Sharpening scans for internet printing

[Michael Wasley]

I have asked this elsewhere and got no response. I have emailed some
of the companies involved with no clear answer.

If I scan a transparency and then send the scan to be printed by a
web-based lab what do I do about sharpening?

Normally for on screen viewing of an image of about 600 px long side I
find USM is typically around 150%, radius 0.7, TV 3. Obviously it
varies.

But for a 6x4 print from a digital lab the long side is going to be
1200+ px.

How do I judge the result of USM in that case?

1) Work at full size on the screen?

2) Zoom out to 25% or whatever produces an on-screen image about the
size of the print?

3) Produce a trial version that at 100% will be about the right size
on screen (say about 500 px long)?

4) Something else?

Two other questions.

Would people recommend reducing the pic to the dimensions recommended
by the lab for good results (eg 1200 px for 6"), or send the full size
scan, which may be about 1900 px long?

Any thoughts on whether the JPEG being sent should be uncompressed or,
if compression is acceptable, compressed to what extent?

Thanks,

Michael

 

[Wayne Fulton]

I have asked this elsewhere and got no response. I have emailed some
of the companies involved with no clear answer.

If I scan a transparency and then send the scan to be printed by a
web-based lab what do I do about sharpening?

Normally for on screen viewing of an image of about 600 px long side I
find USM is typically around 150%, radius 0.7, TV 3. Obviously it
varies.

But for a 6x4 print from a digital lab the long side is going to be
1200+ px.

How do I judge the result of USM in that case?

1) Work at full size on the screen?

2) Zoom out to 25% or whatever produces an on-screen image about the
size of the print?

3) Produce a trial version that at 100% will be about the right size
on screen (say about 500 px long)?

4) Something else?

Two other questions.

Would people recommend reducing the pic to the dimensions recommended
by the lab for good results (eg 1200 px for 6"), or send the full size
scan, which may be about 1900 px long?

Any thoughts on whether the JPEG being sent should be uncompressed or,
if compression is acceptable, compressed to what extent?

JPG files are compressed, by definition of JPG. Your choices when
saving the JPG file are a higher quality setting (a larger file, less
compression), or a lower quality setting (a smaller file, more
compression). You want to be closer to high quality, maybe not 100,
but 90, or at least 80, all the quality you can stand the file size of.
(I am assuming the setting is labeled Quality, instead of labeled
Compression, which goes the opposite way).

If you have it, I'd keep the full 1900 pixels for a 6 inch print. The
idea is to print them closer to 300 dpi, meaning 300 pixels per inch of
print, meaning 1800 pixels for 6 inch dimension. 10% more or less
than this is no big deal. They will tell you they accept smaller
images, because they dont want to turn away people with smaller images.
You will have to be well under 100 dpi to be turned away, but if you can
give them 300 pixels per inch (300 dpi size), then do. If you give them
1200 pixels for a 6 inch print, it can only print at 200 dpi, and 300
may look better.

When asking this, be sure they know you already have larger images, and
are asking about optimum size for very best quality. They will likely
tell you 300 dpi then if they know it is not a problem for you. But they
will likely have different smaller numbers for their prints larger than
8x10 inches, they may not be able to handle such a large image, and it
is less important in a large print.

Sharpening with Radius of 0.7 is good for video screen size viewing, but
a larger Radius of say 1.5 is good for printing at 300 dpi. You cant
really judge it the same way on the screen before printing. Large
radius can reduce the smallest detail at actual screen size, but the
large print image needs more radius to show up. Maybe judge sharpening
at zoomed out reduced size, more the size it will print, and this also
effectively reduces the large Radius you see too, but it remains larger
in the full size data (however, you can see other artifacts when
zooming, due to the screens sloppy resampling). Dont overdo the
sharpening, and judge it again when back from the printer, for next
time. You could test once by including 2 or 3 of the same image with
different radius, marked subtilely in a few image pixels so you know
which is which in the print batch you get back.

 

[noisy]

JPG files are compressed, by definition of JPG. Your choices when
saving the JPG file are a higher quality setting (a larger file, less
compression), or a lower quality setting (a smaller file, more
compression). You want to be closer to high quality, maybe not 100,
but 90, or at least 80, all the quality you can stand the file size of.
(I am assuming the setting is labeled Quality, instead of labeled
Compression, which goes the opposite way).

If you have it, I'd keep the full 1900 pixels for a 6 inch print. The
idea is to print them closer to 300 dpi, meaning 300 pixels per inch of
print, meaning 1800 pixels for 6 inch dimension. 10% more or less
than this is no big deal. They will tell you they accept smaller
images, because they dont want to turn away people with smaller images.
You will have to be well under 100 dpi to be turned away, but if you can
give them 300 pixels per inch (300 dpi size), then do. If you give them
1200 pixels for a 6 inch print, it can only print at 200 dpi, and 300
may look better.

When asking this, be sure they know you already have larger images, and
are asking about optimum size for very best quality. They will likely
tell you 300 dpi then if they know it is not a problem for you. But they
will likely have different smaller numbers for their prints larger than
8x10 inches, they may not be able to handle such a large image, and it
is less important in a large print.

Sharpening with Radius of 0.7 is good for video screen size viewing, but
a larger Radius of say 1.5 is good for printing at 300 dpi. You cant
really judge it the same way on the screen before printing. Large
radius can reduce the smallest detail at actual screen size, but the
large print image needs more radius to show up. Maybe judge sharpening
at zoomed out reduced size, more the size it will print, and this also
effectively reduces the large Radius you see too, but it remains larger
in the full size data (however, you can see other artifacts when
zooming, due to the screens sloppy resampling). Dont overdo the
sharpening, and judge it again when back from the printer, for next
time. You could test once by including 2 or 3 of the same image with
different radius, marked subtilely in a few image pixels so you know
which is which in the print batch you get back.

USM sharpening with a big radius like 1.5 can result in pixel noise.
Reducing the noise after sharpening with Gaussian Blur will decrease the
sharpness.

 

[nikita]

As an addition to Waynes excelent info, if you mean sending to a lab
that uses a *continuous tone color* system like Durst Lambda,
Lightjet, minilab or a smaller Pictrograph;

Then the sharpening is something else than rasterized outputs like a
press or what we traditionally target for sharpening. These cont.
printers needs very little sharpening. Many labs uses their own
sharpening within the machine or workflow that doesn't relates to what
we are used to by the USM values. Most of them use a very low setting.

The first you should ask them is what is the maximum quality
pixeldimension for their machine. It can be a 254 ppi at the physical
size on paper when recounted to a pixel per inch formula. Or 400 ppi.
When you know that value you can target to that and the physical size.
Then sharpen slightly and Save out a copy as a Jpeg Max quality.

By doing so you're avoiding several things. Many online webbased
systems are making a jpeging at a much harder jpeg compression if they
get a tiff loaded into the sendprogram in your own computer before
uploading to them. That program will always resave a Jpeg before
sending. If the program get a very highrez file it will also make a
downrez to a specific limitvalue before sending from your
place....then when it arrives to their site it may target *one more
time* to the endvalue for their preferable pixeldimension. In other
words two downsamplings after each other – and your own sharpening
sits far away in the beginning of the flow. This is if you're sending
a very highrezfile.

If the program gets a max quality file regarding pixel-x-pixel or the
equal ppi-x physical size for their machine, and also "feels" that it
gets feeded with a jpeg file it will let it alone and just brings it
into the flow as is for printing. So, YOU decide the compression, not
the sendingprogram.

This may vary from place to place but it's a common concept used by
these online printingsevices and it's hidden if not asking the right
individuals about it.

The sharpening I use for our own Pictrograph as a standard sharpening
bump is 200/0,5/0. I've done a lot of tests with different amounts and
radius but when using a 200 ppi file this has been the perfect value
for general output sharpening. We have been using the same value for a
254 ppi onlinelab and it worked really good. But as they applied a
slight sharpening on their own we could use 150 % with 0,5. The worst
effect had the higher radius settings in this continues tone systems.
As for sharpening, it will always depend on the file, how it has been
*pre-sharpened* in a digcam or cropped etc. But for a general
sharpening with these machines – keep it low.

Make some trial versions at max quality ppi and max quality jpeg. Send
them and look at the difference between a 150% 200% 300% usm at the
same radius around 0.5-0.7. (make some selections and sharpen
differently on the same print) It will show you very little
difference. Increase the radius and you will find that it very soon
will mess up things.

When it comes to using a lower rez than the "max quality
pixeldimensions", it may work extremely well when doing larger prints.
When doing a 50x70 cm you can let the ppi fall down to 160 ppi and
still have a very good print. This is because the larger viewing
distance for such a large print on a wall. Actually it's quite amazing
to see what 160 ppi can produce under certain circumstances. That's
also what the lower limitvalue they tell you is. It's good to know if
not having a 4000-5400 ppi scanner....ask the onlinelab where the
lowerlimit is.

nikita

 

[Wayne Fulton]

The sharpening I use for our own Pictrograph as a standard sharpening
bump is 200/0,5/0. I've done a lot of tests with different amounts and
radius but when using a 200 ppi file this has been the perfect value
for general output sharpening.

If you are going to drop back to 200 dpi, then yeah, I too would drop
the Radius back to 1.0 to compensate for the larger pixels. And since
Radius and Amount work together, if you're going to use 200%A, then the
Radius should drop back even more to compensate. I wasnt considering
200 dpi and 200%A, but at 300 dpi and judged on paper, I think 100%A and
1.5R are fine values (moderate values). Since A/R do interact, there's
not so much difference between more of one and less of the other, but on
paper, the smaller pixels can use more R and less A. Video is generally
the opposite.

 

[Michael Wasley]

Many thanks to everyone for very helpful responses.

Of three companies I asked for advice I now know that two use Fuji
Frontier, which I believe is 300 dpi. Ofthese two one said don't
sharpen because the machine does it, but results suggested this was
not enough. The other, very honestly, said they didn't really know.

The third company (Ofoto UK, so do they use Kodak related kit?) didn't
even understand the question.

Other places also drew a blank, hence the post here.

I will experiment.

Apologies for loose use of 'compression' in OP. I really meant amount
of compression, or, as Wayne pointed out, I would have done better to
talk about quality.

Michael

 

[Wayne Fulton]

The third company (Ofoto UK, so do they use Kodak related kit?) didn't
even understand the question.

Other places also drew a blank, hence the post here.

I'm sure it depends on who you speak to there, some staff know, some dont.

I inquired about "optimum" of Ofoto US a couple of years ago, and got this:

Thank you for contacting the Ofoto Customer Service Team.

Our native pixel dimensions for printing are as follows:

8 x 10 = 2400 x 3000
5 x 7 = 1500 x 2100
4 x 6 = 1200 x 1800 this is 300 dpi of course)

The optimal resolution for our larger format prints are:

16 x 20 = 2140x1800 (This isnt, even smaller than 8x10 pixels)
20 x 30 = 2140x1800

Although Ofoto will print any of your photos in the size you choose,
some photos - depending on the number of pixels in the image - look
better when printed at a specific size. Use the following guidelines as
a general rule.

Minimum recommended image resolutions for printing:

4 x 6 Print / 640x480 pixels minimum
5 x 7 Print / 1024x768 pixels minimum
8 x 10 Print / 1536x1024 pixels minimum
16 x 20 Print / 1600 x 1200 pixels minimum
20 x 30 Print / 1600 x 1200 pixels minimum
Wallet Prints / 320x240 pixels minimum (Actual image size: 2" x 3")

Our printers print at a maximum of 300dpi.

It is not recommended to interpolate images up to meet these numbers.

In addition, cropping the images to the correct aspect ratios will
prevent trimming of the contents.

For a 4 x 6 print: Your image should have an aspect ratio of 1:1.5
For a 5 x 7 prints: Your image should have an aspect ratio of 1:1.4
For an 8 x 10 print: Your image should have an aspect ratio of 1:1.25
For a 16 x 20 print: Your image should have an aspect ratio of 1:1.25
For a 20 x 30 print: Your image should have an aspect ratio of 1:1.5
For a Wallet-size print: Your image should have an aspect ratio of 1:1.5

 

[nikita]

Wayne,

thanks a lot for the information. The Lab values for pixels etc. are
very good info.

However, even if Amount & radius works in an ballance act it very
much depends on the complete flow. From source to destination. The
classic formulas for sharpening is a guideline, but as many times the
file is presharpened in different ways with unknown mixes and
technics, the radius can be tuned standing alone. These machines are
very sensitive to higher radius values. The sharpening we're used to
do for inkjets or presses according to known formulas doesn't apply.
In my own tests fr the Pictro the Radius is what seems to be a real
screw up thing. Keep it very low with these kind of machines. Sure, if
using 400 ppi values are increased even here.....saying something else
would be stupid.

If we're using a new thinking like the Bruce Frasers concept that
reminds of a modern colormanagment flow things gets *special* even
more. That is; slight presharpening for sourcedevice losses only (and
archive that one) and further sharpening at different stages along the
path to the destinationdevice, step by step for each use including
selective sharpenings. The same goes for different kind of maskbased
sharpeningstechnics or plugins. We're then off the traditional path
and the only thing that counts is what the outputdevice needs.

As for rez like 300, 254 or 200 it can as well be an
*in-machine*-at-the-lab sharpening that works very differently from
the USM thinking, that effects the flow and is tuned to different
machines rez. It very much depends on who set up the machine.....and
their "taste" ragarding what is sharp enough based on the big mass of
customers materials. So, the sharpening at customerstage would very
much be for the losses the capturedevice (scanner, digicam or
whatever) is responsible for, if a good sharpening in the machine is
done. So, even at this stage no standard rules except from what visual
tests or experiences fed you with. The eye is the only real judge.

200 ppi for the Pictro is enough for any photograpic picture that
doesn't hold any textlayout. Even the most experinced geeks and
Pictrofanatics agree on that. It also has an internal memorybuffer
that can get problem when fed with larger pics at a 400 ppi. It simply
goes beyond the built in memory and the memorymodules / add-on stuff
is more expensive than interesting.

The bottomline when sending in files – if we want complete control –
is to gather as much info as possible for the labs maximum quality
needs and how low we can go in the other direction. How to avoid the
flow to the lab that changes our controlled parameters. Then it
doesn't matter that much which one is highest in rez or so. Make some
tests and stick to a good choice.

Another very good way of judging a lab is to ask them for an ICC
profile that describes their machine. Some labs expect to get a
Colormatch/Adobe RGB or s-RGB from you. OR they want you to make the
conversion TO their ICC profile at your place before sending. If they
do *their own conversion* at the lab from a normal workingspace like
Adobe RGB, you could still ask for the profile for softproof use. The
best thing is if *they* do the conversion as they can update their
profile any time. But if you always check if a new version is out for
download even such a flow when you do the conversion at home would
work very well. Anyway, they must be able to answer what they want
sent to them colorwise. A wellprofiled machine in this field of
outputs can deliver amazingly clean prints that matches a
wellcalibrated/profiled monitor. How the lab meet you when asking
about colormanaging stuff is important. many lab people are coming
from the traditional filmexposed>wetpaper field and lacks
understanding for the digital capture part. But the most important
thing is that they holds a STABLE wet flow. Then it's always possible
to build your own profile, buying this elsewhere or get it made for
free. It's good if many asks these questions about colors. That will
push things faster.....

Again, Wayne, that info from the lab you posted is exacly what a lab
should be able to give us. Thanks for posting this as an example.
Really good.

nikita