Archiving: TIFF or PSP, 16 bit or 8 bit?

  • Thread starter Thread starter Robert A
  • Start date Start date
Uni said:
I believe, humans should never be limited to what dumb computers
typically provide.
My eyesight (and/or hearing) isn't digital and never will be.
That isn't a problem even with 8-bpc computer graphics because your
analogue eyes have more SNR than is available on an 8bpc image. Most
studies have indicated that the SNR of your retina is somewhere between
6 & 7 equivalent bits. Even with the requirement to convert from a
linear encoding scheme onto a display gamma, 8-bpc is still more than
you can visibly discern.
 
: "Nevertheless, since you clearly believe that
higher resolution scans are softer and thus, by default, that lower
resolution scans are sharper, can we expect to see your Minolta film
scanner appearing on e-bay whilst you "trade up" to a sharper 300,
perhaps only 100ppi, piece of antiquity? Why don't you just go the
whole hog and flash a single photodiode at your slides to get an
ultrasharp 1x1 pixel rendition of the entire image on each slide. the
next step is just to remove the sensor completely and type a random
character into a file called "image.raw" and observe the infinite
sharpness of it all. ;-)"
---------------------------

Let's remember that images don't have resolution until they are
displayed/printed. A 3800x2500-pixel image is just a 3800x2500-pixel
image. If printed at a 12x8-inch size, it is "hi-res" (~300ppi). If it
is printed at 12x8-feet, it is "lo-res" (25ppi). When displayed on a
monitor of 1024x768-pixel dimensions, it will have more detail when
viewed at 100% (pixel-for-pixel, showing only a small portion of the
image) than at 25%, showing the full image on the screen.

As Dan Margulis points out, it is the pixels in an image that are *not*
the main subject that provide the detail in an image. For example, in
the horse-picture in the reference cited, it is the non-grass pixels
that provide detail/texture in the grassy area. If all pixels are
"grass", the texture is a uniform carpet with little detail.

There is an optimum range for print resolution: below a certain level,
the print appears "pixilated"--above a certain level, the "detail"
pixels get overwhelmed by the "main subject" pixels and disappear,
yielding an image lacking detail (i.e. "softer").

I could go on, but I suspect Kennedy knows this difference, and I don't
want to repetitive. *Large* images generally have more detail than
*small* ones, but very "hi-res* images (above 600ppi) will generally
appear softer and less detailed than their lower-res (say, 225ppi)
brothers.


Preston Earle
(e-mail address removed)
 
Let's remember that images don't have resolution until they are
displayed/printed. A 3800x2500-pixel image is just a 3800x2500-pixel
image. If printed at a 12x8-inch size, it is "hi-res" (~300ppi). If it
is printed at 12x8-feet, it is "lo-res" (25ppi). When displayed on a
monitor of 1024x768-pixel dimensions, it will have more detail when
viewed at 100% (pixel-for-pixel, showing only a small portion of the
image) than at 25%, showing the full image on the screen.

As Dan Margulis points out, it is the pixels in an image that are *not*
the main subject that provide the detail in an image. For example, in
the horse-picture in the reference cited, it is the non-grass pixels
that provide detail/texture in the grassy area. If all pixels are
"grass", the texture is a uniform carpet with little detail.

There is an optimum range for print resolution: below a certain level,
the print appears "pixilated"--above a certain level, the "detail"
pixels get overwhelmed by the "main subject" pixels and disappear,
yielding an image lacking detail (i.e. "softer").

I could go on, but I suspect Kennedy knows this difference, and I don't
want to repetitive. *Large* images generally have more detail than
*small* ones, but very "hi-res* images (above 600ppi) will generally
appear softer and less detailed than their lower-res (say, 225ppi)
brothers.


You initially presented your argument as the subjective qualities of
High Resolution Images vs Low Resolution Images (whatever that means),
but your conclusion seems rather backwards when you claim "high
resolution images are soft".

Dan Margulis was discussing undesirable results due to attempts to print
images at an EXCESSIVE RESOLUTION, greater than the specific media
screen can handle, but you failed to make that distinction. He did use
the word HIGH, but he made the context be very clear. There is a
conceptual difference between high and excessive.
 
Kennedy McEwen said:
As did many readers and writers for GIF, until Unisys decided to enforce
their IPR which they had previously been quite happy for everyone to use
openly.

I'll say it again. The PSP reading and writing code is unencumbered;
Jasc does not have a patent on LZ77 and cannot patent RLE. While its
enforcement was something of a surprise at the time, Unisys did
actually have a patent (now expired, apparently), where Jasc does not.
So it is not the same situation at all.
Patents are not the only form of IPR and Jasc certainly do own IPR in
the PSP format. What they choose to do with that in the future is
anyone's guess, just as nobody would have predicted Unisys enforcing
their IPR in the GIF format.

There is no enforceable patent etc. in the PSP format through v8. I've
read it and implemented it and my implementation is under GPL (with
acknowledgement to Adler & Gailly, of course, for zlib).

Toby
 
Preston Earle said:
: "Nevertheless, since you clearly believe that
higher resolution scans are softer and thus, by default, that lower
resolution scans are sharper, can we expect to see your Minolta film
scanner appearing on e-bay whilst you "trade up" to a sharper 300,
perhaps only 100ppi, piece of antiquity? Why don't you just go the
whole hog and flash a single photodiode at your slides to get an
ultrasharp 1x1 pixel rendition of the entire image on each slide. the
next step is just to remove the sensor completely and type a random
character into a file called "image.raw" and observe the infinite
sharpness of it all. ;-)"

On the contrary - some of the systems I design provide data to equipment
designed by colleagues which is totally dependent on resolution, yet an
actual image may never be produced at any stage of the process. At
most, all that is ever output by the system after photons enter the lens
is an angular coordinate! However none of the system would function
without resolution.

Indeed, the very measurement of resolution does not even depend on an
image being produced!
There is an optimum range for print resolution: below a certain level,
the print appears "pixilated"--above a certain level, the "detail"
pixels get overwhelmed by the "main subject" pixels and disappear,
yielding an image lacking detail (i.e. "softer").
No, and that is NOT what Dan is saying either. As I mentioned in my
previous response, Dan is specifically addressing how the image is
decimated by the rendering process. There is no resolution beyond which
an image becomes apparently softer. The absolute proof of this is to
take your best low resolution unsharpened image that you consider to be
the most sharp reproduction you can create and prop it up on a stand so
that it can be viewed next to the original scene at the same relative
scale. The original scene has effectively infinite ppi, with a
resolution limited only by your eyes. Guess which will be sharper!
I could go on, but I suspect Kennedy knows this difference, and I don't
want to repetitive. *Large* images generally have more detail than
*small* ones, but very "hi-res* images (above 600ppi) will generally
appear softer and less detailed than their lower-res (say, 225ppi)
brothers.
Sorry Earle, but that is complete rubbish. I have in front of me at
this moment, two identical images. One printed at 240ppi, selected
because it fits integrally with the resampling density of the Epson
printer I used to create it. The other is scanned and printed at
720ppi. You might want to try to explain why the latter print not only
looks sharper to the naked eye, but considerably sharper when examined
under a x4 magnifier. (If you search through the archives on this group
you will find a previous thread where I addressed this and my normal
practice of printing all of my contact sheets at this resolution
specifically for this reason.)
 
Kennedy McEwen said:
That isn't a problem even with 8-bpc computer graphics because your
analogue eyes have more SNR than is available on an 8bpc image.

There should, obviously, be a "no" after "your analogue eyes have"
there.

Work is just so enjoyable I was clearly rushing to get off there before
posting that at breakfast this morning! ;-)
 
Toby Thain said:
I'll say it again. The PSP reading and writing code is unencumbered;
Jasc does not have a patent on LZ77 and cannot patent RLE. While its
enforcement was something of a surprise at the time, Unisys did
actually have a patent (now expired, apparently), where Jasc does not.
So it is not the same situation at all.
And I'll say it again too - patents are not the only form of IPR!
 
The latest couple of versions of Photoshop have all of the processing
options for 16bpc that I believe I need. Sure, there are some plug-ins
available which automate some of those functions for pulling detail out
of the deep shadows, but the capability is there in the application
itself if you are prepared to do it. The only improvement likely in the
technology is true 16bpc processing as opposed to PS's 15-bit
approximation, but 1 bit isn't gonna make a great deal of difference.

Yes, I agree that it is unlikely that further improvements in the
capability of Photoshop will make much difference. Then again, the
head of IBM, in 159 (?) said the world would only need 5 computers. I
never say never.
 
Preston said:
: "Nevertheless, since you clearly believe that
higher resolution scans are softer and thus, by default, that lower
resolution scans are sharper, can we expect to see your Minolta film
scanner appearing on e-bay whilst you "trade up" to a sharper 300,
perhaps only 100ppi, piece of antiquity? Why don't you just go the
whole hog and flash a single photodiode at your slides to get an
ultrasharp 1x1 pixel rendition of the entire image on each slide. the
next step is just to remove the sensor completely and type a random
character into a file called "image.raw" and observe the infinite
sharpness of it all. ;-)"
---------------------------

Let's remember that images don't have resolution until they are
displayed/printed. A 3800x2500-pixel image is just a 3800x2500-pixel
image. If printed at a 12x8-inch size, it is "hi-res" (~300ppi). If it
is printed at 12x8-feet, it is "lo-res" (25ppi). When displayed on a
monitor of 1024x768-pixel dimensions, it will have more detail when
viewed at 100% (pixel-for-pixel, showing only a small portion of the
image) than at 25%, showing the full image on the screen.

I agree. A very pixelated image will show a moiré pattern when zoomed.
This does not mean a moiré pattern exists in the image.

Uni
 
And I'll say it again too - patents are not the only form of IPR!

In order that I can correct my understanding - Please be *explicit*
how Jasc can "revoke" existing free implementations of the PSP format,
without holding any relevant patents? Since I have published one, this
is of more than academic interest. What attack should I expect? Which
parts of the implementation are encumbered?

Toby
 
Toby Thain said:
In order that I can correct my understanding - Please be *explicit*
how Jasc can "revoke" existing free implementations of the PSP format,
without holding any relevant patents? Since I have published one, this
is of more than academic interest. What attack should I expect? Which
parts of the implementation are encumbered?

For starters, they could change the specs, leaving other implementations
non-functional with respect to the current version of the program.

David J. Littleboy
Tokyo, Japan
 
Toby Thain said:
In order that I can correct my understanding - Please be *explicit*
how Jasc can "revoke" existing free implementations of the PSP format,
without holding any relevant patents? Since I have published one, this
is of more than academic interest. What attack should I expect? Which
parts of the implementation are encumbered?
I don't think you should expect any attack at all, but it is niaive to
believe that you will be able to use their proprietary standard
indefinitely. That may well be the case but, since it is proprietary,
it is not guaranteed - unless you have some written agreement from Jasc
specifically stating otherwise.

How sure are you that every aspect of their format has indeed been
openly published on the reference you provided? Before answering, you
would be well advised to read again the disclaimers at the start of
their publication, which would appear to provide Jasc with specific
legal protection from anyone making claims in just such an event. Do
you know that the next version of Jasc software will not search for a
specific byte sequence which only their software has written to the
files since the format was first used?
 
: "You initially presented your argument as the
subjective qualities of High Resolution Images vs Low Resolution Images
(whatever that means), but your conclusion seems rather backwards when
you claim "high resolution images are soft".

"Dan Margulis was discussing undesirable results due to attempts to
print images at an EXCESSIVE RESOLUTION, greater than the specific media
screen can handle, but you failed to make that distinction. He did use
the word HIGH, but he made the context be very clear. There is a
conceptual difference between high and excessive."
---------------------------

Point taken. If a 3900x2600-pixel image is printed to 6"x4", it will be
at 650ppi RESOLUTION, an excessive amount for generally-used output
methods. If a 1800x1200-pixel image of the same scene is printed at
6"x4", it will be at 300ppi, generally considered hi-res but not
excessive.

In the context of Archiving, particularly when the object is making
prints of modest size (4"x6"), I thought it was important to make the
point that bigger is not necessarily better. I find it difficult not to
scan even casual images at "Maximum Resolution", even when this results
in files four times bigger than I really need and clean-up times four
times longer than for files of Optimum Resolution. I scan everything at
full frame, even though I know I won't be needing the unnecessary parts
of the image. I know this is a waste of time and resources, yet I find
it hard not to do.

Preston Earle
(e-mail address removed)
 
: "I have in front of me at this moment, two
identical images. One printed at 240ppi, selected because it fits
integrally with the resampling density of the Epson printer I used to
create it. The other is scanned and printed at 720ppi. You might want
to try to explain why the latter print not only looks sharper to the
naked eye, but considerably sharper when examined under a x4 magnifier.
(If you search through the archives on this group you will find a
previous thread where I addressed this and my normal practice of
printing all of my contact sheets at this resolution specifically for
this reason.)"
----------------------------

Well, are they "identical" or not? If identical, how do you tell them
apart? Do you mean they are images of the same scene or scans of the
same piece of film? It would be hard to get two "identical images"
without duplicating one to make the other.

If you are talking about duplicated images from the same file, one
printed at one print-resolution and one at another, I'm not sure what
this shows. There is an optimum pixels-per-output-dot range, usually 1.5
to 2, for conventional and stochastic offset screens, and I haven't seen
any discussion that this is not also true for ink-jet printers. If not
duplicate files, then scanner software and image-manipulation software
can introduce all sorts of unknown manipulations to the file data. Print
drivers can also do "funky" things to particular images. And some image
issues are subject-related, moiré being the first to come to mind.

If you believe that more pixels-per-output-dot is always better than
fewer, I guess we disagree. If I'm misunderstanding your points, perhaps
you need to explain them in words of fewer syllables for this simple
printer. (I'm trying to be cute, not making light of you or your
position. I apologize if this reads the wrong way.)

Preston Earle
(e-mail address removed)
 
In the context of Archiving, particularly when the object is making
prints of modest size (4"x6"), I thought it was important to make the
point that bigger is not necessarily better. I find it difficult not to
scan even casual images at "Maximum Resolution", even when this results
in files four times bigger than I really need and clean-up times four
times longer than for files of Optimum Resolution. I scan everything at
full frame, even though I know I won't be needing the unnecessary parts
of the image. I know this is a waste of time and resources, yet I find
it hard not to do.


I think it is the typical behavior of human male animals <g>

Margulis was specifically discussing 150 lpi prepress screens, and he
begins that chapter saying "it's silly to assume greater resolution is
better".

Prepress conventional wisdom says dpi should be in the range of 1.4 to
2.0 times lpi. These are intended to be limits, but we can find some
writers calling this multipler to be a "quality factor". We all assume
2.0 must be better than 1.4. But Margulis is saying and showing in his
printed book that the low end may give sharper final images.

(note that PDF files can only show print-size images on the video screen
resampled to about 1/4 size, and the RGB JPG images in this file are not
even screened. We need the printed book to see what he shows. It is
about printing).

But the automatic thinking of most males is that if the maximum of 2.0 is
good, think how great 4.0 must be. <g> We males dont always grasp the
concept of appropriate, and we understand maximum as a goal to be
exceeded. <g>
 
Preston said:
: "I have in front of me at this moment, two
identical images. One printed at 240ppi, selected because it fits
integrally with the resampling density of the Epson printer I used to
create it. The other is scanned and printed at 720ppi. You might want
to try to explain why the latter print not only looks sharper to the
naked eye, but considerably sharper when examined under a x4 magnifier.
(If you search through the archives on this group you will find a
previous thread where I addressed this and my normal practice of
printing all of my contact sheets at this resolution specifically for
this reason.)"
----------------------------

Well, are they "identical" or not? If identical, how do you tell them
apart? Do you mean they are images of the same scene or scans of the
same piece of film? It would be hard to get two "identical images"
without duplicating one to make the other.

I think I explained the situation quite clearly, but obviously not
clearly enough. The images are "as identical" as the examples you cited
in Margulis' text.

These images are contact sheets - small thumbnail images of negative
scans which are roughly 1.25x the actual size of the negatives
themselves. They are stored interleaved with my negatives in archive
files for ease of identification of negatives and search purposes. Of
the first several pages I have two sets. The initial set made by
printing the composite sheets of 4000ppi scans at exactly 240ppi on the
page. The second sheet, printed at 720ppi on the page. This latter
format I now use as a standard because the images are *much* sharper
under normal viewing conditions.
If you are talking about duplicated images from the same file, one
printed at one print-resolution and one at another, I'm not sure what
this shows.

What it shows is exactly the topic of this sub-thread - the statement
that higher resolution results in softer images - is simply not true.
There is an optimum pixels-per-output-dot range, usually 1.5
to 2, for conventional and stochastic offset screens, and I haven't seen
any discussion that this is not also true for ink-jet printers. If not
duplicate files, then scanner software and image-manipulation software
can introduce all sorts of unknown manipulations to the file data. Print
drivers can also do "funky" things to particular images. And some image
issues are subject-related, moiré being the first to come to mind.
As I wrote previously, one image is printed at 240ppi *on the page* the
other at 720ppi *on the page* both sizes selected to specifically match
the native resampling pf the Epson desktop printer range. There is *no*
issue here of the driver doing "funky things to particular images", the
images were resampled using known algorithms. In addition, being
several contact sheets, there are approximately 100 images in the
selection - not massive, but significant enough quantity to demonstrate
that this is the norm.
If you believe that more pixels-per-output-dot is always better than
fewer, I guess we disagree. If I'm misunderstanding your points, perhaps
you need to explain them in words of fewer syllables for this simple
printer. (I'm trying to be cute, not making light of you or your
position. I apologize if this reads the wrong way.)
Well, by way of explanation, you partially address the issue in your
paragraph above. Epson desktop photo printers all resample to a native
resolution of 720ppi before the stochastic dot rendering process, which
can be between 1440x720dpi to 5760x1440dpi depending on the printer.
Since all such cases are well above your nominal 2 output dots per pixel
it is clear that all of the Epson range of printers (the very wide
format range resamples to 360ppi) operate perfectly well in the region
where it is not possible to even encounter the decimation artefacts that
Margulis is referring to. Consequently the situation simply never
arises where pixel decimation results in higher resolution images being
printed with softer results than lower resolution equivalents.
 
Kennedy McEwen said:
I don't think you should expect any attack at all, but it is niaive to
believe that you will be able to use their proprietary standard
indefinitely. That may well be the case but, since it is proprietary,
it is not guaranteed - unless you have some written agreement from Jasc
specifically stating otherwise.

How sure are you that every aspect of their format has indeed been
openly published on the reference you provided? Before answering, you
would be well advised to read again the disclaimers at the start of
their publication, which would appear to provide Jasc with specific
legal protection from anyone making claims in just such an event. Do
you know that the next version of Jasc software will not search for a
specific byte sequence which only their software has written to the
files since the format was first used?

I think it is obvious that in *future* versions they reserve the right
to make an incompatible spec and/or not make it public (as Adobe has
withdrawn theirs), but I am only discussing the versions through v8
which are published.

Since the version 8 application is also released, it is easily
confirmed that no such retrospective "revocation" of interoperability
has already occurred. Likewise, I still insist that they have no legal
power over *existing* 3rd party implementations.

To date Jasc shows more common sense than Adobe, by choosing to
encourage 3rd party developers and opening their specifications.

Toby
 
Wayne Fulton posted "...
Prepress conventional wisdom says dpi should be in the range of 1.4 to 2.0 times lpi.
These are intended to be limits, but we can find some writers calling this multipler to be
a "quality factor". We all assume 2.0 must be better than 1.4. But Margulis is saying and
showing in his printed book that the low end may give sharper final images.
...."

That discounts the cases where the scan shows a moiré pattern (as in textured cloth), or
fine lines that have "the jaggies."

In a simple scanning calculator I wrote in October 2001, I added the following note:

Note 1: Recommended values for the Quality Factor are between 1.5 and 2.0. Usable scans
can be obtained with values over the range of roughly 1.0 to 3.5. Values as high as 6.0
can be necessary when the original has very fine lines or a texture that causes a moiré
pattern. See below for additional information.

The above was partly based on the following text on the subject that was copied from 'The
Format' newsletter #37

There was a good discussion of the merits and otherwise of scanning resolution "rules of
thumb" on the QuarkXPress list recently. There were the usual comments about scanning at
various mathematical formulae ranging from pixels-per-inch numbers of from 1.2 to 2 times
the eventual printing resolution in lines per inch.

One writer commented that while these are good starting points, those producing scans need
to look at what they are scanning: On a scan of an automobile at 2x the lines around the
doors may look rastered. The reason? They are made up by a 1 pixel line. Rescan the same
original at 4x or 6x lpi and it looks much better. No jaggies. Areas of solid color don't
look any better; but fine lines, fine detail, require more resolution to avoid rastering.
The writer added: "The formula can't anticipate the subject matter, only the scanner
operator can."

On the other hand a picture of clouds would show that 2x, or even 1.44x can be overkill on
a soft image.

So 2x or whatever as the working formula should be a starting point. For soft images, you
can drop down, but images with fine lines may need to be increased provided it doesn't
produce an impossibly large file.

It is also necessary to relate the original size to the final use size.

Text copied from:
FORMAT newsletter No. 37, March 2000
E-mail: (e-mail address removed)

HTH
 
RSD99 said:
Wayne Fulton posted "...
Prepress conventional wisdom says dpi should be in the range of 1.4 to 2.0 times lpi.
These are intended to be limits, but we can find some writers calling this multipler to be
a "quality factor". We all assume 2.0 must be better than 1.4. But Margulis is saying and
showing in his printed book that the low end may give sharper final images.
..."

That discounts the cases where the scan shows a moiré pattern (as in textured cloth), or
fine lines that have "the jaggies."
[snip]

One also has to consider the problem of moiré between the image (e.g.
fabric pattern) and the halftone dot screen itself - the only solution
to which is to use a stochastic or screenless halftone. (The other
step usually necessary to reduce moiré is to scan at a higher
resolution.)

T
 
Back
Top