-linear curve first, your 10 or 12 or 14-bit data that was mapped into
adjacent values in 16-bit space may be shifted within that confined area
causing pixels to be moved from their original value to an adjacent one. If
this happens there is a loss of information at one level and an addition of
pixels at another value or "combing."

If you set the B&W points first (and they can be set well outside any
clipping points) the low bit data is redistributed in 16-bit space. Then
when you apply a curve adjustment or a gamma change in Levels the pixels can
be reassigned to different values without pilling up on top of each other.

This is where some scanning software, Silverfast in my case, can really give
you a hard time if you use it to make on the fly adjustments.

My workflow has been to first adjust the raw 16-bit scan file with the
Levels function in Silverfast HDR setting the white point slider just beyond
the lightest pixel and the black point slider exactly at the lowest pixel.
(I was probably unclear in my earlier post and gave the impression that I
was using the eyedropper tools in Levels.) My second step is to use Curves
in Photoshop to shape the overall contrast. My only point was that doing
this process in two separate steps and in this order maybe kinder to the
image data than doing them both at once or in the reverse order.

You can accomplish the same thing with two separate curve adjustments. The
idea is just to make sure the data is spread out before adjusting the slope.
Of course if the software is intelligent enough to do things in the proper
sequence then you don't have to worry about it. PS seems pretty okay in this
regard but after my experiences I just play it safe.

* to 
* really start filling things in I think you have to introduce 
* something that does a little averaging... curves just scale your 
* values as-is... so every 159 level gray, for instance, is going to be 
* multiplied by the same value via the curve, but in a convolving 
* filter it will depend on the adjoining pixels and thus introduce 
* variation and kind of fill things in.

Yes but if the result of the multiplication is already occupied by other
pixels you are losing the information that separates one tone from another.
Better to give your data some breathing room before you start pushing it
around.
* 
* Actually, I don't always like setting my B&W points first in PS 
* because it clips your data right there in that one step and then if 
* you do more out of range adjustments you get even more clipping... 
* too bad PS can't maintain the out of range data...  If I have to, I 
* will usually expand the data a bit, but not right to true b&w pts, do 
* other stuff and then do a final b&w pt setting (if I haven't done it 
* already using curve adjustments) as a last step.

At this initial stage I always set my B&W points outside the min and max
values shown on in the Levels histogram so that there is no clipping.
* 
* 64 bits per channel is way over-kill, but it is just a default in my 
* program which could be changed... the real thing is that my own 
* program is more a learning tool for me than anything else.  When I 
* got started in digital image editing with PS 3 years ago it was all 
* too baffling to me and the tripe that is written in most PS for 
* Photographers-like books just shows you what buttons and sliders to 
* use.  I never felt I got a good enough grasp of what I was actually 
* doing from just the smoke and mirrors explainations in these books to 
* do it intelligently... but I found that after I went through the 
* effort to write and de-bug my own code, for something like an unsharp 
* mask routines etc. that I then *knew* pixel for pixel exactly what I 
* was doing and why, even when it was hidden "under the hood" in PS... 
* just my mentality and my method... probably not shared by many others 
* on the list:)

I agree (especially about the books and I have a stack to send out for
recycling!). Trying to understand what is happening behind the curtain after
seeing scan files "combed" lead me to this sequence of steps. It often is
not clear what PS or your scanning software may be doing to your data. For
instance if your scanner software mapped the 14-bit scanner data of 16382
shades of gray into the values 10001 through 26382 within the 65536 shades
of gray in 16-bit space you still only have 16382 shades of gray even if
they are in a larger working space. (This is why raw scans are so bunched up
when you look at the histogram.) If you change the slope of the curve of
that data you will move pixels from one value to another within that narrow
range.

If instead you first re-map the data so that the 0 point of the 14-bit data
falls on say 500 and the tonal value of 16381 on 65000 there are now empty
values between the 16382 shades of gray recorded. At this point if you apply
a curve type function and shift pixels from their recorded value to a
calculated one roughly 75% of the moved pixels will fall on unused values in
16-bit space increasing the number of tonal values in the file rather than
reducing them.

This less important if we can remain in 16-bit space but I hate to loose any
data. When I need to drop down to 8-bit to use layers it has been very
critical to maintain all the data I could at each step of the process 

Martin Wesley
http://www.carolyn.cc/Guests/MartinWesley/pages/MW_01.html
http://www.borderless-photos.de/guests.html