11 Bit depth and file formats
I’m getting a bit far out here, so may be even less correct than usual. But it’s worth at least attempting to cover these topics on a thread about scanning.
First, bit depth. I think this appears in a couple of different ways. In older posts on t’Internet (from some years ago) you may well find discussions about scanner bit depth. This refers to the output of the A/D converter at the heart of the scanner. In modern scanners this appears to be 16 bits, which is enough to encode a density reading of 4.8, so for all practical purposes this is not now an issue… unless you are buying an older scanner. I think you’d need to be sure your older scanner had a bit depth of at least 12 bits to avoid that being a bottleneck in scanning.
When you scan, you can save your image as a JPEG (8-bits per channel), a 24-bit TIFF (sometimes referred to as an 8-bit TIFF… confusing, isn’t it) or a 48-bit TIFF (16 bits per channel). I did a
thread a few years ago comparing a file saved as JPEG with the same image saved as a 24-bit TIFF (both actually came from Filmdev rather than my own scanner). The generally accepted advice is that JPEGs deteriorate quite quickly if you save and then re-open the files, so I made 10 minor manipulations of both files, saving and re-opening each time, and compared the difference. My opinion at the end was that for me (NB!!) there was negligible advantage in 24-bit TIFFs as I could see no significant degradation in the JPEGs, while of course the TIFF file size was much larger. I guess the answer here is likely to depend on the extent of the image manipulation you normally do. However, a good point was made in that thread, that BOTH 24-bit TIFFs and JPEGs already suffer from data loss when the scanner software deuces the 12, 14 or 16 bits per channel from the scanner down to 8 bits per channel.
I have not attempted a similar comparison with 48-bit TIFFs. I’d expect the larger files to be ever so slightly better for “normal” scans, but for my purposes the re-doubled file size would not make it worth while. The 48-bit TIFFs don't suffer from the same "data loss at source" issue though. You pays your money (for disk space) and you makes your choice!
Your scanner software might allow you a choice of saving your black and white scans in a greyscale colour space, which only requires 8 or16 bits per pixel. Just a word of warning, however, after doing this for years I started to move towards Capture One Pro for my image processing. This software does not recognise greyscale images, and will generally neither load nor process them. So remember to scan your black and white in a RGB colour space, with 24 or 48 bits per pixel.
Both Vuescan Pro and Silverfast AI Studio will also let you save the raw data from the scanner, for processing later (either re-imported into the same program, or using a post-processing application that recognises them). I presume any given scanner provides the exact same bits to either Vuescan or Silverfast to process (or any other scanner application, come to that). But I don’t know at this point whether the Raw files produced are the same. I’m guessing probably not, they’ll likely be packaged up in different ways. Anyway, Vuescan will let me save the raw data from my 7500i scanner as either a 48-bit TIFF or a 64-bit TIFF. In the latter case, it includes the infra-red channel used to detect dust and scratches. I don’t know if any other application is capable of processing the 4th channel, but I thought it worth a mention.
EDIT: I have recently scanned some Provia 100F slides on my Plustek 7500i with Vuescan Pro, to JPEG and also Vuescan RAW. I imported both to Capture One Pro, which happily recognised the RAW files and allowed me to process them. Results were much better starting with RAW files than with JPEGs, though disk space was 105 MB for the RAW and under 2 MB for the JPEGs. I also discovered that 2400 samples per inch scans were in fact 3600 samples per inch scans, downsampled by Vuescan. /EDIT
As mentioned above, I normally scan my images as JEPGs… but this is NOT advice, just what I do! YMMV. However, if I were making a linear positive scan of a negative for later inversion in an external software package, I think it would definitely be advisable to save as 48-bit TIFFs (or even better Raws). In the process of inversion all the large values become small, and all the small values become large. As in a matrix inversion, IMO you need as many bits as possible, fewer bits are more likely to give poor results. You may remember the warning from ColorPerfect back in Section 10 that the file input had too few bits. This is one of the advantages of making your inversions in the original software package at the time of the scan: the package has access to all the bits available from the scanner!
That was effectively all I wanted to say about file formats; there might be more to say in threads on the individual applications.