Lunch with Henry Wilhelm

Recently I had the opportunity to have lunch with Henry Wilhelm, founder of Wilhelm Imaging Research (WIR) and author of the classic book on image permanence.
Recently I had the opportunity to have lunch with Henry Wilhelm,
founder of Wilhelm Imaging Research (WIR) and author of the classic
book “The Permanence and Care of Color Photographs: Traditional and
Digital Color Prints, Color Negatives, Slides, and Motion Pictures”.

I had spoken to Henry previously, at a Hewlett Packard event over in
Singapore (see a separate story) but this was a great opportunity to
talk to him one-on-one for some two and a half to three hours. Henry
strikes you, up front, as an on the level sort of guy. He speaks well
and can be hard to shut up when you get him going on something he is
passionate about. Clearly honest and open testing and rating of image
permanence is one of these.

Many have criticized his work because the WIR is funded from the
companies whose products he tests. In answer to this Henry is clear
that WIR develops a standard test and that companies that contract them
to do testing have NO say over the testing method. In fact WIR’s
contracts with these companies make it clear that WIR can publish the
permanence performance of any product that is in the marketplace. This
last part is important as WIR does get called in by companies to help
in the development stages. Also WIR funds some testing themselves,
either because a key manufacturer prefers to do their own less rigorous
testing or because WIR deem a particular product category to be very
important. A recent example of this is their testing of 4 x 6 inch
photo printer display permanence (both digital and photographic). See 4×6 test.

Then we got into issues of their testing procedures. I had asked Henry
in Singapore about how he calibrates accelerated tests to real time.
Here we got a chance to get into the question in more detail. All WIR
(and other people’s) permanence testing is done using an accelerated
testing procedure. It is obvious that we can’t wait 100 years to find
out how long a print will last. WIR has specified a viewing environment
in terms of light levels, etc. They test using much stronger
illumination to have any fading visible in a shorter time. The issue is
one of reciprocity failure, something all photographers should know.
Reciprocity failure touches on one of the fundamental photographic
principles. In photography an exposure is determined by two things:
light level (aperture) and shutter speed (or duration of exposure).
They are supposed to be linked, so that if you double the light level
and half the exposure time (or visa versa) you get the same exposure.
For conventional photographic film this holds true across a wide range
of light levels and shutter speeds. However it is known to break down
at both very short and very long exposures. People doing
astrophotography (how I got into photography) and high speed flash
photography see this as a need to compensate with more exposure than
you would expect.

WIR address reciprocity failure and other issues by using the lowest
possible accelerated light intensity (lower than anyone else uses).
This means their tests take longer. They have also gone to great
lengths to calibrate their tests. They have been lucky enough to find
well documented examples of photographs displayed in known
circumstances for known periods of time and then have tested the same
materials under WIR’s testing procedure. This has allowed a good degree
of confidence to be built up at WIR that their accelerated years
ratings are close to the real number of years that could be expected.
However they are continuing to work on refining this. I also asked why
they used a years figure rather than a numeric or star rating system.
Henry feels such a system is just too complex for the average consumer
and also that, as materials get better, it becomes harder to show
differences in materials that are getting closer together in terms of

There is also some controversy about the light source used in these
accelerated tests. WIR uses fluorescent tubes. Henry acknowledges that
there are some issues with fluorescents, but not because of the
emission peak quality of their light emission. Rather it is because
they have significantly different spectral characteristics at the blue
end of the spectrum. WIR still considers them best though because of
the heat issue. Alternatives, like Xenon, create a heat problem. Heat
lowers moisture  which has a major effect on image longevity. WIR
is testing Xenon but have not been able to adequately solve the heat
issue, plus Xenon has higher UV than they would like. They are also
testing other daylight simulants.

WIR not only test inkjet printing methods but also conventional
photographic ones, though they limit themselves to four color silver
halide processes. They do not test film now as Henry observes that it
is on the way out as a mainstream capture method. Besides, he also
finds inkjet the most interesting area at present. Fuji’s Crystal
Archive paper is currently the highest rating silver halide color paper
at a 40 year display life. I asked Henry if the printing method
affected display life, as in whether there was a difference for prints
done on that paper in a Noritsu or Frontier digital photo printer. In
Henry’s tests there are no differences in longevity caused by the
printer chosen. The quality of processing can obviously have an impact.

WIR is doing some testing of third-party papers and inks. One problem
is that the third party people will typically not pay for testing as
they have no real interest in accurate longevity tests. So the testing
that they are doing they are funding themselves. The major problem with
doing any of this in a systematic way are the number of permutations
when you add third party papers, say, with even the variations in ink
from one manufacturer across their printer models. Henry has major
concerns about much of the third party product. For example, a new
brand of paper is on sale in the US under the National Geographic
brand. The packaging claims 100 year life for the prints ‘on any
printer’. Henry said that he knows that some of the printer models they
then list inside the packaging cannot achieve 100 years even on the
printer manufacturer’s best papers under WIR’s testing procedures, so
he remains concerned about such claims. The issue is that anyone can
come up with a testing criteria (and Henry is very critical of Kodak
for exactly this) but unless it is a reasonable one then the claims may
not bear much resemblance to what people will actually get.

A note here. Henry is also constantly evaluating their own test
criteria. WIR use a standardized light level that is meant to be
representative of typical display conditions. However, he notes that
light levels in houses have been steadily increasing due to the much
heavier use of glass and skylights letting more natural light in. This
may mean that at some point they will have to raise their standard
light level.

We then talked about artist’s materials and the interrelationship with
digital prints. He said that they had looked at the paint permanence
rating systems used by artist paint manufacturers to see if they could
cross-calibrate the two systems, but they found that the artist’s one
is quite loosely constrained because it is a
fairly old standard and
thus they couldn’t get them to cross relate. What he could say was that
in lots of old prints that have been retouched it is usually the
retouching material that has lasted better than the photographic color
print, suggesting that the artist’s materials are reasonabl
y ok. I
raised the issue of reactions occurring if one overworked a digital
print with artist’s material. Henry had seen nothing to suggest that
this occurred. Again he pointed out that the number of permutations
made systematic testing impossible. I also asked about framing
materials and buffered vs. non-buffered matt boards, for example. So
far he has seen nothing to suggest any issues with it and felt that
buffered matt boards were probably an advantage for long term
atmospheric exposure issues (which still happen in a framed print).

WIR is starting the testing of water-based coatings and also lamination
products, especially on canvas. They are limiting it to water-based
coatings mainly, because of a belief these are better for the
environment and the users, as long as they work well. Such coatings are
important not only from a surface protection perspective but also for
light filtering and protection from environmental gases, which can have
a major impact on unprotected prints.

Henry is involved with other people interested in true international standards covering inkjet prints and other mediums.

Henry made a particularly interesting point towards the end of our
lunch. Because there is often a difference in longevity from colour to
colour within an ink set, with the blacks often being the longest
lasting, he observed that the driver version being used, and the driver
settings, can have a major impact on print longevity. This is through
things like whether rich blacks or pure ink black is being used,
variation in the use of light ink colors, etc.

Henry was out in Australia to talk at a photo industry conference about
the new Epson 7800/9800 printers, whose prints he rates very highly for
both display and storage life. He has a lot of good things to say about
the Epson’s monochrome print life because the black inks use carbon
black pigments, about as long lasting as you can get. We’ll have more
to say about this in a separate article coming soon.

Both Henry and WIR are doing a great service to those of us who use digital prints.

That’s all we had time for over lunch but expect more in the future.

When Size Does Matter รข

Some images just need to be printed large.
Whilst most of us stick with small prints, 4″ x 5″ up to 8″ x 10″, we
are really missing out on things. Photographs, like all other forms of
visual art, have a natural size. Highly detailed, intricate works
perform very well in the smaller sizes, such as 8″ x 10″ or 11″ x 14″,
because they invite the viewer to move in close and become absorbed in
the detail. Other images work much better in a larger size. In
this article we will examine all the issues around how and why to print
in poster size or larger.

How To Print Large

Step 1 – Assess the image

Have a good look at the image in whatever form you have it, whether it
is a normal print from the drug store or a digital file from your
camera. Assess things like its sharpness, level of detail and subject
matter. Subject matter is a highly important issue to how large you
should print. Shocking or confrontational images may be better small.
At a recent exhibition I saw some wonderful photographic images of
Auschwitz that were printed small, around the 8″ x 10″ size. They
worked well in this size because their level of detail invited you in
close. But by being small the subject matter did not push you away.
However such images could also have been printed large. In this case
the emotional response of the viewer would have been very different
because of the impact the larger size would make. Thus you also need to
evaluate what impact you want your image to have on the viewer, what
statement are you trying to make. This is the big difference between a
fine art photographer and a snap shooter. The fine art photographer
will have a clear intent and message that guides how they present their
images. Now this could be as simple as “the beauty of nature” or as
complex and confrontational as “man’s inhumanity to man”, but it will
still be there.

Step 2 – Size and resolution

From the above work out what size you wish to print at and with what
technology. The type of printing you will use determines the resolution
in dots per inch that you must supply to the printer. (see the endbar
Resolution Issues) Continuous tone devices like the digital
photographic printers available at professional bureaus will need
higher resolution input than your ink jet printer may, but they will
produce a smoother result. Tied into these considerations are the
surfaces you will print on. Glossy photographic type paper will need
more resolution than printing on canvas or watercolor paper. This is
because the texture of the material hides many inadequacies in the
image data and adds apparent detail. Plus when you print on artists
materials, like uncoated watercolor paper, the ink dots will spread more, making
the image smoother.

Take your chosen size and multiply it by the dpi rating you need to use
to get the absolute size of the image you require in pixels. For
example a 24″ x 36″ print at 100 dpi requires a 2400 x 3600 pixel
image, or an eight and a half mega pixel image.

Step 3 – Input Options

If you work with a digital camera you simply need to transfer your
images to the computer. With digital cameras I always shoot at the
highest resolution the camera is really capable of, unless I know I
will only ever use the image for a low-res task, like a web site.
Shooting at the highest resolution gives you the most options.

Those of us who still shoot on film and then scan have more to think
about. You can either scan the film or a print. It is always better to
scan the film if possible. A print is a second-generation version of
what is on the film. It will be slightly softer than the original on
film plus there is always some texture to the paper it is printed on
that will also scan in. This said, I have achieved wonderful results
scanning prints on a cheap flatbed scanner. Just make sure that the
scanner is clean and dust-free, and that the print is also clean and
free of fingerprints, etc. Scanning the film will offer the best
results and more potential resolution. 35mm film scanners are
relatively inexpensive and produce lots of resolution. Remember that you can also get
scans done at a bureau at reasonable price with as much resolution as
you need. This is a cost-effective choice if you only need high-res
scans occasionally.

When scanning photos I recommend using the maximum optical resolution
of your scanner. A 1200 x 2400 dpi flatbed scanner has a real optical resolution of 1200 dpi even though
you can set the resolution as high as 9600 dpi. I find it is better to
interpolate up in PhotoShop if necessary. Also if your scanner can
provide more than 24 bits of color data per pixel over to PhotoShop (or
whichever imaging program you use) do so. This gives you more data to
work with. If this is not possible, so as much of the color correction
and contrast adjustment within the scanner software as possible. This
ensures that you will have the most useable color information available
in PhotoShop.

Step 4 – Removing information

There may be some information in the image that you need to remove. At
this stage you should generally cleanup the image by using the clone
tool to remove all dust and hair marks or scratch marks. Another thing
to examine is noise in the image. This may be digital noise in a
digital camera image or film grain in a scanned image. Especially if
you have to enlarge the image substantially this may be very
noticeable, so it is better removed now. One solution is to apply a
Gausian Blur filter, the other being the Speckle filter. Note though
that you have to be careful not to soften the image too much. Since
noise is most noticeable in areas with little detail you could mask
areas and only blur those.

Step 5 – Adjusting resolution

If your input medium provides enough or more resolution than you
require there is not much to do at this stage. Use the Image ->
Image Size dialog in PhotoShop to set the size and resolution you
require, making sure that you select the Resample Image checkbox.
Images of the right resolution will still normally go though this to
adjust the dimensions and dpi rating but without changing the file
size. Say you have scanned a 4″ x 5″ photo at 1200 dpi and you are
printing it at 40″ x 50″ at 120 dpi. The number of pixels won’t change
but you will still need to tell PhotoShop the correct dimensions and
resolution so it can print correctly.

What do you do if you have to increase the
size of the image? Simply
use the Image -> Image Size dialog as above to increase the
resolution with the interpolation method set to Bicubic, which
generally offers the best results. For larger size increases (more than
2x) also do this but it is sometimes better to perform this in a number
of smaller steps (say 1.5x to 2.0x) with some sharpening applied inbetween rather than in one big jump. Try both ways and come the result.

Step 6 – Sharpen the image

All scans and digital camera shots usually benefit from some amount of
sharpening with PhotoShop’s Unsharp Mask filter. This is especially
true if you have increased the resolution of the image. The danger here
is that as well as sharpening detail you want additional detail, like
noise and film grain, will be sharpened. If you got step 4 right this
should not happen. Remember that the dialog box allows you to adjust
the threshold of color difference below which no sharpening occurs. The
radius of the sharpening should vary depending on the size of your
image and beware of going above 150% sharpening. Adjust the settings
until you get the effect you want. Then examine carefully various parts
of the image. If you are not happy with the result use Undo or the
History list to go back and try again. Remember that you can also
sharpen only selected parts of the image.

Step 7 – Adding information

If you have done a major resolution increase you may need to add
information to the image. This all depends on the intended use of the
image. If you are a dab hand with the mouse or graphics tablet you can
draw in fine detail, like small branches in the trees, etc. You may
also be able to clone parts of the image. When doing all this work
first duplicate your image layer, then create a new layer above this
and draw on that. By working with separate layers you can ensure that
you can easily change your mind or even create different images from
the same photo by selectively turning layers on and off.

Resolution Issues (as opposed to issue resolution, which is a job for a psychologist or guru☺)

There is a lot of confusion about resolutions, mainly caused from the
fact that there are two ways to specify it. You can express image
resolution in absolute numbers of pixel terms. This is the best and
least confusing way as you can easily convert to the other form from
here. We know that a 1280 x 1024 pixel image contains 1,310,720 pixels,
and with three bytes or 24 bits of color information per pixel this
will occupy around 3.9Mbytes without compression. The second expression
of resolution is as measurement dimensions and the number of pixels or
dots per unit of measurement. Thus it could be an 8″ x 10″ image at 300
dpi. This in itself is not confusing but its usage can be.

Printers and scanners often express their resolution in dpi. For
scanners this is a reasonable approach. An older Nikon film scanner has a
resolution of 2,700 dpi. Thus a full frame 35mm negative has a size of
24mm x 36mm, or 0.945″ x 1.417″. It thus produces a scan of 2700 x
0.945 by 2700 x 1.417 or 2551 by 3825 pixels. Within PhotoShop, by using
the Image -> Image Size dialog, we can change the dimensions or the
dpi resolution to any setting we like without changing the numbers of
pixels by un-checking the Resample Image checkbox. Thus we can make
this roughly 1″ x 1.4″, 2700dpi image a 8.5″ x 12.7″, 300dpi image
without the need to change anything except how PhotoShop sizes the
image. The important thing is the number of pixels you have. You can
spread these thinly over a large area or pack them tightly into a small
one, but the number can stay the same.

The problems with dpi resolution expressions come with printers. When a
printer says it has a resolution of 1,440dpi all this means is that it
can place 1,440 dots per inch of cyan, magenta, yellow or black ink. It
can not put 1,440 dots of pink, purple or skin tone per inch. Note that
I am not talking here of continuous tone printers, like dye sublimation
printers, only of inkjets and color laser printers. To get colors other
than the actual ink color, numbers of dots need to be laid down close
together in a process called dithering (not the sort a Libra does when
confronted with having to make a decision). If the size and closeness
of the dots is good enough and your eyes are far enough away you see
this as smooth, continuous color. This means that such a 1,440dpi
printer may not be able to print photographic images at much better
than 300dpi. So you don’t need to make your images 1,440dpi resolution
to print them on such a printer. You can easily determine what
resolution you really need by testing. Take an image with lots of
detail at various scales. Change its resolution to that of the
printer’s dpi rating. For example we’ll do this for the author’s 300dpi
large format printer. Then generate additional image files by
resampling this original down to, say, 150, 100 and 50dpi. Print all
the versions and compare. You will notice that, at some point, you
really stop getting more detail even though the resolution sent to the
printer was higher. You can do more at in-between resolutions to
fine-tune this. For my printer it is around 80dpi. This means I gain
nothing by preparing files of higher resolution for printing. They are
just bigger to store, slower to work on and take longer to send to the
printer. Whilst 80dpi might seem very low to you it is important to
remember that most of us never view a huge print up close. So an 80dpi
36″ x 48″ print at a normal viewing distance can look as sharp or
sharper than a 400dpi 8″ x 10″ print in your lap.