8 bit vs. 16 bit files for LF scans?

greenspun.com : LUSENET : Large format photography : One Thread

I have struggled getting any good responses to this question... this forum seems to have some of the advanced minds in the photo world.. I am very interested in hearing any opinions on this....

Scanning 8x10 film creates huge files and many storage issues. I am trying to estimate the max. amount of data a "high end" digital printer can utilize to benefit the final print. Knowing this, it will prevent me from storing files excessively large with no current or possibly future benefit. (This relates to saving 8 bit files vs. 16 bit files only, not the size of the file determined by output dimensions)

I was curious if anyone has ever ran the following test on a "high end" digital printer. (Like the LJ 5000 or other very high end ink jet, such as 8 or 12 color printers)

Scan a perfect chrome in 16 bit, save file, then print the 16 bit file on "high end" digital printer.

Take the same file and open in Photoshop and use one of the tools, therefore the file would be compressed into an 8 bit file when saved...print this 8 bit file on the same printer and compare the results. (Use Ektaspace for your work space to eliminate data clipping)

Does anyone think (or know) the 16 bit file would create a better looking print? Or is it possible, even todays best printers are so inferior to this massive amount of data (8 or 16 bit file size), that the additional data 16 bit provides is completely wasted?

If there is no perceived difference in these prints using todays best printers, I wonder if it's possible to try the same experiment comparing 4 bit and 8 bit files on a print. The benefit of knowing this information, would be to determine the optimum file size to work with (8bit vs. 16 bit) then we can make an educated guess on what size file will be obsolete or inferior in the near future.

Some peoples opinion is to always save the largest file size possible to accomodate the improved printers of the future. Of course this appeals to my common sense, however if current "high end printers" make equal prints between 4 bit and 8 bit files, I can see NO reason to save 16 bit data now. It seems 8 bit files would be very sufficient to accomodate the improved printers of the future.

It's my gut feeling, in the next 10 - 20 years, there will always be limitations of printing papers. Considering todays digital printers already match darkroom quality, I doubt manufacturers will invest heavily into R&D to further the resolution of the papers and ink delivery systems. This seems logical considering the current "high end" digital printers output is nearing the capacity of the human eye.

Of course this discussions excludes saving files for other purposes other than printing to paper, such as film recorders, which have the capacity to record 10x - 20x more data than printing papers. Thank you all in advance.

-- Bill Glickman (bglick@pclv.com), January 20, 2001

Answers

Bill:

The real issues are whether printers event accept 16 bit data (I don't think the LJ does) and whether the human eye can discriminate that many tonal levels. As I understand it, the main advantage to 16 bit scanning is the availability of the extra dynamic range while making digital adjustments. This prevents "roundoff" posterization of smooth tonal variations and preserves enough information to make full use of the 8 bits/pixel/color going to the printer.

-- Glenn Kroeger (gkroeger@trinity.edu), January 20, 2001.


The Lightjet does not accept 48 bit (16 bits per channel) files. I am not aware of any printer that does. I doubt that most human eyes can discern the difference between a GOOD, CLEAN, 24 bit file and a 48 bit file. As Glenn said, the advantage of 48 bit files is when you're doing a lot of editing; even in those situations, however, I wonder how much of a visible difference there is? Having considered the same question you pose, Bill, my solution has been twofold: (1) get the highest quality 24 bit files I can and when I make adjustments to them I always do this in a Photoshop layers adjustment so that all the adjustments are combined at the end and applied to the file only once rather than degenerating the original data repeatedly and creating multiple rounding errors, and (2) I archivally store and freeze all of my originals so that if it ever turns out that I've miscalculated about the value of 48 bit files, I can go back and have them rescanned and the colors won't have faded too much. The other problem with working always in 48 bit files is the amount of RAM necessary. I currently have 1.1 Gigabits of RAM on a Mac G4. This is fine for 160 MB files with 24 bits of color. (Large enough to print a 20" x 30" image at about 305 dpi, or a 30" x 40" at 200 dpi.) If I wanted 48 bits of color than that would require files of 320 MB each. Photoshop will currently only permit an allocation of 999 MB of RAM (actualy OS 9, not Photohsop imposes this limitation). That's not enough RAM, IMHO, for efficiently working on a file in Photoshop that's 320 MB.

-- Howard Slavitt (info@enaturephoto.com), January 20, 2001.

Hey guys some great input... so I guess I can narrow down my question to seeing if there a difference between final output between 4 bit files vs. 8 bit files. Although I would never use 4 bit files, not even sure if I can make them, I would like to see if on a final print there is any difference between 4 and 8 bit files. This would give us some indication of wheter 8 bit files is overkill or maybe on the edge of current printing technologies. Any input on this logic? Thanks again!

-- Bill Glickman (bglick@pclv.com), January 20, 2001.

Another consideration to the scan once-use many philosophy. Some authors suggest "down-resing" in steps: says 50 dpi at a time, rather than taking a 600 dpi scanned file and then down sampling to 300 dpi. As a practical matter, I have decided to scan as a function of my scanner power, which is 1200 x 2400. I make a LO scan at 1200 and a HI scan at 1800-and that's it. If the largest print will be fairly small, then I may scan at 600. Have fun. The time issue isn't scanning adjustments but the careful investment is getting the scanner glass and negative as clean as possible. Every once in a while do a scan of the bare scanner bed-good way to see how really clean the

-- David Stein (DFStein@aol.com), January 20, 2001.

4 bits/color will be very evident in tonal posterization... most computer graphics cards can operate at 16-bit color which uses 5 bits per channel. For most graphics, this is a good as 24-bit color, but on smooth gradients you can begin to see steps. Loosing another bit per channel would be very dramatic on screen, and even more dramatic on a continuous tone print.

-- Glenn Kroeger (gkroeger@trinity.edu), January 20, 2001.


As an aside... about 7 years ago, Microsoft along with another media company wanted to develop a CD-ROM tour of the National Gallery of Art. At the time, most people had only 8-bit color graphics capabilities, and storage space was at a premium. Since using 24-bit color (8 bits per color) was out of the question, and 3-3-2 subdivision of the available 8-bits gave terrible results and they didn't want to dither color (set you monitor to 256 colors and look at some web pages to see how bad this is!), they selected an optimum 256 color lookup table for each image. The results were suprisingly good. Some of these algorithms are included in Photoshop. If you want very compact copies of images, try saving images as a 256 color image with optimized lookup tables. It works well for many classes of images.

-- Glenn Kroeger (gkroeger@trinity.edu), January 20, 2001.

Glen, your latest post certainly gave some indication that 8 bits / channel may be the optimum number. Here is a quote from Jan Steinman from another board, I am curious of your opinion to his response. I am starting to beleive our eyes are the bottleneck if we exceed 8 bit files...

I don't see ANY reason for printing a 16 bit file. Human eyes simply aren't sensitive enough to detect such changes -- especially once they've been compressed down to the ~2.0 Drange of reflective material.

On the other hand, if you're archiving a file that will possibly undergo multiple edits in the future, it makes sense to save as much information as possible.

I believe printers will continue to be 8-bit devices for the foreseeable future, and if they ever do become 16-bit devices, it will be because cheap memory and fast processors have made 16-bits the standard for storage, NOT because the output will be any better.

-- Bill Glickman (bglick@pclv.com), January 21, 2001.


Bill, I am still an apprentice and this discussion is interesting to me too. I have read somewhere that correcting an image in 16 bits mode then passing it in 8 bits for output helps get richer tonal range in the shadows. But I too like to use the layers in Photoshop, not to degrade the original image. The 16 bits mode does not support layers yet, even in Photoshop 6. Anyway, the difference in a 8 or 16 bits image is most of the time not perceptible. It might make sense when the image contains subtle variations of the same color, to make transitions smoother. But then as you said, the output device must be capable of making use of the available datas. Another application for 16 bits image is in raw scans. A 16 bits image can be saved as is and then converted in 8 bits RGB or CMYK without loss. Some scanners have this 16 bits HDR feature. Anyone knows if the LightJet or Lambda printers are supporting higher than 8 bits data files yet?

-- Paul Schilliger (pschilliger@smile.ch), January 21, 2001.

Paul... you wrote.. Another application for 16 bits image is in raw scans. A 16 bits image can be saved as is and then converted in 8 bits RGB or CMYK without loss. Some scanners have this 16 bits HDR feature.

My Howtek does scan at 16 bits per channel. But PS6 brings the files down to 8 bit. I am unsure what you mean when you say a 16 bit file can be converted to 8 bit without loss? You are loosing half the data in the transition? Maybe I did not understand what you were saying?

-- Bill Glickman (bglick@pclv.com), January 21, 2001.


Bill, I think this has to do with color conversion algorythms. Normally to achieve the best possible quality, you would scan an image in RGB for inkjet or Lightjet prints or for the web, and in CMYK for pre-press. Now, if you don't know what you are scanning for yet, or are scanning for multiple purposes, the 16 bits HDR data gives you the total information available in both color environments. So converting an RGB image in CMYK from the 16 bits files should be just as good as scanning directly in CMYK. The drawback is that on some scanners, the 16 bit file is produced without the possibility of intervening on the process. If the original is too dark, for example, it will be scanned too dark, even if the scanner had the possibility to correct the curve in 8 bits mode. So, not yet perfect.

-- Paul Schilliger (pschilliger@smile.ch), January 21, 2001.


More info on this process from Scitex website at: (SOOM, Scan Once, Output Many concept)

http://www.creoscitex.com/products/input/eversmart/oXYgen.asp

-- Paul Schilliger (pschilliger@smile.ch), January 21, 2001.


Hi Bill. Everything I've read tells me that there is no printer or output device that can make use of 48 bit files, in fact there are only 3 file formats I know of that can store the 48 bit image format.
However, as previously mentioned, it's easy to see the difference in quality that you get from doing all your manipulation and correction on a 12 or 16 bit depth image, then converting to 8 bits. For instance, if you start with a RAW scanned file from a colour negative, you can easily do the inversion and RGB channel alignment to get a perfect positive from a 12 bit/channel file. The same thing's impossible if you start with only 8 bits/channel.

I suppose the question that you need to ask and answer for yourself is: "Is this image the best I can ever get it, and will I never need, or want, to make changes to it in the future?"
If you can answer "yes" to that, with 100% certainty, then you can change the mode down to 8 bits per channel and store 24 bit files with a clear conscience.
OTOH, if there's any doubt, then 48 bit files would be the safer bet.
West Coast Imaging have a little monogram on the philosophy of 'Sc an once, purpose many' that's relevant to this.

-- Pete Andrews (p.l.andrews@bham.ac.uk), January 22, 2001.


I always scan at 16bit (Linocolor scanner) and then open the document in photoshot, color correct in 16 bit, then convert to 8bit to do any manuiplation of the image. If you look at the histogram you have more "color data" to adjust when in 16 bit than 8.

-- Malcolm Matusky (malcolm@malcolmm.com), February 28, 2001.

Moderation questions? read the FAQ