8 bit vs 16bit editing

[Dylan & Marianne]

I was just wondering how many of you work in 16 bit before converting back down to 8 bit for output (web jpeg or printing)
The reason I ask is that someone has recently brought it up on another forum and was strongly advocating working in 16 bit though I've had my doubts.
I don't disagree that issues like banding and colour handling are better in 16 bit. In fact, I used to work in 16 bit all the time (as indicated by my out dated signature)

My main reservation is that in the past, I distantly recall that on the final conversion, it can (though rarely) introduce some artefacts.
As a result, I changed my workflow a long time ago to just work in 8 bit from the start so that I'll know what I'll be dealing with at the end rather than having to fix a 'nasty surprise' on the final conversion.
Is my memory playing tricks on me? Does a final conversion from 16bit to 8bit do anything to the image that's untoward (eg introduce banding) ?

I appreciate the theory behind the 16bit argument but I wasn't strongly convinced to keep doing it from past experiences since the final file will always be an 8 bit one.
Love to hear your experiences regarding this topic ?

[ameerat42]

Sometimes, depending on what I can't do in raw and as an 8-bit. I haven't noticed any banding that resulted from the conversions back and forth.
Am.

[William]

Not very informative answer from me Dylan, But I have processed both and have'nt noticed a change , I process at 8bit now all the time , I could do a couple of tests on some of my recent Sunrise shots , I noticed some colour banding coming in , I could reprocess one or two from RAW at 16bit and see what happens

PS : But I'll still have to convert to 8bit in the end

[Dylan & Marianne]

i might try a direct side by side processing of 8 and 16 bit versions of the same file and see what happens I guess
- thanks for the replies!

[Xenedis]

I always work in 16-bit mode.

I cannot say I've been tempted to tr working in 8-bit mode, even though the final JPG will be an 8-bit image.

I use Photoshop's native PSD format, and in my PSD files (which I keep in 16-bit mode), I have all my layers and the maximum bit depth. I consider my PSD files my blueprints, so I prefer to preserve the best quality there, and then produce JPGs as required or desirable.

I have never compared bit modes when processing, but by very nature of coverting to JPG, you're introducing pixel destruction anyway.

[Cris]

I have wondered this myself and would be interested in your findings-Thanks

[etherial]

I always work in 16bit as well, always thought it was better to work with the best file you could for as long as you could (ie until whichever output). I'd be interested in the comparison.

[WhoDo]

When working exclusively in GIMP, I had no choice but to work in 8-bit colour from go to whoa. It wasn't uncommon to end up with images that were colour banded, especially in areas of distinct transition as around a sunset. The blues of the sky would be smooth but the transition of the yellows and oranges around the sun would separate and leave the result looking pretty tatty when going for higher contrast.

Now that I use LR4.1 for 99% of what I do, I work exclusively in 16-bit colour and the results are so much better. The only time I am forced to drop back to 8-bit colour depth is if I want to perform some layering in PSE that requires 8-bit colour mode. Can't understand why they do that, but I'm sure there's a good reason.

Bottom line, Dylan, is why throw away 75% of your colour information if you don't have to? Sure, if you're only publishing to the web in JPG then you are going to have to drop a lot of colour data at the end of the process, so working at that level all the way through is probably worthwhile so you know what you're getting. OTOH, if you're going for prints with full colour profiles then you should make sure you've got as much colour data as possible IMHO. Subtlety in colour is easily lost, especially when going a bit heavy-handed on the saturation. That won't matter unless you print big and your viewer's are discerning. JMHO of course.

Update: I thought you might appreciate this assessment of the value of lossy vs. lossless compression for web images. It shows just how much colour information can be sacrificed by a lossy compression algorithm like JPG. If you're already going to lose some colour data, wouldn't you prefer that to be the LAST thing you lose before publishing, just to ensure you have the most possible data for the compression algorithm to work from? Personally, I'd prefer it if PNG, a lossless compression, became the standard. When JPG was devised there were heavy restrictions on bandwidth for web publication. That is no longer true; and a PNG image uncompresses to the identical data of the original while a JPG does not so why not take advantage of the improved bandwidth to display richer images?

[Lance B]

I always work in 16bit until the final upload image for web. The reason being, you never know what you're going to do with the TIFF file as you may want to print big or whatever.

[Warbler]

I always edit in 16-bit. When you're doing some editing like selective levels work, if you're in 8-bit, you'll notice the histogram chunking up more than if you do the same in 16-bit. I equate this to using a fine grate and a coarse grate. Rub your cheese against the coarse grate and lots more of the cheese has to be moved to the nearest hole than being able to pass through its own hole on the fine grate. Alter the values of your pixels and they have to fit a coarser colour profile. There are less values available in 8-bit than are available in 16-bit. You still have to bring it down to 8-bit to save as jpeg, but you're not rounding already rounded numbers. The more editing you do, the greater the effect. The worst effect of this is colour banding which I'm sure we've all seen at one time or another.

Of course you may not notice a difference in side-by-side comparisons, but different image content with be affected differently. From a purely personal anecdotal experience. I notice the tools in PS seem to have a different "sensitivity" when applied to JPEG than they do a 16-bit PSD file.

[Dylan & Marianne]

guys, can I say that I don't doubt 16bit is better to work in - but there you have it , a beautiful 16 bit image that needs to be converted to any kind of output (unless there are 16 bit compatible printers out there?)
my real question was whether the edits in 16 bit hold up when in the end, for any kind of output, you need to sample back down to 8 bit - something I have to try out as I am now comfortable with working in 8 bit (banding and artefact introduced means I know I should either try to correct it or ease back on the PP - if the downsampling is the last step, then the last thing I want to have to do is do add more PP steps)

[ameerat42]

(Ult., it's about what you're trying to achieve from what you've got. I doubt if there will ever be a definitive and general answer to the question as posed, because the very Q covers a very wide range of conditions. That is, there's no single answer to a host of (implicit) questions.)
Just a thought.
PS: As posed, it was 16-bit vs 8-bit"

[I @ M]

guys, can I say that I don't doubt 16bit is better to work in - but there you have it , a beautiful 16 bit image that needs to be converted to any kind of output (unless there are 16 bit compatible printers out there?)

Dylan, I think that part sums it up.

Many commercial printers ask for JPEG files which are by default 8 bit and those that like to receive TIFF files ask for them in 8 bit -----

[Dylan & Marianne]

thanks Am an Andrew -the printer we've been using for years always used to covert the 16bit files to 8bit anyway which greatly influenced the change in our workflow to all 8 bit
however, I am tempted to go back to editing in 16bit if I find that I can do cleaner edits which don't lose their fidelity on the last step of downsampling - just have to push some limits when I get a spare moment to compare the differences.
Ps. everyone else who has mentioned their experiences with 16 bit, I agree with you - there really isn't a reason to work in 8 bit in terms of quality of the edit - it's the final 8 bit product I was more asking about via the 2 pathways (sorry if that was't clear in the opening post)

[WhoDo]

it's the final 8 bit product I was more asking about via the 2 pathways (sorry if that was't clear in the opening post)

No, Dylan, it was clear. It's just that the question is a loaded one. That's why I posted the link to the various compression algorithms and their uses. I can't understand why professional photography printers would want to take 8-bit JPG (a lossy compression) if they want to be proud of the results they can produce. Far better to request a lossless compression; either lossless TIFF or PNG, and know they are reproducing what the photographer captured in total. They've got to expect the colour variations demonstrated in that link to result in a poorer product from the photographer's perspective, surely.

[Mark L]

.... I can't understand why professional photography printers would want to take 8-bit JPG (a lossy compression) if they want to be proud of the results they can produce. ...

Sorry to side track, but how many professional photography printers accept anything else, and if they don't .........?

[Kym]

Speaking from a IT perspective there is a principle that applies here...

• Keep as much data for as long as possible! Once its gone you can't get it back.

This applies to transactional systems as much as imaging systems.

Personally I keep my original raw images in Lightroom,
then significant edits in PSD (usually smaller than TIFF, complete with layers),
and only output 8bit JPEG for we or print as needed.

I don't always edit in Photoshop, so most of my edits are LR only, which is the raw + the sidecar file, i.e. very small.

There is NO point in going from 8 bit back to 16 -- you have already lost information and the resultant 16 bit is at best an approximation.

[Dylan & Marianne]

Mark, mine accepts 8 bit TIFFs which is the reason why i work in that format from the start
I can only rely on anecdote from distant memory which led me to believe that the 16->8bit TIFF sometimes resulted in changes but if there isn't, there is no question for me to be going back to working in 16 bit since I've had a pc upgrade in between too!
edit: I don't ditch my RAWS either - they sit there waiting for a rainy day in the future for re-edits lol - I consider them to be my base file and not the PSD as I find I'll often edit images quite differently and would need a very big server if every edited file retained all of its layers in 16 bit!

[ricktas]

So say in 12 months time the printing industry starts to move to 16 bit printing. Do you then re-edit your RAW files to create 16 bit versions, or will you leave your older photos in 8bit and change your processing to 16bit at that time?

For me it is called future proofing. I will edit in 16 bit mode. I save a full 16bit file and then I reduce the size and save as jpg. Once uploaded to wherever, the jpg version gets deleted. Seems to me almost everyone posting in this thread is trying to justify their own PP methodology, 8bit / 16bit do it whichever way you want. Neither is wrong, neither is right, they are just different.

My printer is happy to take 16bit TIF files. Gets the results I want, even if they downsize them to 8bit to print them.

[Dylan & Marianne]

To answer your question rick, I would reedit from the RAW (also, because the image is already ready in 8 bit form, I'm already happy with its final result so the fact that the printer now offers 16 bit won't influence the editing that's already gone by unless there were correctable flaws to do with bit sampling ?)
Whenever we'e had new exhibitions and I want to do up an old photo, I rarely go back to the already edited TIFF because I've usually had a new vision for the image
Having said that, we do have a 'print ready' folder for specific size and mediums ( ie different sharpening , different vibrance and brightness etc)
I'd hope that in 12 months , I'd be printing new images and not old ones hehe
ps. I am trying to be objective with the discussion since everyone seems to be quite polarised but only a few people have answered the specific question I had - and that is that it seems few people have had a problem downsampling from 16 to 8 bit so I think I'll switch back to my original workflow!

EDIT: exceptions would be exposure blend images and panoramas - its too painful to go back from scratch unless I spot a major problem!