Re: TIFF vs JEPG

On 13 Apr 2006 09:13:56 -0700, "Alan Meyer" <ameyer2@xxxxxxxxx> wrote:

At 100% magnification (i.e. 1:1) even a JPG image at lowest
compression (i.e. highest quality) stands out like a sore thumb when
compared to the original.

I downloaded the two images that Raphael referenced in his reply
then magnified them 16 times (i.e., 4004 pixels per dimension instead
of the 1001 pixels of the original image.)

I'm not sure I can tell them apart.

Then I magnified them almost 100 times i.e., almost 10 times
as many pixels per dimension. Now, if I looked closely, I could see
that some individual pixels had different colors. But I still couldn't
see JPEG artifact squares, and still wasn't sure which image was
which without looking at the file names.

The point is that's your personal, subjective perception. However,
objectively, there is a marked difference. Just because you yourself
can't see it or find it too minor it doesn't mean it's not there or
that it doesn't influence the workflow.

As I said last time, if you're happy with JPGs (and apparently you
are) that's great! But the problem is that's a subjective impression
which goes contrary to objective facts. But if it satisfies your
requirements, then, of course, that's all that counts.

JPG uses 8-bit precision simply because that's all today's monitors
can display. However, in not too distant future monitors will expand
this dynamic range and then a 16-bit (or higher) dynamic range will
become essential.

Leaving aside the monitors, how much precision can the
human eye distinguish? I suspect the best eyes can only do
around 10 bits, though I'm not at all sure about that.

No, it's more like 8-bits some say as low as 6-bits. *But* (and that's
a big but!) the total dynamic range of the human eye is far larger.
What this means is that even though we only have this 6-8 bit "window"
it moves through a far greater total dynamic range our eyes can see.

For example, when going from a dark area into a bright area you're
temporarily blinded and can't see any detail. But after a while as
your eyes adjust (the 6-8 bit "window" moves to highlights) you start
seeing detail. The same goes the other way around, of course.

So even though the window itself is relatively small, it covers a
large absolute area and can move according to lighting conditions. If
you only use JPGs this total or absolute area is limited to 8 bits and
that's a massive loss of information.

Furthermore, human perception is non-linear. There are some
ranges in which we are more sensitive than others. I believe
that some JPEG compression algorithms know that and take
advantage of it to produce images that are really very close to
the maximum human perception.

However I'm not an expert on this. Someone who is should post the
facts.

We have different sensitivity to different wavelength. In practice
this means we see green much better than red or blue. Specifically,
the perceptors in our eyes are broken down as follows: red = 30%,
green = 59%, blue = 11%. So that's the ratio used to calculate
luminance.

But that's not really the point. It's that data is being lost. First
by going from 16-bit (or 14 in case of some scanners, etc) to 8-bit.
And then if that weren't enough, reducing that further by applying JPG
compression. And that's not really suitable for archiving purposes.

Don't forget also that JPG compression level i.e. "quality" is not
standardized! Each software uses different metrics. You mentioned 10
as the maximum, well, the highest quality in Photoshop is 12. Both
numbers are totally meaningless because there is no reference.

I agree with part of that too. And I agree that people should do
their own tests and draw their own conclusions.

I'll go further and say that whether you are losing "massive"
amounts of data is also a subjective conclusion. There's no
doubt that a computer will find a signficant difference between
TIFF and good JPEG. But whether that's "massive" from a
human point of view is not obvious to me.

It is if you look at it in context. Not only is the difference there
but once you bring in the workflow, JPG is not (objectively speaking)
suitable for archiving.

I have no problem (and indeed do it myself!) with JPGs used "for
consumption" i.e. to distribute on DVDs or upload to a web site.

The problem is editing and archiving. Just because you find it
difficult to see the differences or don't find them objectionable the
loss of data between a JPG image and the original is massive.

Finally, I want to defend my point that good scanning is more important
than saving TIFFs.

The quality of the scanner, the decisions made by the scanning
software, the adjustments for color and contrast, the cleaning of
the image and the glass plate - all have a bigger effect on final
results than TIFF vs. good JPEG.

No, as I mentioned last time that's just factually wrong. All that
work will go to waste if your workflow is based on JPGs and/or using
scanner software editing.

For example, the editing in scanning software is very limited and only
contains a small subset of necessary tools. Any editing decisions you
make there will be based on the tiny preview "keyhole" and the 8-bit
histogram. Etc. All of that will lead to (objectivelly) massive loss
of information. Even more if you do "touch ups" later.

The "proper" workflow is to scan at maximum scanner bit depth and
native resolution without using any of the scanner software editing
features (e.g. curves etc). The only exception is ICE due to the way
it's implemented (for marketing reasons :-/).

Such a scan is known as a "raw scan" and it contains everything the
particular scanner can pull out of the media. *That's* the image which
should be archived! TIFF seems the format of choice but any *lossless*
format will do as long as one does the conversions later if needed.

One then edits this image using an external editor (much better than
the cut-down versions with limited features in scanner software). The
editing should be performed at the original bit depth and resolution.
When done, one may save that image as well for the record and to be
able to go back to it without having to edit all over again.

The last step is to then convert such an image "for consumption". This
may be for printing in which case the resolution and color information
will be reduced to match the printer. Or it may be for viewing in
which case the resolution will be reduced to fit the monitor and then
saved as JPG.

The beauty of this approach is that the original is still available.
Secondly, once the monitor size or bit-depth changes (and they will!)
or a new printer is purchased or the print fades (and it will!) you go
to your original image or the saved edited image and reduce to
accommodate these new requirements.

Don.
.

Relevant Pages

  • Re: DIP9 Default save settings
    ... JPG compression is done on colors and when you select a lower quality it is ... Did DIP9 discard pixels? ...
    (microsoft.public.windowsxp.photos)
  • Re: copy picture out of word with full resolution?
    ... No, jpg compression is progressive. ... placing the image in Word can cause damage and the act of saving the ... I paste a high resolution picture into Word. ...
    (microsoft.public.word.newusers)
  • Re: Is this rendering quality OK?
    ... while both .jpg have the same frame size 1920x ... comes from the MPEG compression and how much comes from ... That's why I asked how it looked when viewing the MPEG ...
    (rec.video.desktop)
  • Re: Distiller PDF/X1A
    ... JPG compression is a compression, side by side with an uncompressed tif, ... Now if the client wants a PDF to view on screen for approval, ...
    (comp.publish.prepress)
  • Re: Best Format After Scan
    ... Yeah, the highest quality JPG is a smaller file, yet better image, than only ... saving on the 24 bit RGB bytes, except that higher compression is possible on ... 40,453,426 Dlzw.tif LZW compression - original image ...
    (comp.periphs.scanners)