1
\$\begingroup\$

I will use an external provider to scan 1000+ old slides ("diapositives"), mostly family snapshots made a few decades ago. The provider offers by default standard definition jpg (512 x 340 pixels) and high-definition jpg (3200 x 4800 pixels, 3500 dpi, estimated file size of 5Mb).

As an option (an increase of 30% of the price !!!) they also offer uncompressed TIFF format. I called customer support, and they told me that the resolution (# pixels & dpi) is the same, but quality is higher. They could not tell me which quality settings they use to generate the HD jpg. They also cannot provide the actual file coming from the scanner (I would think the scanner produces some type of RAW or HDR file?), instead of the TIFF.

So I have two questions:

  1. Intuitively, I would think it makes sense to also order the TIFF file for archiving purposes. But if they confirm they make the jpg with 100% quality setting, would it still makes sense to pay 30% more for the TIFF files?

  2. If I do buy the TIFF files as well, which software should I use to batch-convert into jpg ? A web-search gives me tens of options of converters, and I do not find any information about quality differences. Is this because the jpg compression algorithm is fixed/standardized, so high-end software like Photoshop or Lightroom will give the same results as a basic freeware ? Maybe I am confused by the notion of jpg engines that convert RAW, and where some brands like Olympus produce "nicer" looking (subjectively, I know) jpgs than e.g. Lightroom.

I am familiar with GIMP and Imagemagick (but not an expert), so any tips on settings to use ?

= = = = = EDIT 1 = = = = =

I have sent a request to tech support for additional information about the TIFF file and quality settings of JPG file, and herewith their answer (I have sent a follow-up question, as I did not understand their first two points).

  • 8 bits TIFF files come directly from the scanner
  • This processing allows for fully exploiting the 12bits from the scanner's CCD
  • After post-processing, the images are saved as 8 bits TIFF files (50Mb file size) and 95% quality JPG (typically 3-6MB file size)
  • Visually both are indistinguisable
  • TIFF only makes sense if someones want to further process the file

Looks sensible to me, but I did not understand the first part. Does the CCD scans at 12bits in total (4bits per channel, which seems very low to me), or 12 bits per channel (hence 36bits). If the latter, not sure why the scanning/capturing software cannot put this in a 16bit TIFF file instead of using 8bit TIFF, hence discarding some potentially useful data.

= = = = = EDIT 2 = = = = =

I received a follow-up email with further clarifications (please note that the original email is in French, which I translated in English; as I am not a film expert nor a native English speaker, word choices might not be 100% correct).

  • The scanner indeed provides 12bits per channel, on a linear scale. This is not very suited for the tonal spectrum/scale of photos, which is rather logarithmic. Therefore, it is very well possible to convert the 12bits linear from the CCD into 8bits tonal/color scale for a photo. In any case, the scanner cannot do any better.
  • For more information about the justification for using 8bits, please check the Kodak Cineon format which is based on 10bits, which was largely sufficient to store the images from professional-grade film scanners. For more information, see https://flylib.com/books/en/2.104.1/cineon_log_space.html
  • TIFF is very rarely requested, so no samples available so share.

So a quite comprehensive answer. I also had a follow-up call with them, and they say they use very good professional equipment, but their services are aimed at consumers, and this is still below what specialized laboratories use that cater for the professional market. So their scanner's output can be fully captured in an 8bit TIFF and 16bit would have no additional value.

= = = = = CONCLUSION = = = = =

I will forego the TIFF, and only order the JPG.

If by any chance, some of the diapositives turn out to be really amazing, and we want large prints, it probably is better to re-submit these to a really professional lab, re-scan them with a better scanner, and obtain a RAW or TIFF file for the few photos that are really worth it.

Thanks everybody for your contribution.

\$\endgroup\$
4
  • \$\begingroup\$ 12-bit data means 12 bits per channel. You are correct - to obtain highest quality, it would make sense to store this output into a 16-bit TIFF file. Why they don't do that is probably for some operational reason - maybe something to do with speed and storage space. As an example, if they are sending you a DVD, it will take longer to write more DVDs, if they are working with 16-bit as opposed to 8-bit files. I guess most of their customers simply don't want/need 16-bit TIFF files. Time is money. \$\endgroup\$
    – osullic
    Jan 28, 2021 at 19:42
  • \$\begingroup\$ Let me spam you with a link to some tests I did years ago. It is in Spanish so please use google translate. The point is that I am showing a process to compare the compression loss on different JPG implementations. Jpg also has some configurations that make some difference in the final result. I really need to update that link and make one version in English. otake.com.mx/Apuntes/PruebasDeCompresion1/CompresionJPG-03.htm \$\endgroup\$
    – Rafael
    Jan 29, 2021 at 1:11
  • \$\begingroup\$ @osullic Thanks. I do not know their workflow, but to my understanding (I read a nice review somewhere of a Braun slide scanner), it it is the scanning that takes a really long time; I would assume "processing time" to generate a 8bit vs 16bit is insignificant compared to scan time. TIFF files are only delivered on an external hard drive, which of course will take longer for 16bit than 8bit, but still will take much less time than the scanning itself. Anyway, I did ask these questions in a follow-up mail, so I'll edit my question if they answer me. \$\endgroup\$
    – Peter K.
    Jan 29, 2021 at 8:50
  • \$\begingroup\$ @Rafael Thank you ! \$\endgroup\$
    – Peter K.
    Jan 29, 2021 at 8:51

4 Answers 4

2
\$\begingroup\$

If the TIFF is 8-bit, the fact that the scanner is 12-bit is irrelevant, you have lost the 4 extra bits.

TIFF is uncompressed... or not. TIFF is a very flexible format, and in some cases the a JPEG encoding can be used in the TIFF format. But if it"s 50 megs vs. 5 megs, it is likely a lossless encoding.

A 95% JPEG is likely good enough. This is pretty much the quality you get as the JPEG output from a camera. The really important factor is the chroma sub-sampling. No sub-sampling is best, halved chroma is still about OK.

IMHO, don't over-think it, JPEG is more than enough for your needs. The real question is how good the scanning is (color balance, brightness/contrast). This is going to make a lot more difference than TIFF vs JPEG.

If you go with TIFF, you can batch-convert your TIFF to JPEG, using ImageMagicks's convert command in a terminal/shell, or using utilities such as XnView or IrfanView that have a batch mode. For IM, it is as simple as:

convert image.tif -quality 90 image.jpg

or, for recent versions:

magick convert image.tif -quality 90 image.jpg
\$\endgroup\$
1
  • \$\begingroup\$ Thanks for your help. I guess I'll listen to the advice of all of you, and just take the JPGs... \$\endgroup\$
    – Peter K.
    Jan 29, 2021 at 8:35
2
\$\begingroup\$

Let's assume there is some image data that is going to be stored as a JPEG file and also as a TIFF file.

JPEG is a lossy image file format. Even at 100% quality (whatever that means), the JPEG encoding process discards image information in order to reduce the file size. Many computer scientists have spent many years refining the JPEG compression algorithm such that the image information that is lost (in a high quality encode) doesn't generally have any real impact on an average person's impression of or satisfaction with the result.

TIFF is a widely-supported lossless image file format. Relative to the image data fed into the TIFF encoding process, no image data is discarded. So, if you use TIFF, your conscience can rest easy that you didn't opt for a lossy encode. Can you see any difference between the TIFF file and a high quality JPEG? Probably not. If you are then just going to encode the TIFF file to JPEG, is there any point in paying more for the TIFF? Probably not.

(Note: I'm not addressing the situation where you want to further edit the images. In that case, there can be an argument for choosing TIFF over JPEG, certainly if the TIFF has a higher bit depth than the JPEG.)

There is of course a massive difference between opting for JPEGs with a resolution of 3200 x 4800 over ones with a resolution of 512 x 340 pixels. I think I can quite confidently say, don't choose the lower resolution option for this project!

You can think of audio files as an analogy. You can save audio data as an uncompressed WAV file, as a 320 kbps compressed MP3 file and as a 96kbps compressed MP3 file. The 320kbps file is a lossy format - audio data has been discarded. But it's high quality compression. Can you hear the difference between this file and the lossless WAV file? Probably not. Can you hear the difference between the 320 kbps file and the 96 kbps file? Probably yes. The latter file has much greater compression, at the cost of sacrificing quality. Do you want to pay extra for a lossless format (over a high quality lossy format), only to then just go and compress it to a lossy format? I don't really see the rationale for that.

Addendum: Note that while a scanner may allow access to its raw image data, for most users this isn't a useful format to have images in. Scanner manufacturers (like digital camera manufacturers) make it simple for their hardware device to output images in a widely-supported file format. It's probably not useful for you to start requesting raw scanner data. High-quality, "first-generation" scanned images, saved in a widely-supported file format are what you want.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for your answer. I edited my question with some additional info I received. So TIFF has worse bit depth than I expected, and 95% JPG is better than I expected. I would definitely take the TIFF, if it would have been 16bits, and the JPG would have lower quality. Would you think that I loose a lot if I just convert a JPG to TIFF in case I want to do further processing/editing of some of the images? \$\endgroup\$
    – Peter K.
    Jan 28, 2021 at 18:49
  • \$\begingroup\$ If you want to do further editing, then starting with a high quality JPEG is not the end of the world. Converting to TIFF first is pointless. It doesn't bring you anything. It doesn't restore any additional quality over using the JPEG for editing directly. If you have a high quality JPEG, edit it and re-encode to high quality JPEG once, it's not ideal, but it's probably OK. The advice would be not to perform the edit/encode cycle repeatedly. \$\endgroup\$
    – osullic
    Jan 28, 2021 at 19:34
  • \$\begingroup\$ Thanks. As I responded to someone else, I probably needed a few people to push me to ignore my OCD tendency to always go for "the best", and forego the TIFF. \$\endgroup\$
    – Peter K.
    Jan 29, 2021 at 10:01
1
\$\begingroup\$

Is the service provider able to provide you with samples of the images? If so, you could put them on Google Drive or Dropbox, and someone would be able to pretty quickly tell you what settings they're saved with.

I would think it makes sense to also order the TIFF file for archiving purposes. But if they confirm they make the jpg with 100% quality setting, would it still makes sense to pay 30% more for the TIFF files?

It's pretty much up to you. Depending on the quality of the scans, the TIFF files may or may not be useful. JPGs saved with quality setting 100 have very little loss. The main advantage TIFF would have is the possibility of 16-bit color depth. But you have no assurances that is the case.

... I do not find any information about quality differences.

There aren't really any. Most programs use the same libraries, which are either built into the operating system, open source, or provided by the programming language.

Is this because the jpg compression algorithm is fixed/standardized...

Pretty much. Yes.

... so high-end software like Photoshop or Lightroom will give the same results as a basic freeware ?

Adobe uses their own libraries with different settings. For example, instead of quality settings from 1-100, they use 1-12. The results are comparable though, since, as you noted, JPG is a fixed standard.

... some brands like Olympus produce "nicer" looking (subjectively, I know) jpgs than e.g. Lightroom.

The visual differences come from how the raw data is processed into an image before being compressed by JPG.

I am familiar with GIMP and Imagemagick (but not an expert), so any tips on settings to use ?

On Mac/Linux, from the command line, ImageMagick is convenient.

for file in *.tif ; do
   convert "$file" -quality 100 "${file%.tif}.jpg"
done

The basic convert command on Windows would be about the same, but I don't know the syntax for looping in Windows. You can search for other GUI batch conversion utilities. As noted earlier, results will be about the same.

\$\endgroup\$
10
  • \$\begingroup\$ Windows command-line for loop: https://docs.microsoft.com/en-us/windows-server/administration/windows-commands/for \$\endgroup\$
    – osullic
    Jan 28, 2021 at 18:01
  • \$\begingroup\$ @xiota Thanks for your answer. I edited my question with some additional info I received. I did ask for a testfile in TIFF and JPG, but that request was ignored... I did ask again. So TIFF has worse bit depth than I expected, and 95% JPG is better than I expected. So would I loose a lot if I just convert a JPG to TIFF in case I want to do further processing/editing of some of the images? \$\endgroup\$
    – Peter K.
    Jan 28, 2021 at 18:32
  • \$\begingroup\$ I'll edit later to account for the new info you've provided. In short, I see no point to having the TIFF files. The JPGs should be good enough for light editing, unless the scan is botched, in which case 8-bit TIFF would also be useless. If you have an interchangeable lens camera, you can capture select frames with high-resolution and bit depth using a slide copy attachment. \$\endgroup\$
    – xiota
    Jan 28, 2021 at 20:08
  • \$\begingroup\$ As for the ignored requests... did they not even respond saying they were unable to provide sample files? If so, I'd be hesitant to trust my negatives to them. What if something happens and your attempts at communication are ignored? \$\endgroup\$
    – xiota
    Jan 28, 2021 at 20:12
  • \$\begingroup\$ @xiota Thanks for your help. I probably needed a few people to help me decide to ignore my OCD tendency to always go for "the best", and forego the TIFF ... Re: my request for sample files, this was part of the same email for tech details, on which I got a reasonably detailed answer. Let's see their answer on my follow-up email... But the company has a fairly good reputation in France, as one of the better "mass market" providers. Of course not "fine art" quality, but I checked a few of these, and they are just too expensive. \$\endgroup\$
    – Peter K.
    Jan 29, 2021 at 8:29
0
\$\begingroup\$

If it is a compressed Tiff I would pass on it. Uncompressed Tiff's are worth the extra cost, but it sounds like they are lying to you. Photoshop would be my choice for handling large uncompressed Tiff files. Never overwrite your original files.

\$\endgroup\$
3
  • \$\begingroup\$ Compression of not of TIFF files have no bearing on quality. Only minor inconvenience can be that the compression algorithm is not standardized, so there could be compatibility issues. Not sure how you deduce that they are lying. I thought the answers were quite transparent and complete. \$\endgroup\$
    – Peter K.
    Feb 1, 2021 at 11:04
  • \$\begingroup\$ Peter K. Your first line makes no sense. And, I believe the lab is dishonest. I believe thair marketing is dishonest. \$\endgroup\$
    – user85781
    Feb 2, 2021 at 10:32
  • \$\begingroup\$ Robert, I think you are confusing lossless compression (e.g. a compressed TIFF file) and lossy encoding (e.g. a JPG). Lossless compression is like making a Zip file. You compress the file (can be any type of file), and when you uncompress, you get the same file as before. Having said that, TIFF is often not-compressed (it takes time to compress, and there is not a single compression algorithm defined so there could be compatibility issues). \$\endgroup\$
    – Peter K.
    Feb 2, 2021 at 13:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.