Digital BW, The Print

Hi Ken,

This works well�I have done something similar for scanners within my Curve Calculator II software, which determines curves for digital negatives, but can also linearize a printer, but it uses the Stouffer 21 step reflection step tablet. �When I compared this method to a reflection densitometer, the values are surprisingly accurate, with the maximum variance at about Log .01, or 1/30 of a stop. �I even had one student that did not have a scanner, so we tried his spot meter and got fairly good results with that for making curves for digital negatives.

A couple of suggestions:

1. �Instead of applying Gaussian Blur to the steps before taking measurements, try the Blur Average and clean up any dust first.

2. �Be sure to set black and white points based on the steps of the Stouffers.

3. �If you are trying to determine the reflection densities of the output based on the Stouffer, it is only valid if the density range of the output falls within the range of the Stouffer, which goes up to Log 2.0.


--
Best Wishes,

Mark Nelson
Precision Digital Negatives
PDN Print Forum @ Yahoo! Groups
Mark Nelson Photography

On Jan 14, 2009, at 3:12:41 PM, kwalsh74 <kwalsh74@yahoo.com> wrote:

I just wanted to post a note that some might find helpful. I don't
own a densitometer and wanted to linearize some profiles. I saw
Paul's write up on using a scanner, but unfortunately my scanner is a
junky all-in-one deal with no real controls and thus the highlights
are always blown making it useless. I did however get surprisingly
good results using my camera using the method I'll describe here.

1. I got a calibrated 12-step reflection wedge from Stouffer ($24
shipped, uncalibrated are about $10 less) but any sort of reference
with a few steps will do.

2. Place the 12-step reference right next to the 21-step profile to be
linearized on the floor. Make sure the two are as close as practical
and oriented the same way (white to white, black to black).

3. Put camera (a Rebel XTi in this case) on tripod directly above the
step sheet. Use as long a focal length as possible and get the camera
as high as possible, this will reduce the possibility of picking up
direct reflection of the light sources from the paper under measurement.

4. Place two light sources as far as practical to each side
illuminating at a 30 to 45 degree angle. Try to make sure the sources
are the same brightness and are the same distance from the step sheet.
Make sure the angle of illumination won't produce specular reflection
at the camera. I used off camera flashes, but desk lamps would do
just as well with longer exposure. The goal is to get the
illumination of the paper as even as possible and to not have any
direct reflection from the camera's viewpoint.

5. Take photos at a few different exposures, my XTi has enough dynamic
range that I could get reliable measurements using just two exposures
when shooting RAW but you may want to take more, especially if using
JPEG. The goal is to make sure among your exposures there are some in
which the darkest patches aren't too close to black limit and some in
which the lightest patches aren't too close to saturation. My first
try I did four exposures and it was over kill.

6. Load photos into Photoshop/Gimp or whatever, ideally 16-bit data.�
Convert to Lab color. Apply some Gaussian blur. Measure the L values
of each reference step. Measure the L values of each of the 21 paper
under test steps. You may not be able to get all measurements from
the same exposure, you might have to do mostly highlights with one
exposure and mostly shadows with another.

7. Using the known densities from your reference scale and its
measured L values you can then calculate the densities for your 21
step test using cubic spline interpolation. You can get free cubic
spline tool sets for Excel or you can even do it online here:

http://www.akiti.ca/CubicSpline.html

Obviously you'll need to do different fits for each exposure.

8. That's it, the results from your interpolation will tell you the
densities you need to plug into QT RIP to linearize.

Some notes:

- Getting even illumination of the paper is difficult, that's why
densitometers and scanners are a better choice. But with proper light
placement and aligning the reference steps and test steps close to
each other and in the same orientation you minimize the effect. Also,
you can check by seeing how the paper white varies across the image
and can even use the measurement of this variation to correct for any
error but I believe that to be very much overkill for doing a
linearization. Also you could move the reference wedge to be closest
to the portion of the 21-step that you are measuring with a given
exposure.

- If shooting RAW resist the urge to output linear data from the RAW
converter. It seems like a good idea, but since we're interpolating
to a log scale (density) it actually makes the measurement and
interpolation more fussy and less reliable. Using a standard tone
curve and 2.2 gamma space gets things into a power law that jives with
density interpolation better.

- Even if you have 8-bit data, going into 16-bit mode before doing the
Gaussian blur will gain you more precision.

- Note that (at least in CS3) the eyedropper can actually display
16-bit values. In the info panel where the values are displayed note
that there is a tiny tiny arrow next to the little eyedropper which if
you click you can choose 16-bit display values.

- In the middle portions of the exposure the results were surprisingly
consistent, density interpolations varied by less than 1% from
exposures differing by 1eV. Nonetheless, as stated above you can have
a systematic error if the illumination is not even enough.

- One might be able to use Photoshop's "Merge to HDR" feature to
combine all your exposures into a single one for measurement. I
haven't tried this.

- Given it appears you can get a pretty amazing Epson scanner (16-bit
per channel, LED illumination, 3.2 Dmax, built in transparency
adapter, 4800 optical DPI) for just $100 setting up the camera and
doing multiple exposures may be more trouble than it is worth. That
said, using the reference wedge and doing cubic spline interpolation
to get accurate results would still apply with scanned input.

Well, hopefully this helpful to someone someday. At least it is here
in the archives now...

Ken


[Non-text portions of this message have been removed]