When should I see the difference between JPEGs and TIFFs in film scans?

ChrisR

ChrisR

I'm a well known grump...
Messages
11,026
Name
Chris
Edit My Images
Yes
So, I understand the prevailing wisdom that JPEGs lack "flexibility", and one should prefer TIFFs (or RAWs in the digeri world, but that's a whole different ball-game). For years I have ignored this conventional wisdom, and scanned almost always to JPEGs. Most labs only provide JPEGs when they scan for you. But recently I've been using Filmdev, who will provide TIFFs at no extra cost if you ask. Admittedly they are 8-bit TIFFs, thus losing the 16-bit per channel advantage (if such there is) that TIFFs can offer. Anyway, I've been asking Filmdev for TIFFs as well as JPEGs for a while now, with the idea that for suitable images I would use the TIFFs for the purported extra flexibility. A couple of times I think I did use the TIFFs, although I can't find the actual examples in my Aperture library!

Now, I do understand that if you load a JPEG into, say, Elements, make some changes, save (export?) it as a JPEG, and repeat, there will be a progressive deterioration of the image due to the successive lossy compression steps. But I suspect that most of us use a different workflow. I load my images into Aperture (most these days will likely use Lightroom), where they stay. Edits are non-destructive; nothing happens to the original image until I export a finished image for posting or printing... except that's not quite true, occasionally an image will get some extra tweaking in Elements (or Affinity Photo, these days). But in both these cases the image in Aperture has its edits applied, then is converted to a TIFF and transferred to the external editor, and a TIFF or possibly a PSD file comes back. No extra compression steps there.

I should say, I can see NO difference at all in the histograms of freshly imported JPEGs and TIFFs of the same image. I've just done some edits of an image (crop, auto-levels, definition), and then applied those same edits via Apertures Lift/Stamp capability to the TIFF, and there is the tiniest of differences in the histogram, so slight at first I thought it wasn't there.

A disadvantage of asking for TIFFs as well is that they are 3 times the size of the JPEGs, so my scanned file downloads take 4 times as long, and I lose lots of disk space...

So, what I want to understand is, what sort of situation (presumably, edit scenario) would make importing and using the 17 MB Filmdev TIFFs preferable to importing and using the 5 MB JPEGs?

(I realise this is a rather general question and should perhaps be in a more general forum rather than Talk Film & Conventional, but it is based on a particular film-based scenario, and I trust you folks. Plus, non-film folk may spot the topic and weigh in with insights anyway. But if the mods want it elsewhere, that's OK!)

[ETA: thread revived because the availability of 16-bit TIFFs from Filmdev made me want to revisit this issue...]
 
Last edited:
I would describe uncompressed Jpg and tif as just different modes of packaging the image information. Yes, if you process a file and re-save it, it will progressively degrade (lose information) - this applies to tif's as well as jpg's. So perhaps the best strategy is to keep the original scans intact so that they can be revisited and processed anew if wanted.
 
droj said:
I would describe uncompressed Jpg and tif as just different modes of packaging the image information. Yes, if you process a file and re-save it, it will progressively degrade (lose information) - this applies to tif's as well as jpg's. So perhaps the best strategy is to keep the original scans intact so that they can be revisited and processed anew if wanted.
Click to expand...
This doesn't apply to uncompressed tiffs or tiffs that use lossless compression, which you can save as many times as you like without degradation (unless the data gets corrupted!). Lossy jpeg-style compression is supported by the tiff format (it's an option when you save from Photoshop, etc.), but I doubt it's commonly used. Here's a discussion on the types of artefact you can expect from jpeg compression: https://photo.stackexchange.com/que...peg-artifacts-and-what-can-be-done-about-them . Much depends on the quality settings - I can't say I notice artefacts when I use (say) the quality = 10 setting in PS. I'd hope that good scanning services use high quality settings. Even if the original file is a jpeg, it never needs to go through more than one more lossy compression cycle (to save the final jpeg, if that's the target). Any intermediate steps should use a lossless format. Edit: I agree you should always keep the original scan or in-camera file, though, whatever format it is in.
 
Last edited:
droj said:
So perhaps the best strategy is to keep the original scans intact so that they can be revisited and processed anew if wanted.
Click to expand...
I agree. All my stuff is stored as it came out of the camera (.jpg) and the working copies held elsewhere.
 
DG Phototraining said:
I was just about to give a full and practical answer when I remembered I was a "non-film folk" sorta guy - so bye

Dave
Click to expand...

Hi Dave, I specifically mentioned I was happy with insights from non-film folk; anything you could add would be gratefully received! In that para I was only trying to explain why I didn't post in one of the other forums...
 
droj said:
I would describe uncompressed Jpg and tif as just different modes of packaging the image information. Yes, if you process a file and re-save it, it will progressively degrade (lose information) - this applies to tif's as well as jpg's. So perhaps the best strategy is to keep the original scans intact so that they can be revisited and processed anew if wanted.
Click to expand...

I think JPEGs are always compressed, TIFFs rarely so (although as mentioned, the TIFF format does allow for compression, lossy or lossless). I very strongly doubt that Aperture (or LR) uses lossy compression when writing TIFFs to transfer to external editors. Certainly the file sizes don't suggest that.

EDIT: on the last sentence, Roger, AFAIK that's exactly what Aperture (and presumably LR) does: saves the original scan (JPEG or TIFF) intact until exported for printing or uploading...
 
Last edited:
I guess what I'm trying to say here is, I'm not seeing any reason at the moment to waste the extra file space for TIFFs. And what I'm asking is, what kind of scenario (presumably based around some sort of extreme editing) might actually provide that justification?
 
Retune said:
This doesn't apply to uncompressed tiffs or tiffs that use lossless compression, which you can save as many times as you like without degradation (unless the data gets corrupted!).
Click to expand...
Ah - it's not the saving itself that results in degradation of a tif, but any processing that has been performed. Make that distinction. Processing an image file degrades it. Push curves and levels about as you will, then inspect the histogram - it'll have gaps and bars in it. If you then over-write your original, some file information has gone for good.
 
All cameras and scanning devices start with a raw file ( which itself is based on the tiff format,} but is almost totally unprocessed.

both a camera or scanner will then save that file in the format requested. Usually either raw or Jpeg. or in the case of a scanner often as a tiff.
tiff and raws will be saved in various bit depths depending on the device and your intention

Jpegs are lossy.. that is they chuck away unused data, so are alway less amenable to the Post processing recovery of highlights and shadows.
Jpegs also reprocess the image every time they are saved, so the loose quality on each reprocess.

While scanners can save as raw files, few people choose to do so. they usually save as uncompressed tiffs in 8 or 16 bit. ( an 8 bit fill has massively less data than a 16 bit file)
An 16 bit uncompressed Tiff is the ideal staring point for working in photoshop as it retains most data, but can be very large. ( even larger than the original raw file which was probably only 12 or 24 bit.)
For this reason some people do nor care to store 16 bit tiffs. But only store the original raw file, and a jpeg derived from the worked over tiff.
However while the Raw has all the original data, It does not save the corrections made on the tiff in photoshop. So unless you save the tiff in at least 8 bit, you can not go back to make further changes to near the original quality.

I tend to hold on to the original Raws only for a short time after converting them. I then save only the worked on tiff. I rarely save Jpegs as they can be regenerated at anytime from the tiffs in lightroom, or more easily, in bulk, in IrfanView.

By chucking the raws I lose the ability to return to square one.....Ideally one should keep the Raw and the tiff.... but in most cases I do not.

From My point of view File space is relatively unimportant as it is so cheap and backing up should be automated to a very large extent. It is the Jpegs that I find to be the waste of space.
 
Last edited:
droj said:
Ah - it's not the saving itself that results in degradation of a tif, but any processing that has been performed. Make that distinction. Processing an image file degrades it. Push curves and levels about as you will, then inspect the histogram - it'll have gaps and bars in it. If you then over-write your original, some file information has gone for good.
Click to expand...
Oh sure - any deliberate change you make to the image, even something relatively innocuous like adjusting levels, could have undesirable consequences you might want to reverse, so it's important to keep the originals (and perhaps intermediate stages if they involve a lot of work, and your image format or software doesn't allow you to step back). But multiple lossy compression cycles add another layer of degradation for no good reason. I suppose the question is whether a single extra cycle (receiving a jpeg rather than an 8-bit tiff scan from a processing service) is worth avoiding, at the cost of extra disk space or download time.
 
DG Phototraining said:
I was just about to give a full and practical answer when I remembered I was a "non-film folk" sorta guy - so bye

Dave
Click to expand...

giphy.gif
 
Just so I can properly follow the discussion, could you clarify whether we're discussing colour or black and white? And I have no concept at all of how file sizes translate into the only measure I know and use - pixel sizes (how many along the long and short sides of the negative).

I always save the raw files from my scanner, as I can then reprocess them with different settings if that gives a better result without the time taken to rescan. I got into this habit when my 5x4 scans took 4 hours each, and not having to rescan literally meant I could adjust settings to get a better scan. Otherwise, it was, if not impossible, at least impractical.

I also think that as time goes on, not only will I learn more, but software might develop as well; hence having the original raw data means that in either case I can produce a better final image.
 
I don't think you will see any real difference between an 8bit tiff and an 8bit jpeg. Both have been converted/compressed down to 8bit data by the same program/process/algorithm. And if you are using a program like LR where the edits are stored externally and only applied on export, then there's no real concern about edits/saves.

The main advantage of tiff is as a 16 bit container for 12/14/16 bit raw data from a camera/scanner so that there is no data loss/compression.
 
Doesn't it depend on how many saves you do, jpeg will just deteriorate over time. I save out as a jpeg for the final image but otherwise keep things as tiff
 
StephenM said:
Just so I can properly follow the discussion, could you clarify whether we're discussing colour or black and white? And I have no concept at all of how file sizes translate into the only measure I know and use - pixel sizes (how many along the long and short sides of the negative).

I always save the raw files from my scanner, as I can then reprocess them with different settings if that gives a better result without the time taken to rescan. I got into this habit when my 5x4 scans took 4 hours each, and not having to rescan literally meant I could adjust settings to get a better scan. Otherwise, it was, if not impossible, at least impractical.

I also think that as time goes on, not only will I learn more, but software might develop as well; hence having the original raw data means that in either case I can produce a better final image.
Click to expand...

For these purposes, we're discussing colour from negatives scanned at 2000 ppi. So it's roughly 2000*3000 pixels ie a 6 Mp image (times 3 bytes per pixel, comes to 18 MB).

I guess the same arguments would apply to black and white, though. Also, I know of no lab that supplies raw images from their scanners. I have not yet investigated saving raw from my own scanning, though that is on my to-do list!
 
Thanks. As far as raw goes, when I started scanning the only programs I had were SilverFast (cut down edition) and Epson scan provided with the (Epson) scanner) and VueScan (recommended by Barry Thornton in part because it could save the raw file. From this, I assume that providing a raw file - even as an option - may well be software rather than scanner dependent.

My interest in the colour/black and white side was down to whether the scan would have 8 bits in one channel (equals 256 different tones) or 8 bits in 3 channels (16 million tones). Again, when I started, 8 bits and one channel was an option - or perhaps better, dire possibility.

For what it's worth, my own tests some years back on resaving jpgs using Photoshop and the highest quality setting, showed that viewed full screen (not at 100%) seven saves revealed a noticeable degradation. I expect at 100% it would appear sooner.
 
AshleyC said:
Doesn't it depend on how many saves you do, jpeg will just deteriorate over time. I save out as a jpeg for the final image but otherwise keep things as tiff
Click to expand...
Yes that's good practice, but as mentioned processing degrades tif's too, even if less so. All the more reason to preserve originals, and also perhaps to use adjustment layers in such as Photoshop. If you can work in 16-bit all the better.
 
Retune said:
Oh sure - any deliberate change you make to the image, even something relatively innocuous like adjusting levels, could have undesirable consequences you might want to reverse, so it's important to keep the originals (and perhaps intermediate stages if they involve a lot of work, and your image format or software doesn't allow you to step back). But multiple lossy compression cycles add another layer of degradation for no good reason. I suppose the question is whether a single extra cycle (receiving a jpeg rather than an 8-bit tiff scan from a processing service) is worth avoiding, at the cost of extra disk space or download time.
Click to expand...

But since I use Aperture (and the same applies, I believe, for LR), adjusting levels is reversible; it does not make any changes to the original image until exported, when the edits are applied. I can see that exporting as a JPEG and re-importing agin will have consequences. I suspect that exporting as a TIFF, editing externally and re-imprting, then repeating this process with different editors (as I have seen mentioned many times on 'hints and tips' blogs) could also have negative consequences.

No-one has yet mentioned a realistic editing scenario within my workflow that is likely to have significantly more adverse consequences for images that started as JPEGs rather than TIFFs!
 
Given that jpgs are produced by lossy compression and in this case presumably the tiffs aren't, some fine detail will have been lost with the jpg before you get it. The question then becomes - does it matter? If it's down to a just noticeable difference, you'd need to compare them side by side to judge. But in theory you'll be starting with slightly less data in the case of the jpg.
 
StephenM said:
Given that jpgs are produced by lossy compression and in this case presumably the tiffs aren't, some fine detail will have been lost with the jpg before you get it. The question then becomes - does it matter? If it's down to a just noticeable difference, you'd need to compare them side by side to judge. But in theory you'll be starting with slightly less data in the case of the jpg.
Click to expand...

I have compared the image from the JPEG and TIFF side by side, and the histograms too, and can see no difference in either. As mentioned above, after a few of the sort of adjustments I normally make, there is no visual difference in the images (to my eyes) and only the tiniest difference in the histogram... I had to check back and forth several times before noticing it.

I plan to do some edit/save cycles with both as a test. It's an indication of my skill level that when I loaded an image into Elements at first I couldn't work out how to do a Levels adjustment, then when that was achieved I couldn't get the Export button un-greyed! I'll try something else...
 
StephenM said:
Given that jpgs are produced by lossy compression and in this case presumably the tiffs aren't, some fine detail will have been lost with the jpg before you get it.
Click to expand...
If the original scan was recorded at a bit depth greater than 8 bit, then the 8 bit tiff is also "lossy compressed."

Simply opening and closing a jpeg (or any other file) does nothing to the data, and this is what LR/Aperture/etc do when "editing" an image. It is not until the image is changed and *re-saved* that the data is reprocessed, and that is where the losses start to come into concern.

It doesn't really matter what the file format is. The main problem with jpegs is that "unused data" is discarded... I.e. if you process an image so that a larger area goes to black and re-save, those areas are now black and you're not getting the data back. But the same can happen with a tiff. i.e. if you do the same with a layered tiff in PS, and then flatten/re-save you will have the same issue.
 
Right, I've done 10 cycles of importing an image into Affinity Photo, making a slight adjustment to brightness using an adjustment layer (up and down by 3% on successive rounds), and exporting (using default settings), then re-importing the last version. First for JPEGs, then for TIFFs. Then I have imported the 10th version into Aperture and compared with the original JPEG and TIFF.

The first thing to note is file sizes. The JPEG from Filmdev was 5.6 MB; the first save from Affinity Photo was 8.3 MB, with each successive version being slightly larger until the tenth was 9.7 MB! Meanwhile, the TIFF from Filmdev was 17.7 MB; the first save from Affinity Photo was surprisingly 14.6 MB, and each successive version was either 14.5 or 14.6 MB.

Comparing the 4 images visually, the only discernible difference was that the 10th TIFF save was slightly brighter. I'd put that down to a slight error in my brightness adjustments. Looking at 400% (!) at a few selected areas of the image, I can't really see any differences; certainly no obvious artefacts.

Comparing the histograms, there are very slight differences in each. The differences in the first and last TIFF histogram are, if anything, slightly more obvious than for the JPEG case, but this is mostly down to a slight uptick at the bright end.

I should say that the histograms in Affinity Photo looked different from those in Aperture, and did look more "posterised", and as I went through I felt there were increasing differences, but these aren't showing up in the Aperture histogram. However, I've just been back into Affinity Photo and re-checked the first and last histograms, and although they look different from the Aperture histograms, the differences are no more significant.
 
sk66 said:
If the original scan was recorded at a bit depth greater than 8 bit, then the 8 bit tiff is also "lossy compressed."
Click to expand...

True. But I assumed that the compression (with tonal loss) took place before the final save as tiff or jpg; and that in that case the jpg would receive a further loss. I've never looked into it in detail, but my assumption would be that the loss in the two types of compression would be different. It's possible to argue that fine detail is just another way of describing subtle tonal differences and the two compressions work in the same way, and I'm just assuming that smoothing out the detail for jpg is more aggressive.
 
Last edited:
For me, the best comparison would be between a tiff and a jpeg of the same image, as supplied by FilmDev. Do they routinely give you access to both files if you request tiffs, or do you just get tiffs? If they only provide tiffs, it might be worth asking for a few corresponding jpegs as a special request just to get a handle on how good their compression settings are, since these are outside your control. Jpeg compression does indeed introduce specific sorts of artefacts (typically you might see blocking and 'dirt' or halo effects around edge details, and a general loss of sharpness) and the effects may not be visible in a histogram. Even an 8-bit lossless/uncompressed tiff will in theory be a better starting point than a jpeg, assuming (e.g.) a 16 bit original. Both will have data loss due to bit depth reduction, but only the jpeg will have additional compression artefacts. But the question is - are these artefacts significant enough to care about, given a single pass with whatever quality settings are used by FilmDev? Using Photoshop's most aggressive settings, even a single pass looks terrible. Using the maximum quality settings, you may see no real difference to the tiff.
 
Retune said:
For me, the best comparison would be between a tiff and a jpeg of the same image, as supplied by FilmDev. Do they routinely give you access to both files if you request tiffs, or do you just get tiffs? If they only provide tiffs, it might be worth asking for a few corresponding jpegs as a special request just to get a handle on how good their compression settings are, since these are outside your control. Jpeg compression does indeed introduce specific sorts of artefacts (typically you might see blocking and 'dirt' or halo effects around edge details, and a general loss of sharpness) and the effects may not be visible in a histogram. Even an 8-bit lossless/uncompressed tiff will in theory be a better starting point than a jpeg, assuming (e.g.) a 16 bit original. Both will have data loss due to bit depth reduction, but only the jpeg will have additional compression artefacts. But the question is - are these artefacts significant enough to care about, given a single pass with whatever quality settings are used by FilmDev? Using Photoshop's most aggressive settings, even a single pass looks terrible. Using the maximum quality settings, you may see no real difference to the tiff.
Click to expand...

On request they provide both JPEGs and TIFFs, which is what I have been getting for the past few months. The comparisons described in post 24 were done using a JPEG and TIFF of the same image, both provided by Filmdev. The JPEG appears to be of reasonably high quality (larger file sizes typically than when I scan my own and save to JPEG), but those saved by Affinity Photo are higher (100%, they say).

I will put up the image and the 10th iteration JPEG version in a later post. If anyone wants access to the original files for their own experiments, please let me know.
 
This is the unmodified TIFF version as provided by Filmdev, but shared to flickr via Aperture's inbuilt feature as a JPEG:

R2-03409-0013T by Chris R, on Flickr

This is the 10th iteration JPEG version as described in post 24, similarly shared to flickr from Aperture:

J10 by Chris R, on Flickr

In both cases there may of course be artefacts introduced by flickr, but I uploaded them both at actual size so you should be able to download a larger version if interested.
 
Thinking more about this, I'm reminded that all scans are in some sense interpretations of the negative. Different scanner hardware, software and operators will produce different results. The following 3 images are from an earlier thread about some problems I had scanning Velvia transparency film; I believe colour negative film would be even more open to variations particularly from the software involved, as both the underlying orange mask and the film's characteristic tonal S-curve need to be interpreted. The differences visible here are FAR greater than those between the TIFF and JPEG treatments from one particular scanner...

a) This was my best interpretation, Plustek 7500i with Vuescan Pro:



b) This was from Photo Express in Hull (I sent them the processed, cut film strips and they scanned them); EDIT they use a Noritsu Koki scanner:



c) This last was from UK Film Lab, a highly respected lab now transplanted to Canada as the Canadian Film Lab. They keep their source material for a year, but when they asked me if I wanted it the postage costs were horrific for one set of transparencies, so I declined... so it's about the only set of film I don't have from the last 5 years! EDIT They used a Fuji SP-3000 scanner



Hmmm...
 
Last edited:
Well Chris IMO your scan is the best IF you use autocolor in Photoshop (which gave just a hint of magenta) and just highlight the shadows slightly....Photo express seems to have a blue cast and UK film lab might be accurate to the eye but is boring. All subjective but it is how I liked your scan, mind you what I see is in Photoshop which might look different by the round about way to show in T&C...and what I see on my monitor might look different on others.
 
Last edited:
Arr, but you get extra Browne points from the Cognesentee for saving as a TIFF wear-as your just regarded as a common, unknowledgable ouyk if you use Jpegs !
 
Stumpy said:
Arr, but you get extra Browne points from the Cognesentee for saving as a TIFF wear-as your just regarded as a common, unknowledgable ouyk if you use Jpegs !
Click to expand...

Yes, that's the feeling I had which is why I started the thread. I've now convinced myself to ignore the cognesentee (luvly word that!) and stick to JPEGs for my film scans. Might up the quality of the JPEGs when I scan them myself, however. Nevertheless, I'm now firmly an ouyk! :)
 
If it takes one to know one, and you're so confused that you don't what what you are, then you can't be one. :D

:wacky:

:jaffa:
 
I think you've just proved I don't exist, Stephen. Which I have sometimes suspected!
 
"ai" - that's a NIkon lens term, isn't it?
 
Something to do with rabbits' ears? You're right it IS surreal!
 
ChrisR said:
Thinking more about this, I'm reminded that all scans are in some sense interpretations of the negative. Different scanner hardware, software and operators will produce different results. The following 3 images are from an earlier thread about some problems I had scanning Velvia transparency film; I believe colour negative film would be even more open to variations particularly from the software involved, as both the underlying orange mask and the film's characteristic tonal S-curve need to be interpreted. The differences visible here are FAR greater than those between the TIFF and JPEG treatments from one particular scanner...

a) This was my best interpretation, Plustek 7500i with Vuescan Pro:



b) This was from Photo Express in Hull (I sent them the processed, cut film strips and they scanned them); EDIT they use a Noritsu Koki scanner:



c) This last was from UK Film Lab, a highly respected lab now transplanted to Canada as the Canadian Film Lab. They keep their source material for a year, but when they asked me if I wanted it the postage costs were horrific for one set of transparencies, so I declined... so it's about the only set of film I don't have from the last 5 years! EDIT They used a Fuji SP-3000 scanner



Hmmm...
Click to expand...

What I see here is not really related to scanners, jpeg/tiff, or interpretations so much, but an issue with subject brightness range. You have very deep shadows and very bright highlights, spanning many zones. At some point you need to control the contrast with some sort of local adjustment (e.g., ND grad filter at time of exposure, virtual grad in post processing) to specific parts of the image (e.g., the bright parts).

Unfortunately, labs will ordinarily only make global adjustments.

The reason, I think, that the Photo Express image and UK Film Lab images are different is because Photo Express elected to save the highlights at the expense of shadow detail, while UK Film Lab elected to get more shadow detail at the expense of highlight detail, so this is where some interpretation did come in, but the labs were forced to do more interpretation because of the exposure choices.

If I were taking this photograph, I likely would have used a two-stop ND grad filter at time of exposure. By controlling the subject brightness range, I reckon that the scans from the two labs would have looked more similar as they wouldn't have been presented with a choice of salvaging either shadow or highlight detail.
 
You must log in or register to reply here.
Back
Top