DEE Plugin

  • Thread starter Thread starter Robert
  • Start date Start date
R

Robert

Starting a large slide scanning program, I made very impressive
expieriences with Nikons DEE technique. DEE (Dynamic Exposure Extender)
reproduces in a magic way the last details out of shadows and lights.
In a certain way it is similar to the contrast masking method, but
compared to that, it creates never disturbing halos at contrast borders.

The only disadvantage is, that it nearly doubles the scanning time. I
would prefer to run it as a batch after the scanning. NikonScan is not
able to run batches on files. Because DEE it is a pure software
technique, it should be a possible to be applied as a Plugin for
PaintShop or Photoshop. Does anybody know such a solution?

Regards
Robert
 
Robert said:
Starting a large slide scanning program, I made very impressive
expieriences with Nikons DEE technique. DEE (Dynamic Exposure Extender)
reproduces in a magic way the last details out of shadows and lights.
In a certain way it is similar to the contrast masking method, but
compared to that, it creates never disturbing halos at contrast borders.

The only disadvantage is, that it nearly doubles the scanning time. I
would prefer to run it as a batch after the scanning. NikonScan is not
able to run batches on files. Because DEE it is a pure software
technique, it should be a possible to be applied as a Plugin for
PaintShop or Photoshop. Does anybody know such a solution?

I don't know about DEE, but there are ways to prevent halos in contrast
masking. Certain CM techniques seem to insist on blurring the selection.
This is where the halos come in. You could try a CM technique that does
not include blurring, or just skip the blurring. In some ceases, even
sharpening the selection can be applied if halos persist to show up.
 
Yes, I feel the same way about GEM. It should be able to be applied
later but there is no way to do that. Regarding DEE, it is similar
to Image->Adjust->Shadow/Highlight in Photoshop CS. Have you tried that?

W
 
Robert said:
Meanwhile I have made comparison test between DEE and Contrast Masking:

http://home.t-online.de/home/jm-net/cm.html

I see what you mean. Perhaps you could try this Photoshop technique,
which is similar to contrast masking, but which has less risk of halos:

1. Set the foreground color to black
2. Choose Select > Color range and set the slider to the maximum value
(or slightly lower - this determines how much of the shadows you want to
brighten up) and press OK
3. Press Ctrl-H to get rid of the marching ants
4. Select Adjust > Curves. Select a point towards the left (Dark) end of
the curve and drag it upwards (i.e., increase its RGB value). Use the
preview facility to check out the effect. If you like the amount of
increased shadow detail, select OK.
5. Deselect.

To reduce the risk of halos even more, you could insert the following
steps between 2 and 3:
2a: Enter the mask mode
2b: Perform usnharp masking on the mask to mark the crisp edges in the
selection (you will have to experiment with the sharpening settings)
2c: exit mask mode

Once you have found settings that work well, you can easily automate
this as an action, and run images through it in batch.
 
1. Set the foreground color to black
2. Choose Select > Color range and set the slider to the maximum value
(or slightly lower - this determines how much of the shadows you want to
brighten up) and press OK
3. Press Ctrl-H to get rid of the marching ants
4. Select Adjust > Curves. Select a point towards the left (Dark) end of
the curve and drag it upwards (i.e., increase its RGB value). Use the
preview facility to check out the effect. If you like the amount of
increased shadow detail, select OK.
5. Deselect.

Thank you for these hints. They are working well. Nevertheless I have
the impression, that DEE and Photoshop-CS "Adjust-Shadows and Lights"
are bringing in a magic way more details out of the shadows. I assume,
they are using mathematical algorithms to produce their results.
To reduce the risk of halos even more, you could insert the following
steps between 2 and 3:
2a: Enter the mask mode
2b: Perform usnharp masking on the mask to mark the crisp edges in the
selection (you will have to experiment with the sharpening settings)
2c: exit mask mode

With your second proposal I was not successful. I assume, it is not
possible to apply USM on a mask, which was blurred with the Gauss
filter. For this reason I looked in the past for blurring filters, that
do only blur within the given contrast borders. Neanwhile I was
convinced, that this cannot work. Because each blurring filter is using
for calculating a new pixel all the pixels around the old one, contrast
boarders will be always ignored from blurring filters.

Regards
Robert
 
Robert said:
Thank you for these hints. They are working well. Nevertheless I have
the impression, that DEE and Photoshop-CS "Adjust-Shadows and Lights"
are bringing in a magic way more details out of the shadows. I assume,
they are using mathematical algorithms to produce their results.

Don't know about DEE, but shadow-highlight adjustment in PS CS gives an
additional boost to shadow detail by setting the darkest point to an RGB
value much higher than 0,0,0. In other words, after applying the
adjustment, there is no more solid black in the image. You can check it
in the histogram - IMO it's a way of cheating. You could do the same by
performing a levels adjustment, where you move the black slider on the
grayscale bar *below* the histogram in the dialog box towards the right.
And, BTW, any operation in PS is based on mathematical algorithms.
BTW (2): it is more or less essential that you perform the
abovementioned steps on a 16-bit image.
With your second proposal I was not successful. I assume, it is not
possible to apply USM on a mask,

This is certainly possible. What I meant by 'mask mode' is officially
known as 'quick mask'. Perhaps that caused some confusion.
which was blurred with the Gauss
filter.

Why did you blur the mask in the first place? If you apply USM to
something that has been blurred, the effect is typically invisible.
The essence of this method is *not* to blur the selection ...
 
BTW (2): it is more or less essential that you perform the
abovementioned steps on a 16-bit image.

This would limit the application of the contrast masking method on
Photoshop CS, because this is to my knowledge the only imaging software
with 16 bit AND layer support.

I have read a lot of discussions about the 16 bit advantages and there
are examples, which show the "sliced" form of a 8 bit picture
histograms, which were stressed with postprocessing steps. But up to now
I have not yet seen a picture comparision, were a clearly visible
advantage can be seen.

Perhaps it is, because my software is Paintshop and not Photoshop, that
I have not completely undersood your explanations yet. What is the sense
if the USM step exactely?

In my procedure, the mask is an inverted black&white copy of the picture
itself. I have to blure this mask in anyway, because the picture would
lose otherwise to much details - and there the mask can not be
processed by USM. Is this also your understanding of a contrast mask?

In the case of batch processing, I apply the contrast masking not to
selections, but to the whole picture. Your step 2, does it refer to the
pichture itself or to the mask?

My procedure is the following:
- Duplication of pictire into two layers
- Lower layer is inverted , desaturated and blurred
with Gaussian Blur Filter (40 pixel)
- Upper layer blended with "hard light"
- fine tuning of both layers with "Curves"

I am a little bit confused.

Regards
Joachim
 
Robert said:
This would limit the application of the contrast masking method on
Photoshop CS, because this is to my knowledge the only imaging
software with 16 bit AND layer support.

That may be possible, although the procedure may also work in PS 7,
Corel PhotoPaint an future versions of PSP and The Gimp.
In my procedure, layers aren't needed anyway.
I have read a lot of discussions about the 16 bit advantages and
there are examples, which show the "sliced" form of a 8 bit picture
histograms, which were stressed with postprocessing steps. But up to
now I have not yet seen a picture comparision, were a clearly visible
advantage can be seen.

You're welcome to try it in 8-bit mode. I have seen unwanted effects,
especially when applying curves with steep inclinations.
Perhaps it is, because my software is Paintshop and not Photoshop,
that I have not completely undersood your explanations yet. What is
the sense if the USM step exactely?

If you don't see halos you can forget about the extra steps anyway.

In my procedure, the mask is an inverted black&white copy of the
picture itself.


The quick mask is a channel that displays your selection, not the whole
image (although it may look like it) in grayscale. Sometimes, the
selection created with 'select color range' results in slightly fuzzy
edges (perhaps an aliasing effect) that may lead to halos when
performing the curves adjustment. By applying the unsharp mask with the
correct settings, you can create sharp transitions where there are meant
to be sharp transitions.
I have to blure this mask in anyway, because the picture would
lose otherwise to much details

By blurring, you can never add or preserve details
- and there the mask can not be
processed by USM.

In Photoshop (not only CS) this is possible. Don't know about PSP.
Is this also your understanding of a contrast mask?

I think it's a different implementation of the same thing.
In the case of batch processing, I apply the contrast masking not to
selections, but to the whole picture. Your step 2, does it refer to
the pichture itself or to the mask?

My procedure is the following: - Duplication of pictire into two
layers - Lower layer is inverted , desaturated and blurred with
Gaussian Blur Filter (40 pixel) - Upper layer blended with "hard
light" - fine tuning of both layers with "Curves"

It seems you are combining my procedure with the standard procedure for
contrast masking. That was not what I meant. My method does not involve
layers, blurring and layer blending modes. I haven't tried the standard
procedure as it is known to produce halos. Anyway, I don't know the PSP
equivalents of my steps 1-5. The purpose of step 2 is to make a
selection that covers the blackest black (=selected foreground color) in
the image completely, the not-so-black blacks partially and does not
cover the midtones nor the highlights. Thus, the deepest shadows are the
most affected by the curves, the not-so-deep-shadows are less affected
and the non-shadows are not affected.
 
That may be possible, although the procedure may also work in PS 7,
Corel PhotoPaint an future versions of PSP and The Gimp.
In my procedure, layers aren't needed anyway.

No I have understood your procedure and I applied it sucessfully. There
is a little bit work to do to improve the selection. For example mirror
pictures in window glas should not be seperated by selection, because
they will appear after the curve application like artefacts.

Another way does not work: I tried at a friend to apply your method with
PS7 on 16 bit files. It is not possible, because the selection tool
allows in the 16 bit mode only selection of the complete pichture. I
assume, this has the same technical reason like the lack of layer and
mask features in the 16 bit mode.

Best regards
Robert
 
Robert said:
No I have understood your procedure and I applied it sucessfully. There
is a little bit work to do to improve the selection. For example mirror
pictures in window glas should not be seperated by selection, because
they will appear after the curve application like artefacts.

Perhaps PS CS and PSP treat things a bit differently, then. I haven't
had such problems.
Another way does not work: I tried at a friend to apply your method with
PS7 on 16 bit files. It is not possible, because the selection tool
allows in the 16 bit mode only selection of the complete pichture. I
assume, this has the same technical reason like the lack of layer and
mask features in the 16 bit mode.

You must be right - with every version of PS, more functions become
available in 16-bit, but I can't exactly remember which functionality
became available with which version.
 
It's on a computer - it has to be mathematical algorithms ;-)

Don't know about DEE, but shadow-highlight adjustment in PS CS gives an
additional boost to shadow detail by setting the darkest point to an RGB
value much higher than 0,0,0. In other words, after applying the
adjustment, there is no more solid black in the image.

That is not correct.
There is a setting for the desired black point (and white point) in the
adjustment.
Plus the adjustment works on regions, not individual pixels -- so the
results will vary quite a bit depending on the surrounding image
values.

You can check it
in the histogram - IMO it's a way of cheating. You could do the same by
performing a levels adjustment, where you move the black slider on the
grayscale bar *below* the histogram in the dialog box towards the right.

No, you cannot.
Shadow and Highlight is closer to a contrast masking technique (with
lots of improvements).
You cannot reproduce the effect with levels or curves.

Chris
 
[/QUOTE]
It's on a computer - it has to be mathematical algorithms ;-)

Chris,

Please take care who you are quoting. It's not me who said this. My
comment on this claim was similar to yours;-)
That is not correct.
There is a setting for the desired black point (and white point) in the
adjustment.
Plus the adjustment works on regions, not individual pixels -- so the
results will vary quite a bit depending on the surrounding image
values.
No, you cannot.
Shadow and Highlight is closer to a contrast masking technique (with
lots of improvements).

I did not claim that S/H simply boosts the highlights the same way as
dragging the black slider of the 'output levels' to the right. But
typically, the result has the same symptom: the 'starting point' of the
histo shifts to the right after S/H adjustment. The amount of shift
seems unrelated to the 'black clip' setting in the S/H dialog box: the
shift is usually much more than that.
You cannot reproduce the effect with levels or curves.

Of course, you can't reproduce it exactly. But by applying curves to a
'smart' selection of the shadow areas (or the highlights or midtones)
you can achieve similar effects that don't take so much computing power.
On my computer it takes ages to get an S/H preview of a 16-it scan from
my Minolta DSE 5400 and it takes even more time to perform the actual
S/H adjustment.
 
Back
Top