Moiré interference patterns

[Terry Pinnell]

First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)

What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

[Hans-Georg Michna]

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Terry,

filters exist that take the mathematical approach to remove
moiré patterns entirely. The last time I encountered such a
filter was when I still used a program called Picture Publisher
by Micrografx. This program had an excellent moiré filter which
used clever mathematics to obtain a precise model of the moiré
pattern and deducted it perfectly without losing any visible
sharpness.

Unfortunately I haven't seen anything like it ever since, and I
had to drop Picture Publisher because of its excessive number of
defects.

The moiré filter in Paint Shop Pro is a joke in comparison. It
is only a soft filter that essentially doesn't remove the moiré
at all, it just reduces it along with any other high frequencies
by applying a low pass filter, making the whole picture soft.

I would also like to know a third-party tool that does what
Picture Publisher did.

Hans-Georg

[Terry Pinnell]

Post by Terry Pinnell

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Terry,
filters exist that take the mathematical approach to remove
moiré patterns entirely. The last time I encountered such a
filter was when I still used a program called Picture Publisher
by Micrografx. This program had an excellent moiré filter which
used clever mathematics to obtain a precise model of the moiré
pattern and deducted it perfectly without losing any visible
sharpness.
Unfortunately I haven't seen anything like it ever since, and I
had to drop Picture Publisher because of its excessive number of
defects.
The moiré filter in Paint Shop Pro is a joke in comparison. It
is only a soft filter that essentially doesn't remove the moiré
at all, it just reduces it along with any other high frequencies
by applying a low pass filter, making the whole picture soft.
I would also like to know a third-party tool that does what
Picture Publisher did.

Thanks Hans-Georg.

Since my post, I've discovered something very puzzling. By default, I
was viewing the finished JPG scan in IrfanView. That showed the bad
patterns. And they remained with various DPI changes in Epson Scan
(300, 400, 600, 150, 96). But I happened to open one in PaintShop Pro
8 (in order to see what filters I might use), and the patterns were
gone! This is with both programs filling the screen with the image to
about the same size.

I also found something else that I haven't figured out. Those first
scans were done by opening Epson Scan direct and using its
Professional Mode. But just now I instead tried opening Smart Panel
and then selecting Scan & Save. That doesn't present any options, it
just appears to scan using Fully Auto mode (and takes a fair while,
presumably because it's scanning the entire area, because there's no
Preview allowing cropping). But the resulting JPG was not patterned!

Any help getting my mind around these would be appreciated please.

[Ed Ruf (REPLY to E-MAIL IN SIG!)]

On Wed, 30 Nov 2005 19:25:12 +0000, in rec.photo.digital Terry Pinnell

Post by Terry Pinnell
Since my post, I've discovered something very puzzling. By default, I
was viewing the finished JPG scan in IrfanView. That showed the bad
patterns. And they remained with various DPI changes in Epson Scan
(300, 400, 600, 150, 96). But I happened to open one in PaintShop Pro
8 (in order to see what filters I might use), and the patterns were
gone! This is with both programs filling the screen with the image to
about the same size.

Probably different resampling/display algorithms.

[Frank ess]

Post by Terry Pinnell

Post by Terry Pinnell

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning
at say 600 dpi and then using an image editor to reduce the size?

Terry,
filters exist that take the mathematical approach to remove
moiré patterns entirely. The last time I encountered such a
filter was when I still used a program called Picture Publisher
by Micrografx. This program had an excellent moiré filter which
used clever mathematics to obtain a precise model of the moiré
pattern and deducted it perfectly without losing any visible
sharpness.
Unfortunately I haven't seen anything like it ever since, and I
had to drop Picture Publisher because of its excessive number of
defects.
The moiré filter in Paint Shop Pro is a joke in comparison. It
is only a soft filter that essentially doesn't remove the moiré
at all, it just reduces it along with any other high frequencies
by applying a low pass filter, making the whole picture soft.
I would also like to know a third-party tool that does what
Picture Publisher did.

Thanks Hans-Georg.
Since my post, I've discovered something very puzzling. By default, I
was viewing the finished JPG scan in IrfanView. That showed the bad
patterns. And they remained with various DPI changes in Epson Scan
(300, 400, 600, 150, 96). But I happened to open one in PaintShop Pro
8 (in order to see what filters I might use), and the patterns were
gone! This is with both programs filling the screen with the image to
about the same size.
I also found something else that I haven't figured out. Those first
scans were done by opening Epson Scan direct and using its
Professional Mode. But just now I instead tried opening Smart Panel
and then selecting Scan & Save. That doesn't present any options, it
just appears to scan using Fully Auto mode (and takes a fair while,
presumably because it's scanning the entire area, because there's no
Preview allowing cropping). But the resulting JPG was not patterned!
Any help getting my mind around these would be appreciated please.

My Epson Scan software ver. 2.03A offers a descreening filter with
four Screen Ruling 'level' choices. In professional mode it is near
the bottom of options offered, just below Unsharp Mask Filter.

If that doesn't have the desired effect, do a Gaussian Blur sufficient
to your purposes. On a 3000x2000 ppi image, a blur of something
between 0.8 and 1.3 seems to eliminate the pattern effect without too
much of a deleterious effect on detail.
Loading Image... (99KB file)

My guess about the Fully Auto mode is that it applied the Descreening.

It doesn't require much of an offset or interference to render a
pattern effect onscreen (anoraks please forgive my unfamiliarity with
approprate terms of art). When I look at a scanned programme cover in
Windows' viewer or Photo Shop, rolling the mouse wheel in the view
resizing function passes through several sizes that are alternately
pattern-free and patterned in various configurations. Proportion of
the screen utilized also has an effect.

[Terry Pinnell]

Post by Frank ess
My Epson Scan software ver. 2.03A offers a descreening filter with
four Screen Ruling 'level' choices. In professional mode it is near
the bottom of options offered, just below Unsharp Mask Filter.

Thanks! Now found it from those directions. I hadn't realised I needed
to scroll downwards in order to see all my setting options. And I use
a 1024 x 768 screen, which I don't regard as abnormally small <g>. (I
also see a Dust Removal option hiding at the bottom.)

Post by Frank ess
If that doesn't have the desired effect,do a Gaussian Blur sufficient
to your purposes. On a 3000x2000 ppi image, a blur of something
between 0.8 and 1.3 seems to eliminate the pattern effect without too
much of a deleterious effect on detail.
http://www.fototime.com/02FB5204FA89BE0/orig.jpg (99KB file)
My guess about the Fully Auto mode is that it applied the Descreening.
It doesn't require much of an offset or interference to render a
pattern effect onscreen (anoraks please forgive my unfamiliarity with
approprate terms of art). When I look at a scanned programme cover in
Windows' viewer or Photo Shop, rolling the mouse wheel in the view
resizing function passes through several sizes that are alternately
pattern-free and patterned in various configurations. Proportion of
the screen utilized also has an effect.

Only tried the Descreening option once so far, but happily it does
seem to work. And, unlike using the Auto Mode, I can now preview and
scan only the selected section, which of course avoids later cropping
and saves time.

But I'm still keen to understand what exactly I should be looking at
to *know* whether the patterning has been removed! IOW, what
definitive steps should I take directly after scanning and saving a
JPG to know whether or not I need to repeat it with different
settings, or process it further with blurring, smoothing, or resizing?
I don't want to find at some much later stage (such as when I've used
the JPG in a DVD slide show), that the patterns are still present.

Until this discussion, I'd never realised there was an issue. But I
now see that different image programs can display the identical file
markedly differently. Furthermore, any *one* of the programs can
display differently depending on factors such as zoom level or window
size.

If I'd had JPG associated by default with PaintShop Pro 8 instead of
IrfanView, I'd never have *seen* the patterns when I did my routine
post-scan view. So I could have done all of them and remained
innocently unaware of a potential problem.

So...would the following make sense as a 'definitive test':

Use IrfanView, because limited experience seems to imply it's the most
sensitive at seeing these patterns. Open the JPG full screen and zoom
in and out across a wide range. If no patterns are seen at any level,
then there *are* none.

Any other suggestions would be appreciated please.

[Hans-Georg Michna]

Post by Terry Pinnell
But I'm still keen to understand what exactly I should be looking at
to *know* whether the patterning has been removed!

Zoom in.

Hans-Georg

[Marvin]

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Turning the picture at a slight angle when scanning can reduce the problem.

[Terry Pinnell]

Post by Marvin

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Turning the picture at a slight angle when scanning can reduce the problem.

Thanks, I'll experiment with that when I get these other puzzles
sorted!

[kctan]

calendar scanning should not get moiré pattern. I think it should be the
screen dots pattern which can be removed by de-screen in your Epson scanner
advance setting menu.

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?
--
Terry, West Sussex, UK

[Terry Pinnell]

Post by kctan
calendar scanning should not get moiré pattern. I think it should be the
screen dots pattern which can be removed by de-screen in your Epson scanner
advance setting menu.

Is there such a feature? If so, as I said, so far I haven't found it.
Could you tell me exactly where it is please?

See also my other replies, re apparent difference between viewers.

BTW, why do you say "Calendar scanning should not get moiré pattern"?

[kctan]

Moiré only happens in subject with very fine surface's textures like textile
or mesh's fabric or net. It happens on photography work only. Calendar
images are press printing on smooth paper surface that will never produce
moiré pattern. Even it does, it is from the printing screens not properly in
control that reproduces as moiré image and not pattern.

Post by Terry Pinnell

Post by kctan
calendar scanning should not get moiré pattern. I think it should be the
screen dots pattern which can be removed by de-screen in your Epson scanner
advance setting menu.

Is there such a feature? If so, as I said, so far I haven't found it.
Could you tell me exactly where it is please?
See also my other replies, re apparent difference between viewers.
BTW, why do you say "Calendar scanning should not get moiré pattern"?
--
Terry, West Sussex, UK

[Terry Pinnell]

Post by kctan
Moiré only happens in subject with very fine surface's textures like textile
or mesh's fabric or net. It happens on photography work only. Calendar
images are press printing on smooth paper surface that will never produce
moiré pattern. Even it does, it is from the printing screens not properly in
control that reproduces as moiré image and not pattern.

Well, still not sure I follow the distinction you're making, but
whatever they're called, I don't want them! FWIW, these patterns
change as I zoom in or out with IrfanView.

[kctan]

Post by Terry Pinnell
Is there such a feature? If so, as I said, so far I haven't found it.
Could you tell me exactly where it is please?
Terry, West Sussex, UK

I can't tell it but mine got to be selected during the time that scanner
starts, from the splash screen, click the advance tab to go into advance
mode. There are plenty of manipulations there.

[Terry Pinnell]

Post by kctan

Post by Terry Pinnell
Is there such a feature? If so, as I said, so far I haven't found it.
Could you tell me exactly where it is please?
Terry, West Sussex, UK

I can't tell it but mine got to be selected during the time that scanner
starts, from the splash screen, click the advance tab to go into advance
mode. There are plenty of manipulations there.

Thanks, have now found the Descreening option (hadn't realised I
needed to scroll down to get access to all settings).

[Wayne]

Post by Terry Pinnell
First time I've come across this effect. Scanning a calendar, I get
gross 'interference patterns'. Understand they're more correctly
called Moiré patterns. I'm disappointed that my Epson Perfection (!)
2480 doesn't seem to have any setting to remove them at source.
(Gather some scanners have a 'de-screening' facility to do this.)
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Yes. Scan images in printed material at least at 300 dpi, and at 600
dpi if you can stand it (if the image is small enough it is bearable).
Moire is digital aliasing, caused by scanning resolution insufficient
to resolve the detail. The dots are the detail, and the dots are
typically (often) 150 dpi, which therefore theoretically requires at
least 300 dpi to resolve them (Nyquist). 600 dpi is better.

There are other workarounds, but the best solution is to increase the
scanning resolution, and then resample it to any smaller desired size.

[Terry Pinnell]

Post by Wayne

Post by Terry Pinnell
What is best way to handle this please? Is it a matter of scanning at
say 600 dpi and then using an image editor to reduce the size?

Yes. Scan images in printed material at least at 300 dpi, and at 600
dpi if you can stand it (if the image is small enough it is bearable).
Moire is digital aliasing, caused by scanning resolution insufficient
to resolve the detail. The dots are the detail, and the dots are
typically (often) 150 dpi, which therefore theoretically requires at
least 300 dpi to resolve them (Nyquist). 600 dpi is better.
There are other workarounds, but the best solution is to increase the
scanning resolution, and then resample it to any smaller desired size.

Thanks Wayne. Just tried one (a photo of about 5" x 8") at 600 dpi
then halved its size in IrfanView and it did indeed make a big
improvement. But please see my reply to Hans-Georg. I'm now confused
as to whether this patterning is *real*, or just a flaw in my usually
excellent IrfanView.

Here are a couple of screen shots showing the comparison. Both scans
were at 240 dpi. I then used Print Scrn to capture both and paste as a
new image in PSP 8. As you see, the view in IrfanView shows the
patterns, but PSP 8 does not (or vet much diminished).

Loading Image...
Loading Image...

[Wayne]

Post by Terry Pinnell
Thanks Wayne. Just tried one (a photo of about 5" x 8") at 600 dpi
then halved its size in IrfanView and it did indeed make a big
improvement. But please see my reply to Hans-Georg. I'm now confused
as to whether this patterning is *real*, or just a flaw in my usually
excellent IrfanView.
Here are a couple of screen shots showing the comparison. Both scans
were at 240 dpi. I then used Print Scrn to capture both and paste as a
new image in PSP 8. As you see, the view in IrfanView shows the
patterns, but PSP 8 does not (or vet much diminished).
http://www.terrypin.dial.pipex.com/Images/Screenshot240-IrfanView.jpg
http://www.terrypin.dial.pipex.com/Images/Screenshot240-PSP.jpg

You always want to judge images ONLY at 100% actual size on the monitor
screen (even if you must scroll around on it). If you scanned at 600 dpi,
it is a huge image, but I think you will see that pattern is false and not
actually present in the image when you view it at 100% actual size.
(IrfanView menu View - Actual Size).

Viewing programs only do a quick and dirty (nearest neighbor) resample to
make large images smaller to fit the screen window (fast and immediate,
but relatively poor quality), and it is not unusual to see moire effects
on the video screen, from even real photos, but especially from the dot
patterns in scanned printed images. Photo editors do this in different
ways, some a little better than others, but none are the same result as
actually viewing the actual real pixels, which are seen only when at 100%
actual size.

If this smaller size is what you want to view on the screen, then use the
resample menu to make a smaller copy that is the right size for viewing.
That resample menu is slower, not quite immediate, but wont create that
false pattern, because it uses better (slower) resampling methods.

[Dave Martindale]

Post by Terry Pinnell
Here are a couple of screen shots showing the comparison. Both scans
were at 240 dpi. I then used Print Scrn to capture both and paste as a
new image in PSP 8. As you see, the view in IrfanView shows the
patterns, but PSP 8 does not (or vet much diminished).
http://www.terrypin.dial.pipex.com/Images/Screenshot240-IrfanView.jpg
http://www.terrypin.dial.pipex.com/Images/Screenshot240-PSP.jpg

It looks like IrfanView is using nearest-neighbour resampling to bring
the image to 50% of original size. This will generate aliasing
artifacts, and exaggerate any that are already present. PSP is
obviously doing some low-pass filtering with its downsampling and thus
reducing the moire effects.

However, Irfanview is capable of doing very good quality downsampling if
you ask it to. First go to the Properties dialog and select the
"Fullscreen" tab. Find the option that says "Use Resample function for
display options". This tells IV to use its slower but higher quality
resampling for image display instead of nearest neighbour. And yes, it
applies to windowed-mode display as well as fullscreen, despite where
the option is located in the menus.

Then, with an image loaded, do Image->Resize/Resample, pick some
arbitrary new size, make sure the "Resample" box is checked rather than
"Resize", and select one of the resampling filters. Lanczos is the most
expensive, but does the best job of retaining image detail that can be
retained while avoiding creating new aliasing artifacts. On the other
hand, the other filters smooth the image somewhat more and might be
better in this case.

Once you've done this, you should see your display image improved at any
size below 100%. It's unfortunate that Irfanview doesn't provide a more
straightforward way of setting the filter to be used when viewing.

Dave

[Terry Pinnell]

Post by Dave Martindale

Post by Terry Pinnell
Here are a couple of screen shots showing the comparison. Both scans
were at 240 dpi. I then used Print Scrn to capture both and paste as a
new image in PSP 8. As you see, the view in IrfanView shows the
patterns, but PSP 8 does not (or vet much diminished).
http://www.terrypin.dial.pipex.com/Images/Screenshot240-IrfanView.jpg
http://www.terrypin.dial.pipex.com/Images/Screenshot240-PSP.jpg

It looks like IrfanView is using nearest-neighbour resampling to bring
the image to 50% of original size. This will generate aliasing
artifacts, and exaggerate any that are already present. PSP is
obviously doing some low-pass filtering with its downsampling and thus
reducing the moire effects.
However, Irfanview is capable of doing very good quality downsampling if
you ask it to. First go to the Properties dialog and select the
"Fullscreen" tab. Find the option that says "Use Resample function for
display options". This tells IV to use its slower but higher quality
resampling for image display instead of nearest neighbour. And yes, it
applies to windowed-mode display as well as fullscreen, despite where
the option is located in the menus.
Then, with an image loaded, do Image->Resize/Resample, pick some
arbitrary new size, make sure the "Resample" box is checked rather than
"Resize", and select one of the resampling filters. Lanczos is the most
expensive, but does the best job of retaining image detail that can be
retained while avoiding creating new aliasing artifacts. On the other
hand, the other filters smooth the image somewhat more and might be
better in this case.
Once you've done this, you should see your display image improved at any
size below 100%. It's unfortunate that Irfanview doesn't provide a more
straightforward way of setting the filter to be used when viewing.

Wayne, Dave: Excellent - thanks a lot for those detailed explanations.
Finally think I have my mind around this.

Viewed at 100%, both PSP and IV do indeed show identical views, as you
see here:
Loading Image...
Loading Image...
(In my Firefox browser, the initial display of these still shows
patterns; clicking to zoom gets full-size, and they disappear. In
MSIE6 they come up OK at once.)

Also, here's the original (1 MB).
Loading Image...

I'll experiment with IV's settings later.

Please also see my reply just now to Frank. I'd appreciate your
thoughts on my question about a definitive (but fast and simple) test
please.

[Dave Martindale]

Post by Terry Pinnell
Viewed at 100%, both PSP and IV do indeed show identical views, as you
http://www.terrypin.dial.pipex.com/Images/Screenshot100PSP8.jpg
http://www.terrypin.dial.pipex.com/Images/Screenshot100IrfanView.jpg
(In my Firefox browser, the initial display of these still shows
patterns; clicking to zoom gets full-size, and they disappear. In
MSIE6 they come up OK at once.)

If Firefox is automatically shrinking the image to make it fit the
window, you're seeing the results of the downsampling algorithm built
into Firefox, which you have no control over (other than turning it
off). IE will also downsize automatically, if you selected that option,
and again you don't have any control over the algorithm used. To avoid
these uncertainties, view at 100% size.

By the way, Photoshop is even *more* insidious at doing things behind
the scenes. If you have the "image cache" enabled (this is in PS6; it
might be called something else now), Photoshop automatically computes
smaller-sized versions of large images, and uses them for display when
you are displaying the image at some sizes below 100%. These sub-sized
images are properly filtered when created. But at some magnifications,
Photoshop uses nearest-neighbour resampling to the final size, which
is what you saw in Irfanview. So as you change magnification in
Photoshop, moire effects can get better or worse depending on exactly
what route Photoshop took to create the displayed image at that size.

For the same reason, film grain (in scanned film images) may appear to
get worse (compared to 100% view) or better at smaller sizes. Filtered
downsampling reduces grain, because grain is mostly high-frequency noise
that is reduced by filtering. Nearest-neighbour downsampling doesn't
reduce grain noise amplitude, but it makes the grain look bigger (called
grain aliasing).

Post by Terry Pinnell
Please also see my reply just now to Frank. I'd appreciate your
thoughts on my question about a definitive (but fast and simple) test
please.

Haven't seen it yet.

Dave