8-bit to 16-bit: any better for color correction?

[Swizzle]

While the alterations would be 'smoother' (gradients, for example), you might not be able to see those changes in true 16-bit color. I'm really not sure if there are any monitors (and somebody correct me if I'm wrong) that can display all the colors in 16-bit, so it might not really make much difference at all whether you're working in 16- or 8-bit color. The only difference you might see is in things like, as I mentioned before, gradients.

Why do you want to work in 16-bit? What kind of images are you playing with? Is 16-bit really going to be that beneficial? Can you actually see much difference between 16,777,216 (sixteen million, seven-hundred and seventy thousand, two-hundred and sixteen) colors and 281,474,976,710,656 (two-hundred and eighty-one trillion, four hundred and seventy-four billion, nine-hundred and seventy-six million, seven-hundred and ten thousand, six hundred and fifty six) colors?

I would suggest that you just work in 8-bit mode since any 16-bit-specific changes probably won't be noticeable in 8-bit anyways. Besides, you're not going to be able to work with those images in most other programs if you keep them in 16-bit mode.

[madart]

As far as I remember and can think of, there's not much point in converting an image from 8-bit to 16-bit. You can test this for yourself by making a copy of your image, make the copy 16-bit and make the same adjustment to them both, for example, levels, after you've made the adjustment, open up levels again or the histogram and look at the values. There shouldn't be any difference.

But there is a big difference between an 8-bit photo and a 16-bit photo (which has never been anything else but 16-bit). If you do the same test between a true 16-bit photo and an 8-bit you'll notice the difference in the histogram.

What I used to do when I scanned positives or negatives, I always scanned them at 16-bit, made my adjustments with adjustment layers and saved them like that and then converted them to 8-bit and saved them as JPG or whatever fileformat I needed them as. The reason for that is that there is so much more information in the 16-bit images, and like you and swizzle stick said you do notice it in the gradients. but also, when you make colour corrections, you do get so much more details from a 16-bit photo than you would from an 8-bit.

But when I'm working with images from scratch I always make them 8-bit.

[halo]

sending someting to 16bit from 8 will have benefits if the colour correction builds colours outside of the 8 bit gamut...so whereas you may see banding in 8 bit on certain grads for instance, in 16 you wouldnt.

Put it this way, it may help, so its worth a try.

[madart]

halo, you're more than welcome to convince me and explain to me how an 8-bit image converted to 16-bit can actually get less banding than the original 8-bit.

As far as I remember, it just doesn't work. If an image is supposed to be true 16-bit it has to be 16-bit from the beginning.

If it does work like you say, I'd really like to know, because it would be great. Things might have changed in the last couple of years or so, I haven't worked with 16-bit photos for a while.

[halo]

halo, you're more than welcome to convince me and explain to me how an 8-bit image converted to 16-bit can actually get less banding than the original 8-bit.


No-one ever said that, not even the original poster, its not a one step fix... he's aware that the file doesn't change, ie banding doesnt dissapear when converted to 16 bit.
What he's asking is whether converting to 16 bit before doing a colour correction will help avoid the banding associated with the correction. The answer is yes, it will help reduce the banding it if the banding isn't present already in the 8 bit file (ie hidden in dark gradient thats being lightened) and the file is in RGB (CYMK bands anyway due to its space more than its range). Yes, converting to 16 bit is a different route that avoids banding caused by the file needing colours outside of the 8 bit file's colour space.

[madart]

Eviedently I didn't express myself fully. Naturally I meant that I wanted to know how you were going to get less banding in an 8-bit image converted into 16-bit WITH colour adjustments/corrections AFTER the conversion, I mean that's what we were talking about, colour corrections and adjustments. I am not at all talking about banding in the original files, but as a possible result in some gradients in some images after correction.

I am not saying that you're wrong, I am really interested to know and understand how it works. It's just that according to my experiences and from what I have learnt it doesn't work like that.

If you compare the histograms, of an original 8-bit image and an 8-bit image converted to 16-bit, after the corrections (the same corrections in both images), the histograms are going to look exactly the same and totally different from a true 16-bit image, naturally the "same" image.

If it does work the way you are saying, well fine, great! I am just curious how it works in that case.

[halo]

oh, no worries...i got the wrong end of the stick (again) :) :banghead: doh.

ok, well certain operations wont see any improvement because they will work from the existing colour structure of the file and just offset from that, however some adjustments will take the colour range into areas where 8 bit bands because it doesnt have the colours. Although the histogram doesnt get any better on the conversion, moving to 16 at least gives the potential to move it without colours falling into "holes" in the range.
An example of this is gradients...take most gradients and really squeeze or push them hard in levels, even in 8 bit rgb* and you will start to see grads...try the same function in 16 and you will often see an improvement. Subtle deep bright primary grads are a bitch for this.

*where narrow gamut spaces such as sRGB are used it does take some experimentation to determine if the banding can be avoided within the confines of that space...sometimes 16bit cant help you there because the gamut is the limiting factor, not the range of colour within it.

Of course some corrections dont improve just because they're just offsetting colour from what exists already...its more the operations that exagerate the difference between colours that benefit as these are the ones that tend to show the "gaps" (therefore bands) that 8bit has.

Although I haven't needed to myself, its been noted that conversion to 16 bit before a cymk conversion also can reduce banding. I tend to work in cymk preview and work around colour issues (force of habit more than anything) before hitting them head on, so i haven't had to resort to it for sometime...but I think its worth remembering, you never know when you'll get one of those awkward clients again;)

[madart]

don't worry :-) but thanks for taking the time to explain all of this. I don't usually work with CMYK or print, or at least not very often. I will remember what you said here, about the colours being outside of the gamut and not the color range and that's where you can get improvements in the converted image. great tip! thanks again.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.