Scanning debate (rather long)

[Ed_L]

Hi Folks,

Once again, this concerns high-bit scanning. I have read opinions from recognized people in the digital imaging world about the advantages (or lack of) to high-bit scanning. Katrin Eismann believes there is a definite advantage to high-bit scanning (pg. 47 - 48 of her book), while Dan Margulis has different thoughts on this. Here is part of what he had to say:
_______________________________________________
"if you have a file with >8 bpc Photoshop will open it as a 16 bpc document (often called high-bit) and you can
then horse around with it to a considerable extent, although many commands are disabled. You *can* convert it
to CMYK if you like. Sooner or later you have to go into Image: Mode and change it back to 8 bpc (often called
low-bit), since otherwise you can't print it.

This is a theoretically superior method, but in practice there is considerable question as to how valid the additional
data is. I tried to test this before I wrote my last book, where a couple of pages of the results appear in print at a
size large enough to draw conclusions. I scanned a set of 10 originals on a very expensive flatbed scanner, and a
different set on a medium-priced desktop scanner that is capable of exporting a high-bit file. With my 20 high-bit
files, I made 20 copies and converted them immediately to low-bit.

With the 20 high-bit files, I really tortured them with a series of curves much more drastic than would be used in
real life. As I was doing this I was saving the changes as Actions. When I was done, I converted the high-bit files
to low-bit, as I would have to do if I wanted to print them. Then I played back the same Actions on the files that
were originally low-bit, and saved the results.

I knew that the ones that were worked on in high-bit originally would be better, but I didn't know how much. I
was hoping to find out whether there were certain categories of image where it made sense to work on them in
Photoshop in high-bit mode. High-bit files are twice as large as low-bit ones, so it's a PITA to use them.

The results surprised me. From the cheaper scanner the results were virtually identical. From the more expensive
scanner the quality improvement was there, but it was so marginal that you wouldn't see the difference on output
even on a film recorder. In print, maybe if you applied drastic curves *and* you found out later that you had to
print the file at four times the size you originally thought, it might make a noticeable difference. Or maybe, as
scanners get more accurate, the extra bits will start to make more of a difference.
So, from what I've seen so far, working in high-bit in Photoshop gets you a better-looking histogram, but not a
better-looking image."
_______________________________________________

After conducting a few comparisons of my own images in 8 bit and 16 bit scans, I have not seen a difference worth noting (not a high end scanner). When the original had reasonably decent tones, I did not see a difference in histograms after several tonal adjustments. The one time I did see a difference was when I made adjustments to one channel. The histogram for that particular channel showed gaps in the 8 bit histogram, while the 16 bit file did not. The composit image however, showed identical histograms from both files.

With the limited information I have, it is my conclusion that unless the original is lacking to the extent that I forsee a possible problem, I should scan with an output of 8 bits per channel. If I am concerned with pulling the detail in both the highlights and shadows, it might be better to scan once for the shadows, then again for the highlights. These two scans can then be combined for better overall detail.

I would like to hear from others on their experiences regarding scans that have been compared in 8 and 16 bit scans.

Ed
P.S. I think there was a typo (not mine) in the first sentence of Dan's quote.

[Doug Nelson]

I don't think that was a typo, since I read that same info some time ago when responding to a question on the Adobe forums. It was in response to a question about whether to convert from 8bit to 16bit for work. Margulis and Eismann both agree the answer is "No".

[Ed_L]

Well, that certainly would make a difference. I seems as though every time I make an assumption, I shouldn't make an assumption.

But if he said ""if you have a file with >8 bpc Photoshop will open it as a 16 bpc document", wouldn't that be an incorrect statement? Why would Photoshop open it as a 16 bpc document? He doesn't mention converting it to 16 bits. Not that I'm doubting you -- I'm just trying to get it straight in my crooked head.

Ed