8 bits or 16 bits

[wxnut]

In working on RAW files in CS2 it gives me a choice of saving as 8 bits / channel, or 16. Whats the benifit of either or?

Thanks,
Doug Raflik

 

[Big Mike]

16 is better, there is no doubt of that...but it's not always necessary to work in 16 bit for everyone.

Basically, using a higher bit image, gives you a lot more tones & colors. Most/all of which are in between the tones & colors that you get with 8bit...so you may not even see them or realize that they are there...but the more there are, the smoother the transition between different tones & colors.

Here is a link, which I think is the same article that I read in the magazine several months ago. http://www.digitalphotopro.com/tech/the-bit-depth-decision_2.html

 

[wxnut]

So other than the larger file size, is there a reason why I wouldnt want to work in 16? Do pro labs print in 16 bits or 8?

Doug

 

[Big Mike]

Well, I was/am under the impression that nothing (or very little) actually gets printed in 16 bit. (by the way, I believe that what they call 16 bit, is actually 14 bit).

With this in mind, it seams like it would be a waste to work in 16 bit, if the info gets thrown out when you prints. However, the benefit, which I believe that article hits on...is for editing.

Here is the example that got to me. When you edit the levels in Photoshop, then open the levels dialog again, the histogram will most likely have slots missing. This is because when you changed the image the first time, you changed a lot of the pixels...and since you are not adding pixels, those changed pixels had to come from somewhere...and you are left with slots or gaps. When you work in 8 bit, there are fewer tones/colors to pull from, so the gaps are bigger. In a worse case scenario...this could be visible in your image...usually seen as 'steps' in what should be a smooth color gradient.

When you edit in 16 bit, the histogram on 2nd levels command, will still be slotted...but the gaps are a lot smaller...therefore, the steps of color/tone on the image, will be smaller.

Again, it may not be visible in the image, so it may not be necessary...however, the best 'workflow' would be to work in 16 bit, up until you print the image. And if there are actually printers that can use those 16 bits, well then, that's a good reason to use it.

 

[astrostu]

To sort of expand on what Big Mike said, pretend that your monitor is 8-bit, and that 8-bit means you can have colors ranging from 0 through 9.

Your 8-bit image has lots of dark colors, but almost no saturated parts, so in reality the pixels range from around 0 through 5. In the Levels, you adjust 'em so they span the whole range, 0 through 9. But, all that did was move the pixels of brightness 5 to 9, those of brightness 4 to 7, 3 -> 5, 2 -> 3, and kept 1 and 0 the same. So now, you have brightnesses of 0, 1, 3, 5, 7, and 9 in your image. You are noticeably missing 2, 4, 6, and 8, and it will appear as brightness "steps" in any kind of gradient.

Now let's say you have a 16-bit image, which supports brightnesses from 0 through 99. Your monitor will effectively scale this down from 0 to 9 for display purposes, even though the image has more information.

But say it's the same image, so the brightness only goes from 0 through 50. You adjust the levels to scale it from 0 to 99. So you've moved 50 -> 99, 49 -> 97, 48 -> 95, and so on. When you look at the histogram after this, there will still be those noticable gaps at 98, 96, 94, and so on, but because it's compressed to the same scale as the 8-bit, they'll look smaller.

And then when you convert the final 16-bit image to 8-bit, it will basically divide all the pixel brightnesses by 10. So because you had brightness information at 71, 73, 75, ... 99, you will now in 8-bit mode have brightness information at 7, 8, and 9. You will NOT be missing the brightness at 8, which you would've if you'd originally done everything in 8-bit.

Make sense?