The diagnosis of hardwood

vessel picking in offset prints

By H.U. Heintze
HE OFFSET PRINTING PROCESS is prone to
contamination of the offset blankets,
generally with a resulting visible
defect in the printed material. Print-
ing press contamination can arise
from a number of paper sources, as well as from
the pressroom. Any attempt to diagnose and solve
an offset press picking problem must be guided
by a proper definition of the nature of the prob-
lem [3]. An in-depth physical or chemical analy-
sis should be preceded by documentation of the
frequency and appearance of the defects, fol-
lowed by estimation of their sizes. However,
despite the fact that such defects are commonly
encountered in complaints, benchmarking stud-
ies and product development trials, the evalua-
tion of how many defects appear in a print and
how big they are remains subjective.
The vessel elements that are an essential com-
ponent of every hardwood species give rise to a
peculiar offset printing defect variously known as
lint, fuzz, or vessel picking in woodfree fine
papers. Overviews have been published that give
a good introduction to the vessel pick problem
and possible remedies [2, 4]. The objective of the
present work is to provide a practical guide for
mills and their technical service representatives to
address the question of how many picking defects
are in a print area and how big they are. In par-
ticular, an outline is given of a method to quanti-
fy the severity of hardwood vessel picking in the
offset printing of fine papers.
High-resolution page scanners now cost less
than a thousand dollars and are capable of pro-
ducing very good images. Disk storage on a PC is
also inexpensive and, with little effort, it becomes
feasible for a mill to establish a pictorial database
of offset picking and other print quality prob-
lems. Such a reference can help to identify the
source of the picking in present and future com-
plaints. While this may not be as good as having
photomicrographs prepared by an expert micro-
scopist/photographer, it is within the reach of
any mill that deals with offset press contamina-
tion problems.
EXPERIMENTAL
The samples used in this study were commercial-
ly printed uncoated fine papers, primarily direct
34 108:1 (2007) PULP & PAPER CANADA
offset printing T1
mail advertising pieces and a variety of financial
statements. All samples were printed by offset.
The selected prints were scanned at an optical
resolution of up to 94 pixels/mm (2400 dpi)
using a Canon D2400UF scanner with an optical
resolution of 94 pixels/mm x 189 pixels/mm
(2400 dpi x 4800 dpi). Some prints were also
scanned at an optical resolution of up to 189 pix-
els/mm (4800 dpi) using an Epson 4870 photo-
scanner with an optical resolution of 189 pix-
els/mm x 378 pixels/mm (4800 dpi x 9600 dpi).
The pages printed on both sides were backed by
a black sheet to minimize the effect of print-
through from the reverse of the sheet
Higher magnification images were obtained
using a Canon 10D 6.3 mega-pixel digital camera
fitted to a Wild-Heerbrugg microscope. Image file
manipulations such as cropping and rotating
were done with Jasc PaintShop Pro 9.0. Image
analysis measurements were made using ImageJ
(version 1.34), the public domain software devel-
oped at the United States National Institutes of
Health by Wayne Rasband and available on the
Internet at http://rsb.info.nih.gov/ij/. This soft-
ware runs under the Windows operating system,
as well as under Linux and on Apple computers.
Over twelve hundred users of ImageJ support the
development of new measurement routines that
address a wide range of problems. All files were
converted to 8-bit grayscale images for the opera-
tions carried out in this study.
CAPTURING A DIGITAL IMAGE
OF A PRINT DEFECT
Definition of the problem
The diagnosis of offset press blanket contamina-
tion normally starts with a visual examination of
the defect in the print or on a suitable tape pull
from the press blanket. In an ideal situation, a
microscopist is available to produce photomicro-
graphs that illustrate the defect and produce esti-
mates of their size.
Figure 1 illustrates the appearance of three typ-
ical defects, as captured with a 2400 dpi scanner.
All three pictures in Fig. 1 are at the same magni-
fication. The resolution may not be as good as can
be achieved with a microscope, but these images
can be captured quickly and inexpensively.
The image on the left is a typical hickey, the
H.U. HEINTZE,
Consultant
Montreal, QC
[email protected]
T
Abstract: Offset printing of fine papers containing hardwood-fibre can give rise to small, repeat-
ing defects in printed solids or halftones. The evaluation of this and of other picking problems is
subjective. Public domain image analysis software and consumer-level page scanners can help with
the quantification of this problem. An analysis of the issues is given, along with typical results. Rec-
ommendations are made for the implementation of this low-cost approach.
most common defect appearing in offset
prints on fine papers. The centre image
displays a common problem caused by
several types of contaminant. The small
white repeating defects on the right
image in Fig. 1 are typical of hardwood
vessel picking. These are smaller than the
defects in the other two images and are
not likely to be confused with them when
comparing magnified images such as
these. The size and shape of these repeat-
ing defects is characteristic of the hard-
wood species from which the vessel ele-
ments are derived.
It should be clear that the ability to
show manufacturing groups pictures such
as these will be more helpful than simply
saying we have a picking problem. Clear
definition and communication of the
defect appearance are essential to ensure
an effective search for a solution.
The structure of a scanned image
The image captured by a scanner or by a
digital camera is made up of discrete units
called pixels and each pixel represents a
certain area and average intensity (gray
level) on the scanned original picture.
Figure 2 illustrates the sequence from a
continuous tone photomicrograph origi-
nal at the left to a coarse pixel represen-
tation at the right. The continuous-tone
image can be considered as divided into a
grid representing the individual scanner
elements or pixels. The gray level of each
pixel can be thought of as approximately
the average brightness of the image area
captured by each pixel. Instrumental fac-
tors and image capture software also play
a role and have to be evaluated for specif-
ic applications. A detailed understanding
of digital image capture and manipula-
tion can be found in sources such as The
Image Processing Handbook [1].
Scanner resolution setting
The resolution setting on a scanner
defines the number of pixels that make
up a captured image feature. Figure 3
compares an image taken with a micro-
scope and camera (left) to the same area
of print captured with an Epson scanner
at resolution settings of 4800 dpi (centre)
and 1200 dpi (right). The differences in
image resolution are obvious. The
scanned image at 4800 dpi gives a reason-
able approximation of the white area
caused by a vessel on the press blanket
and even gives an indication of the fibre
structure in the print. The image at the
right, scanned at 1200 dpi, clearly has
poorer resolution, making it more diffi-
cult to characterise the shape and size of
the vessel defect. Since image analysis
works with the individual pixels in an
image, it should be evident that a lower-
resolution image, such as the 1200 dpi
image on the right of Fig. 3, will provide
less precise measurements such as image
area or length.
Although a higher scanner resolution
is desirable to give better definition to the
individual print defects, there are trade-
offs against the resulting scan times and
image file sizes, especially as the total area
to be scanned is increased. Scan times on
the CanoScan for a 50 mm x 50 mm area
of print increase from 20 seconds at 600
dpi to 145 seconds at 2400 dpi. Compara-
ble times on the Epson Perfection 4870
Scanner are 12 seconds and 101 seconds.
At 4800 dpi the scan time on the Epson
increases to 214 seconds. Higher scanner
resolution settings also increase the size of
the resulting image file. TIFF (tagged
image file format) 8-bit grayscale file sizes
for the 50 mm x 50mm areas scanned on
the Epson increased from 1.3 MB at 600
dpi to 21.2 MB at 2400 dpi and to 85.1 MB
at 4800 dpi. These scan times and file sizes
are a clear indication that it may not
always be desirable to scan larger areas at
very high scanner resolution.
Scanner exposure setting
Image capture always involves some form
of exposure setting to balance the specif-
ic requirements of the original image, the
illumination and the sensitivity of the
image capture device. This is true for old-
er film-based methods and also for newer
digital cameras and scanners. It is impor-
tant to evaluate the effect of scanner
exposure control in developing a database
of scanned defect images. Page scanners
are designed to capture relatively large
images, such as documents or pho-
tographs, for digital enhancement. Scan-
ner control software is frequently installed
with a default option to optimize the
quality of the captured image. Such auto-
matic image enhancement is not desirable
for the present purpose. It can have unde-
sirable consequences, such as reducing
contrast or producing images that are too
light or dark for the intended purpose.
Suitable scanner exposure settings must
be identified and controlled for each
application. Figure 4 illustrates the effect
of extreme differences in exposure set-
tings. The high contrast image on the
right was scanned with an automatic
enhancement setting and it is obvious that
this has changed not only the overall gray
levels, but also has increased the size of
the vessel pick features. This indicates the
need for a high-quality scanner, adequate
lamp warm-up time and the use of known
and controlled scan exposure settings.
This is not a major problem; it just
requires attention to detail and a docu-
mented method for scanning images. A
control sample is always a good idea to
check on the stability of any test over time.
IMAGE ANALYSIS
MEASUREMENT OF A
PRINT DEFECT
Basic considerations for image analysis
size measurement
A database of scanned image defects can
serve simply as a basis for visual compar-
isons, but it is better to also estimate the
size of printing defects such as hickeys or
repeating vessel picks. A scanned digital
image defect can be quantified by the sim-
ple expedient of using a ruler on the
monitor and a scaling factor or by the use
of image analysis software.
Image analysis systems and software
have been available for a long time [6].
Proprietary image analysis software is fre-
quently bundled with dedicated instru-
ments for specific applications. The cost
of such software and the apparent com-
plexity of image analysis concepts have
discouraged many people from consider-
ing it as an option for their less-common
measurement needs. However, the avail-
PULP & PAPER CANADA 108:1 (2007) 35
offset printing T2
FIG. 1. Examples of offset blanket contamination with fine
papers. Hickey on the left, hickeys and long threads in
centre, and hardwood vessel picking on the right.
FIG. 2. Illustration of the transformation of a continuous-
tone, high-resolution photomicrograph image at the left to
a low-resolution pixel image at the right
ability of high-quality public domain soft-
ware, such as ImageJ, provides a path for
the casual or infrequent user to gain
access to the power of image analysis for
applications where canned commercial
products do not exist or cannot be justi-
fied economically.
The basic difference between the
human visual system and an image analy-
sis system is the ease with which we can
analyze patterns in complex images. The
eye-brain combination is so efficient that
we rarely consider how it works. On the
other hand, an image analysis system has
to be given specific rules to separate the
desired features from their background.
This is done by adjusting the gray level of
a features boundary until its outline
agrees with visual assessment. This process
is called thresholding and is most com-
monly done on 8-bit grayscale images
where the gray levels can take on values
ranging from 0 for black to 255 for white.
The definition of measurement condi-
tions that quantify small white defects in
offset prints is remarkably easy for an oth-
erwise uniform printed solid area. It is up
to the user to ensure that the white spots
have the shape and size characteristic of
hardwood vessels. Photomicrographs of
vessel elements for various hardwood
species have been published [5, 7].
Lengths of larger vessel elements for vari-
ous species were measured from pho-
tomicrographs [7] and are summarized in
Fig. 5. Length is not an unambiguous
indicator of vessels by species, since there
will be differences in vessel length and
width for any given species due to season-
al growth patterns [5,7]. In addition,
some species such as oaks have earlywood
vessels that, in the tree, are wider than
they are long, giving a square appearance
in offset vessel picking.
It is evident from Fig. 5 that maples
tend to have smaller/shorter vessels than
birches or poplars. Thus, the length of
vessel pick images can provide an indirect
indication of the wood species causing the
problem. People planning on using this
approach should first familiarize them-
selves with their pulp furnish and the ves-
sel shapes/sizes corresponding to the
hardwood species being used [5,7].
How to define image analysis settings for
vessel picking
The details of developing a method will
vary slightly, depending on the image
analysis software that is used. The follow-
ing steps are kept general in nature, but
they should guide anyone wishing to use
this approach.
The first step is to scan the print area
containing the defect(s) at a suitable scan-
ner resolution and exposure. The selected
print area should be a solid print with
100% ink coverage. Vessel picking in
halftone screen areas takes the form of
missing screen dots that are noticed visu-
ally but are lacking the clear boundary
required for image analysis. The selected
area should be neither too small nor too
large. A print area of 60 mm x 160 mm
that is scanned at a resolution of 1200 dpi
will produce a TIFF file of about 22 MB.
This gives a good balance of image area,
resolution and file size for hardwood ves-
sel picking. The maximum image file size
that can be analysed depends on the avail-
able computer system. A typical PC with
512 MB is adequate for most work, but it
should be noted that ImageJ will only use
up to about 1.5GB of RAM. Other limita-
tions may arise from video card memory.
The second step is to define an appro-
priate threshold for the image. Thresh-
olding is done on a grayscale image and,
if the original was scanned in colour, it
must be converted to an 8-bit grayscale
image. The easiest definition of a suitable
threshold is to set the gray level threshold
at the midpoint of the average grayscale
levels for the light and dark parts of the
image in the vicinity of a defect. Doing
this is easier than it may sound. Magnify a
section of the grayscale image around a
typical white spot and measure the gray
level profile of a selection line drawn
across the spot. Set the threshold halfway
between the gray level of the white area of
the defect and that of the dark area of the
surrounding print. The defect area seen
at this threshold setting should agree with
visual assessment of the defect size before
thresholding. Make sure the image on the
monitor has been zoomed to a sufficient
level to show individual pixels.
Features below the smallest size of a
vessel pick defect are to be considered as
noise. They are excluded from the mea-
surement by a suitable size criterion. The
minimum size criterion is chosen to be
smaller than the minimum likely vessel
size in the problem prints [5,7]. Initially,
measurements are done at several values
of the exclusion criterion to confirm that
the desired results are stable.
The next step is to define the mea-
surement that is desired for the defects.
This will depend on the hardwood species
that give rise to the problem, since that
determines vessel size and shape. Exami-
nation of Figs. 2 and 3 indicates that the
longest dimension of a vessel pick has
more pixels than the width. Thus, at any
scanner resolution, measurement of
longest dimension may provide better
definition of vessel pick size than either
width or area. Longest dimension of ves-
sels is obtained by selecting Ferets diame-
ter as the measurement. Measurements
can be made more complicated, but there
may be little return on the additional
investment of time and effort.
ImageJ provides options for various
output formats, but it is prudent to save
the output image of the measured fea-
tures as a separate file for comparison
with the original and to copy the detailed
feature measurements to a spreadsheet
for future reference. Image analysis mea-
surements should never be performed on
the only copy of an image file! Measure-
ments should always use a copy of the file
in case something goes wrong and the file
is lost.
Image analysis software generally has a
macro capability that allows a multi-step
analysis to be recorded for future play-
back. For repeat measurements of vessel
pick or other defects in an offset print, it
is easy to set up an automated image anal-
36 108:1 (2007) PULP & PAPER CANADA
offset printing T3
FIG. 3. Comparison of the decrease in image resolution in
going from a photomicrograph at left to scanned images at
4800 dpi (centre) and 1200 dpi (right).
FIG. 4. The effect of scanner exposure setting on image
appearance. Note that vessel pick images on the right are
larger than those on the left.
ysis routine activated by keyboard function keys in order to min-
imize the risk of operator error.
A detailed written method should be prepared for any image
analysis method and it should clearly spell out all the steps in
sample preparation, scanner set-up, image analysis operations
and analysis. Minor details like scanner cleanliness are very
important since a speck of dust on the scanner plate will end up
looking like a speck in the scanned image. It should be obvious
that the print sample to be scanned also must be cleaned of any
loose dust with a soft brush or air blower before putting it in the
scanner. Loose fibres or other contaminants on the surface of a
printed sheet are easily mistaken for something picked out of
the print.
As pointed out in the discussion of vessel lengths by species,
the measured values of vessel pick length in the prints can be
used to help pinpoint the hardwood species causing the prob-
lem. Some time spent examining vessel images [6,7] will allow
even the non-expert to see the similarity with the shapes of the
defects in the prints.
Examples of image analysis for vessel picking
The methodology described above was evaluated on both the
Canon and Epson scanners at resolution settings of 600, 1200, and
2400 dpi. In addition, evaluations on the Epson at 4800 dpi quick-
ly showed the problems arising from very large file sizes at this high
resolution. A 20 mm x 20 mm print area scanned as a grayscale
image at 4800 dpi produced a 14 MB TIFF file. Thus, it is suitable
for scanning small defect areas to examine their fine structure, but
this resolution is too high to be used for larger areas.
The ImageJ macro was modified for each scanner resolution
setting so that the minimum size exclusion limit corresponded
to the same physical size on the print for each resolution, but
other factors were unchanged. Figure 6 shows a print with very
heavy vessel picking scanned at 1200 dpi and the resulting out-
line of measured features as output by ImageJ.
Area and longest dimension were also recorded. It should be
evident from Fig. 6 that these defects do not all fit the model of
simple straight lines. Thus, longest dimension is an approximation,
but one that appears to be adequate for the present objective.
The eight features in Fig. 6 range in longest dimension from a
measured low of 0.44 mm to a high of 1.00 mm. On the basis of
these lengths it would be reasonable to exclude maples as the like-
ly cause of this particular problem and one would focus on species
such as birch, poplar or alder. A knowledge of the pulp furnish
would obviously simplify the puzzle of relating sizes to species.
A length of 1 mm at a scanner resolution of 1200 dpi corre-
sponds to 47 pixels and this defines a best case measuring pre-
cision of +/- one pixel as equivalent to +/- 0.02 mm. Obtaining
this degree of precision by microscopy or by other means would
be a lot more cumbersome and slower than the simple and low-
cost approach described here.
Measurements on a range of commercial samples in both
black and colour provided similar results, and enumeration of
other examples would add little to the understanding of the
method. The key is to avoid halftones, convert all images to
grayscale and set the threshold at the value appropriate to the
sample. The image analysis quantification of hickeys and other
press contamination defects has not been covered in detail here,
but it is simply a variant on the approach described for vessel
picking.
Scanner stability over time is an issue that must be addressed
and that will depend on the equipment used. As already indicat-
ed, the use of a control sample is a useful way to identify problems.
CONCLUSIONS AND RECOMMENDATIONS
The investigation of picking defects in offset printing has long
suffered from the difficulty in communicating to all parties the
exact nature of the problem in the pressroom. The defects are
generally small and appear at low frequency, and the original
printed samples are not always available to all parties involved in
fixing the problem. Photographic capture of the defects is the
ideal solution, but requires a skilled operator with a very good
macro-lens set-up or a microscope. These options are not always
available in a paper mill. Where they are available, they tend to
be time-consuming and expensive.
High-resolution page scanners provide a fast and low-cost way
to document the appearance of picking defects in offset print-
ing. A mill-specific catalogue of defect images is an essential tool
in the efficient communication and resolution of picking prob-
lems. Without such a guide it is very easy for different percep-
PULP & PAPER CANADA 108:1 (2007) 37
offset printing T4
FIG. 5. Comparison of the larger vessel element lengths for
various hardwood species.
FIG. 6. Example of vessel pick measurement on a 10 mm x
10 mm image at 1200 dpi. Scanned image is shown at left
and ImageJ output of measured features is shown at right.
tions of a problem to slow down progress
in finding a solution. Scanned defect
images can be shared with all parties
involved in solving the problem and this
will help to ensure a common under-
standing of what is being addressed.
A high-resolution scanned image cov-
ers a much larger area than a microscope
view, and panning across a zoomed ver-
sion of the image on a computer monitor
provides a convenient virtual micro-
scope for the examination of individual
print defects.
ImageJ public domain image analysis
software allows the rapid measurement of
defect size and shape parameters for
defects in uniform solid offset prints.
This zero-cost software offers the potential
to take basic image analysis out of the
domain of research centres and make it
an everyday mill tool for a large variety of
problems. It should no longer be neces-
sary for anyone to have to resort to the
evaluation that sample looks like it has
more picking.
LITERATURE
1. RUSS, J.C., ed. The image processing handbook. CRC
Press (1999).
2. SHALLHORN, P.M., HEINTZE, H.U. Hardwood
vessel picking in the offset printing of uncoated fine
papers. Pulp & Paper Can. 98(10): T353-T356 (1997).
3. EAMER, M., HEINTZE, H. Your paper, your cus-
tomer, his press. Pulp & Paper Can. 96(11): 21-22
(1995).
4. HEINTZE, H.U., SHALLHORN, P.M. Hardwood
vessel picking and the manufacturing process. Pulp &
Paper Can. 96(11): T365-T367 (1995).
5. PARHAM, R.A., GRAY, R.L., ed. The practical identi-
fication of woodpulp fibers. Tappi Press (1982).
6. GARTAGANIS, P.A., HEINTZE, H.U., GORDON,
R.W. From Printograph to Videoscanner. Tappi J.
59(12): 113-117 (1976).
7. CARPENTER, C.H., LENEY, L. 382 Photomicro-
graphs of 91 papermaking fibers. State University of New
York, College of Forestry at Syracuse, Technical Publi-
cation 74 (1952).
38 108:1 (2007) PULP & PAPER CANADA
offset printing T5
Reference: HEINTZE, H.U. The diagnosis of hardwood vessel picking in offset prints. Pulp &
Paper Canada 108(1): T1-5 (January, 2007). Paper presented at the 92nd Annual Meeting in Mon-
treal, QC, February 6-10, 2006. Not to be reproduced without permission of PAPTAC. Manuscript
received September 29, 2005. Revised manuscript approved for publication by the Review Panel
on April 24, 2006.
Keywords: FINE PAPERS, PICKING, OFFSET PRINTING, IMAGE ANALYSIS .
Rsum: De petits dfauts rcurrents dans les imprims en aplat ou en simili apparaissent lors
de limpression offset des papiers fins contenant de la fibre de feuillus. Le processus dvaluation
de ces dfauts et dautres dfauts associs larrachage est de nature subjective. Les logiciels
danalyse dimage du domaine public et les analyseur pleine page lintention de la population
en gnral peuvent servir quantifier ce problme. Une analyse des problmes est prsente, ain-
si que des rsultats typiques. Des recommandations sont offertes pour la mise en oeuvre de cette
mthode conomique.