Sharpness in Tiff Scans

[1Scan]

I have done a number of scans of 35mm and m/f frames, using a Nikon
Coolscan and an Epson 4990. On the Nikon I've used their scanning
software, on the Epson SilverFast.

Each image has been scanned twice. First 2000 dpi jpg; second 4000 dpi
tiff. Apart from that all other settings were the same.

I am puzzled because in every case on enlargement the tiffs look
markedly inferior to the jpgs which are sharper. I had expected the
opposite to be the case.

Can anyone throw light on why this might be, and more importantly
suggest a solution? All suggestions gratefully received.

Jeff Underwood

 

[Scott W]

I have done a number of scans of 35mm and m/f frames, using a Nikon
Coolscan and an Epson 4990. On the Nikon I've used their scanning
software, on the Epson SilverFast.

Each image has been scanned twice. First 2000 dpi jpg; second 4000 dpi
tiff. Apart from that all other settings were the same.

I am puzzled because in every case on enlargement the tiffs look
markedly inferior to the jpgs which are sharper. I had expected the
opposite to be the case.

Can anyone throw light on why this might be, and more importantly
suggest a solution? All suggestions gratefully received.

Jeff Underwood

An image scanned at 2000 ppi will look sharper then on scanned at 4000
ppi, when viewed at 100%.
Try resizing the 4000 ppi scan to 50%, to match the 2000 ppi scan and
see how it compares.

As Barry said there really should not be a differents in sharpness
between a tiff and jpeg, unless you use a lot of compression on the
jpeg.

Scott

 

[Charlie Hoffpauir]

This goes to the heart of questions I had for a long time and had
hoped been solved. But I'm not sure. I am scanning/achiving family
slides with a Nikon Coolscan for my children and their children's
children. I'm not concerned how they look today but in a 100 years
with the technology of that time.

My approach (supported by the opinions of others) has been to scan at
the greatest resolution technology provides, and this has been 4000
dpi and 16 bit color. Of course, this generates 100MB image files,
but that's just the cost of this approach.

As much as possible, I avoid "over-processing" the image by not using
algorithms for sharpening and noise reduction. All I do is balance
faded colors, use dust removal, and crop the image on the philosophy
that it's better to give future generations the raw product to enhance
with their technology.

I wonder though. I too am disappointed at those high res images.
They're just not as sharp as I had hoped, and definitely not as sharp
as the 35mm slide projected on an 8 by 5 screen. Yes, I could sharpen
the image, be that introduces artifacts that could defeat the
algorithms of future genertations.

Perhaps I'm getting all the resolution there is. It's a lot of work
for something that's not a good as I would have liked.

Anyone have thoughts on the subect?

Mike

Everyone has an opinion, here's mine. (Actually, I have two.)

1. I've seen very few 35mm slides or negatives that are good enough
for a 4000 dpi scan. Certainly no "hand-held" 35mm images, as most of
mine are.
2. If your scans are really not as sharp as a projected image, then
you are really doing something wrong, or your equipment is really
screwed up. Try scanning your "best" slide on someone else's scanner,
or better yet, have someone else scan it for you, just as a
comparison.

 

[OldMaven]

You're likely getting all the film has to offer, as others have noted.
That's not to say that future technology might not be able to improve on
hi-res image appearance, so your approach of capturing all you can now
seems like a good one for you so long as large storage is not a problem.

What you might do as an interim step is to create smaller jpeg files to
pair with your larger tiffs -- the smaller file reduced by a factor of
four or more -- then sharpen, color-adjust and otherwise edit images to
look good today. Your large tiff remains unaltered for the future.

Note that some photo editing programs, such as Adobe Photoshop
Lightroom and Apple's Aperture edit nondestructively. That is, they
save the original plus an instruction set from which they can recreate
your edits when you view or print the images from within the program.
(This is similar to the image and metadata from a camera's RAW file.)
They can also save a copy with edits incorproated as a JPEG, TIFF, etc.

 

[Charlie Hoffpauir]

Sorry, I forgot to mention this in my previous post. I agree
with the others who said that 3000 dpi is really all you need. If you
scan a 35mm slide at around 2800 dpi, that will allow you to print out
an 8x10 picture. Scanning at a higher dpi than that would be good if
you wanted to print out something larger than an 8x10, but that is
rarely the case. I think that you'd just be wasting storage space if
you scanned at anything over 3000dpi.
Think of it like this....if you scanned them at 4000dpi, and then
decided to make an 8x10 printout of one of them, you would resize the
image to be 8x10 at 300dpi to go to the printer. Most all printers
require the image to be 300 dpi when it's sent to it. If you resized
the 4000dpi image to an 8x10 image at 300 dpi, you'd be throwing away
a lot of info when you resized it. At 3000dpi, you'd still be
throwing away data, but it wouldn't be that much. Why save the image
at 4000dpi if you're going to throw away anything over 2800 dpi when
you go to print it?

Talker

Your logic makes good sense, unless you need to crop the image before
printing, as is the case for most images. That being so most of the
time, your logic should be to go ahead and scan at the maximum
resolution possible on the scanner.

However, I look at it from another point of view... from what is
available IN the image that you're going to scan. For almost all 35mm
slides, camera shake, slight mis-focus, limitations of the lens, etc
limits the quality of the image more than does the dpi of the scan.
That is, there simply isn't 4000 dpi worth of information in most 35
mm slide images. Looking at it from the point of view of the image, if
you scan a soft image at 4000 dpi, you won't get any better data than
scanning the soft image at, say, 2800 dpi. Try it on a slide, looking
critically at blown up prints scanned both ways (crop a small portion
fo the image and print to the same physical size image on paper). Then
if your 4000 dpi scan gives a better image, you need to scan at 4000
dpi. But if (as in most cases) you can't tell the difference, then
you're wasting a lot of time and storage space on those 4000 dpi
scans.

 

[MoiMoi]

Does 16 bit color provide real value in my scheme?

IMO, no.
Histograms prove that certain editing functions on 16 bit files are
better.
Eyeballs prove that it doesn't matter.

MM

 

[Toni Nikkanen]

I try and start a discussion about archiving for 100 years in the
future about once a year. ....

Sigh--Not yet! Opinions are still firmly on all sides of the issues.

Expect the opinions to become closer to each other in about 100 years.

 

[MoiMoi]

Expect the opinions to become closer to each other in about 100 years.

I can hardly wait.

MM