Another 8/16 bits question

[Dropt]

Good morning everyone,

As there is no small questions thread, I find myself opening up a whole new subject for a small question. Here's the deal and it's quite simple : My boss has given me a 16 bits camera file to convert in 8 bits, then retouch. So far so good. But he's also adding (quoted here) "the files are 16 bits, first I increase the physical size by two then I step down to 8 bit files... please follow this always as the depth of the pixels in shadows and highlight will suffer otherwise".

I suspect him to have no ideas of what he's talking about, as he's no brilliant retoucher or anything near that. I've red quite a few arguments about 8 vs 16 bits here and on the internet, but never something likewise. I understand that 16 bits creates more tonal range, so I guess it's about conserving them in his sayings, but anyway it all comes down to a 256 levels range in 8 bits, so what's the point ? It's not like the resolution change between 8 and 16 bits.

[Tony W]

Re: Another 8/16 bits question

IMO if you are given a 16 bit file to work on then it is preferable to edit at 16 bit depth and only convert to 8 bits after your editing manipulations prior to output.

I cannot see how increasing the physical size size of a 16 bit file by a factor of 2 actually gives any benefit - the bit depth and the levels you have to play with (shadows and highlights included) remains the same, the file size will be larger but no true gains to resolution as the increase in size has been accomplished by software interpolation.

[Dropt]

Re: Another 8/16 bits question

I agree entirely with your first statement, also the pic is to merged with others which are in 8 bits, so I had to downgrade it pretty quickly. Also I could've taken all the other files to 16 bits, that was easier to do the first take for practical reasons.

Thanks for confirming my doubts, that just made no sense to me...

[andrewrodney]

Re: Another 8/16 bits question

I understand that 16 bits creates more tonal range, so I guess it's about conserving them in his sayings, but anyway it all comes down to a 256 levels range in 8 bits, so what's the point ? It's not like the resolution change between 8 and 16 bits.

It doesn’t provide more tonal range. The range is the range.
Think of this like a staircase. The height of the staircase is what it is (the tonal range). The number of steps in that staircase is the bit depth. You can have a 9 stop staircase height with 8 steps (bit depth) or 16 steps. But that doesn’t change the tonal range.

If given 16-bit data, keep it in 16-bit data forever (or save off an iteration in 8-bits per color if necessary while retaining the high bit archive).

[mistermonday]

Re: Another 8/16 bits question

It doesn’t provide more tonal range. The range is the range.
Think of this like a staircase. The height of the staircase is what it is (the tonal range). The number of steps in that staircase is the bit depth. You can have a 9 stop staircase height with 8 steps (bit depth) or 16 steps. But that doesn’t change the tonal range.

If given 16-bit data, keep it in 16-bit data forever (or save off an iteration in 8-bits per color if necessary while retaining the high bit archive).

+1; A really great analogy
Regards, Murray

[Dropt]

Re: Another 8/16 bits question

Re-thinking about it, the tonal range is indeed still the same. I mixed tone and color range here, how stupid. Thank you Andrew.

[daygraphics]

Re: Another 8/16 bits question

I had a nice long answer for you, but for some reason, my login had timed out by the time I hit submit. I hate it when that happens.

Longer story shorter. You are right.

I agree with the 16 bit suggestions put forth here if you are working with imagery for photographic output.

But if you are in advertising like me - then cut to the chase. Convert it to 8 bit right out the gate. You will hardly ever even notice any differences. You are going to press and are going to be dealing with CMYK conversion and out of gamut issues anyway. You'll work on smaller files, get done faster and make your deadlines easier.

In any event, you boss is whacky and is confusing some resolution issues with color depth ones. His process will do absolutely nothing good and may in fact, end up softening the end image.

[Flashtones]

Re: Another 8/16 bits question

My boss has given me a 16 bits camera file to convert in 8 bits, then retouch. So far so good. But he's also adding (quoted here) "the files are 16 bits, first I increase the physical size by two then I step down to 8 bit files... please follow this always as the depth of the pixels in shadows and highlight will suffer otherwise".

It sounds to me like your boss wants to end up with a file uprezed to contain twice as many pixels, and would like you to the uprez in 16 bits before beginning retouching rather than after.

Uprez first, then retouch.

Seems like a fair and simple enough request. If his testing bears this out as superior, so be it. I doubt the reverse would prove itself better. I'd prefer to sharpen and clone, etc, at size rather than uprez any flawed retouching.

[daygraphics]

Re: Another 8/16 bits question

It sounds to me like your boss wants to end up with a file uprezed to contain twice as many pixels, and would like you to the uprez in 16 bits before beginning retouching rather than after.

Uprez first, then retouch.

Seems like a fair and simple enough request. If his testing bears this out as superior, so be it. I doubt the reverse would prove itself better. I'd prefer to sharpen and clone, etc, at size rather than uprez any flawed retouching.

While appreciate your thinking and agree with you on the facts, I don't however think that is what her boss is thinking. I say that because it sounds like she is getting pretty hires (16 bit to boot) files as her originals to start with. I seriously doubt that there may be anything of any significance gained, or lost in the process. And sharpening can be harder to control, and even cause a graininess in a large file with lots and lots and lots of pixels. And she says that her boss' "formula" is in response to her having to convert to 8 bit eventually - so I really think he thinks he is counteracting the effect of the 16 bit to 8, by upsizing by a factor of 2. Not doing what he thinks - and is just likely wasting time and losing money.

[Flashtones]

Re: Another 8/16 bits question

I don't however think that is what her boss is thinking. I say that because it sounds like she is getting pretty hires (16 bit to boot) files as her originals to start with.

"My boss has given me a 16 bits camera file to convert in 8 bits, then retouch. So far so good. But he's also adding (quoted here) "the files are 16 bits, first I increase the physical size by two then I step down to 8 bit files... please follow this always..."

I don't see anything indicating hires from the OP, and what does "hires" even mean"? Where exactly is the dividing line between hires and low res? If the res is so high why do they need to step it up? And what assumptions about output size and resolution are you working from?

While we're at it, what assumptions are you making about the OP's computer, RAM, storage capacity, etc. to justify going against their bosses explicit instructions to upres in 16 bits before stepping down to 8?

I seriously doubt that there may be anything of any significance gained, or lost in the process. And sharpening can be harder to control, and even cause a graininess in a large file with lots and lots and lots of pixels.

Wow that a lot of assumed pixels. How many are we talking about here?

And she says that her boss' "formula" is in response to her having to convert to 8 bit eventually - so I really think he thinks he is counteracting the effect of the 16 bit to 8, by upsizing by a factor of 2. Not doing what he thinks - and is just likely wasting time and losing money.

I think he believes there will be less rounding errors by uprezing in the native 16 bits. While it's debatable (and testable,) whether or not it yields a visible advantage, I see no reason to defy one's boss over it. Even marginally inferior results will rest solely on the shoulders of the employee. One has to make a lot of assumptions to justify the risk.

[kav]

Re: Another 8/16 bits question

Unless you're going through a ton of images on a computer that is short on ram or very old, I don't see what the problem would be doing this one action at 16 to a presumably single layer file. Practically any computer from the last five years does this in a few seconds. If not, you may have an IO bottleneck.

[daygraphics]

Re: Another 8/16 bits question

Based on some recent comments, one would "presume" that I suggested that Dropt should defy her boss and refuse to follow his directions. Nothing could be further from the truth. I am simply expressing my opinion on the subject and YES, my assumptions on the situation. Should the facts prove otherwise, then my opinion and assessment of her inquiry would change. But as it stands, I continue to believe that she is getting hires (larger than 3000 x 2000 pixels) - approximately 7 x 10" at 300 DPI - largely due to the fact that she is even getting 16 bit to start with. Stock photography at online sites (even Getty) is never 16 bit upon download.

What am I basing these assumptions on? The likelihood that she is going to press and printing at a standard 150 ls ruling. I also presume that fact because she is merging multiple images, which is usually done for design and print and less for photographic output. But I could be wrong.

In any event, none of this is relevant and all pure conjecture unless Dropt logs back on and fills in the blanks.

[Decoboy]

Re: Another 8/16 bits question

Hi Dropt,

If I had to bet my life on it, I would say that the boss thinks that by doubling the file size by 2 and then going to 8bit he ends up with much the same thing - i.e. 8 is half of 16 so you just double it first right?
We all know that's just plain wrong, but most people don't get the idea of bit depth, and can't comprehend that each single bit increase is a doubling. Furthermore the average Joe doesn't even understand resolution/dpi/ppi etc

Whilst the 8vs16 argument is best avoided, I'll run the risk and make a couple of observations;
As others have noted the range doesn't increase - just the fineness of the steps - and although 256 steps might not be enough in a few cases, we hardly ever need 4096 steps. The trick is learning/recognizing when it might be worth working in 16bit. As others have noted, always keep the 16bit, just in case. I am however firmly in the Dan Margulis camp - it rarely matters-assuming you are working with a non trashed original.

BTW, it should be noted that, apart from a couple of med/large format exceptions most cameras are only 12bit and a few 14bit - thus the whole 16bit thing is a bit of con - i.e. there were never 16bits captured- it is an extrapolation from the original 12-14 bits - nonetheless 12-14 is a BIG increase over 8.

At the end it comes down to what you are trying to achieve, i.e. the final output and how much torture/manipulation you intend. Maybe you could let you boss know about 'this great site I found that explains...'

[daygraphics]

Re: Another 8/16 bits question

DECOBOY... you hit the nail on the head. While most here who have commented, are pretty well versed on the subject - though they may have some varying opinions - the average Joe, has "no idea" what all of this means. At least not in relationship to each other. Image capture to begin with (16 bit vs 8 or 12) - going all the way back to the high end pricey scanners, has been debated and argued. There was always squabbling and debate over this issue of "who had the better" scanner. Most people who spent less than 40K on their scanners and professed that they were capturing 16 bit, failed to read their own equipment's technical specs which often clearly said the scanner captured 12 bit (or even 8 bit) raw data. The 16 bit was always interpreted from that original capture. But it sounded good when competing against the big boys. Nothing seems to have changed when it comes the current crop of mid-range to hi-end cameras out there.

Now you talk about resolution? Input vs output, dpi vs ppi and throw in there LS (line screen), bits and bytes, levels of gray, etc., many of the folks working in the field can't fully comprehend it all.

I attribute a lot of this confusion to a lack of education, misinformation and oftentimes a short lived history in the field. Had you been doing this since the beginning of electronic imaging, you were forced to understand these principles and make the appropriate choices, at varying stages during production process.

Finally, I know you meant it tongue-in-cheek, but I would not have Dropt have her boss check out "this site" since some have roasted him - most importantly "she".

[Dropt]

Re: Another 8/16 bits question

Hello everyone,

So many answers ! daygraphics and decoboy, you're absolutly in the right; I quite sure my boss thought that it might be a resolution matter between 8 and 16, which is not. You pretty much said it all, and my case is now solved ! Thanks a lot.