8 BIT or 16 BIT

[mdaddyrabbit]

Could someone tell me what this means and which I should use. I actually dont know anything about it. I am running PSCS and I really would love for this to be broke down to me.

[Hellashot]

16 bit is actually 12 bit with 4 empty channels. Takes up a lot more memory/disk space than 8 bit, but you get more color range. JPG images taken are 8bit with a camera, RAW takes 12/16 bit.

[tim]

16 bit is actually 12 bit with 4 empty channels. Takes up a lot more memory/disk space than 8 bit, but you get more color range. JPG images taken are 8bit with a camera, RAW takes 12/16 bit.

That's not accurate, and it's somewhat misleading.

Canon Camera RAW is 12 bit, sure, if you choose 16 bit in CS RAW it does pad the last 4 bits with zeros. JPG is 8 bits, but it's meant to be logarithmic, not linear, so to the eye it will look like more than 8 bits.

Photoshop is proper 16 bit - perhaps some higher end cameras are 16 bit, or scanners perhaps. It gives you a much wider range of colors to work with. If you open a RAW image in Photoshop (not brigde) then make adjustments with levels or colors, the 16 bit images will stay more accurate than 8 bit images. How important it is I don't know. Personally I just use 8 bit images, and I will until I can see a difference myself, or need to do a lot of manipulations involving color.

[PhotosGuy]

Personally I just use 8 bit images, and I will until I can see a difference myself, or need to do a lot of manipulations involving color. For small prints & the web that should be OK. ;-0

[tim]

For small prints & the web that should be OK. ;-0

I do A1 prints (20*30) and i've not had a problem. Unless you're doing color manipulations then I don't think 16 bit gives you anything except more patience from having to wait for the machine ;) Many places state that saving in JPG level 12 will give you a print just as good as a 16 bit TIFF, and these are places that do a lot of large prints.

[robertwgross]

If you were trying to represent a red dot in 8-bit JPEG, then it would be a red dot of one of 256 shades of red.

If you were trying to represent a red dot in 12-bit RAW, then it would be a red dot of one of 4096 shades of red.

If you were trying to represent a red dot in 16-bit TIF, then it would be a red dot of one of 65,536 shades of red. If you start from here, it is not hard to see how the 8-bit JPEG causes a loss of color depth, and some people will see it, and some people won't.

---Bob Gross---

[tim]

How many shades of a color can a human eye see? And how many shades can a printer generate? Anyone who knows much about computers can understand the 8/16 bit difference easily enough, but how important it is is another question.

[quickben]

Most people have a varying degree of colour-blindness. In some, it means they cannot see entire colours e.g. red or green. In most people, we cannot see a certain part of a particular colour's range or gamut (I have trouble differentiating between light browns and kaki greens) but they are not diagnosed as having colour-blindness, it is that slight. If you can display more shades of a particular colour, more people will be able to see it as that colour.

That's why I shoot RAW and process in 16 bit and save to 16 bit TIFF.

Just my two pennies worth.

Gary.

[scottbergerphoto]

Take the same picture and convert it to 8 and 16 bit tiff copies. Do a number of edits in each(crop, curves, levels, hue/sat, sharpening). Then look at the histogram for each one in Levels and tell me which file you would rather have. I'll take the 16 bit.

[pcasciola]

It's not about how many colors the eye can see, or how many colors a printer can generate, it's about precision. Doing edits on an 8 bit image results in a loss due to rounding off on each step. Doing edits on a 16 bit image results in a lot more precision. Even if the final result is converted back to 8 bit, it will still be more precise than the one that was edited in 8 bit all along because of all the rounding that occurs in each step with 8 bit.

It's like using currency without decimal points. If you have $1 and you earn 30% interest, without decimal points you still only have $1, but with more precision you have $1.30. Earn another 30% an now you have $1.69, but in the less precise world you still only have $1. Convert the higher precision number now, and you have $2, which is more precise even though it was converted back to the lower precision in the end.

Sorry, that's the only analogy I can come up with at 9AM. I haven't had any coffee yet.

[PhotosGuy]

I agree with what Bob & the rest of you are saying & you've addressed the original question quite well. 16 bit IS better than 8 bit.
OTOH, I like to play in PS & sometimes wind up with a 100MB file at the end. If I'm never going to print a 16X20" pic, do I need to keep that huge file on my hard drive? I don't think so. If I change my mind (haven't yet) I'll go back to the original RAW file & start over. I'll probably know more about PS at that point & come up with a better processing job anyway!

So I'm sticking with my comment, " For small prints & the web that should be OK." ;-)

[Gerdav43]

If you have $1 and you earn 30% interest

Could you post the website or phone number to your bank. I'm ready to move all my money there. Then I could afford more L glass and more hard drives to convert to 16bit on interest alone.:rolleyes:

[Longwatcher]

How many shades of a color can a human eye see? And how many shades can a printer generate? Anyone who knows much about computers can understand the 8/16 bit difference easily enough, but how important it is is another question.

From my memory refreshed by some web research.

I was always told that humans can differentiate 125 shades of gray, so 256 provides more then enough gray for the human eye to see a smooth transistion. This means that we can see a band line between a shade of gray that is 1/125th of a difference between black and white (web research says about 1/100). We can only see three bands of color RGB, so if I remember my algebra correctly that would be 125x125x125 colors that we can differentiate. no double to 256 shades of gray and you get about 16.7 million shades required so that the average human can not see a difference between two colors.

Then there are color blind people ( about 10-12% of males, 0.5% of female) and people whose vision is not centered on the average (like mine going slightly UV side of normal).

In other trivia a 1200 DPI printer should be more then enough for the average human eyes resolution and 2400 DPI covers enough for just in case, with 4800 being a touch of overkill. Which is why you don't see significant improvement in printer DPI anymore.

Just my memory and some limited research.

[Jesper]

Have a look at this article:

Out of Gamut: The High-Bit Advantage (http://www.creativepro.com/story/feature/7627.html)
Why bother working with 36- or 48-bit color when you're stuck with 24-bit output? Color guru Bruce Fraser explains why.

And Norman Koren explains in detail why you have more dynamic range when you shoot RAW (12 bits per channel) instead of JPEG (8 bits per channel):
Tonal quality and dynamic range in digital cameras (http://www.normankoren.com/digital_tonality.html)

[lostdoggy]

How many shades of a color can a human eye see? And how many shades can a printer generate? Anyone who knows much about computers can understand the 8/16 bit difference easily enough, but how important it is is another question.

Here is a test. Go to your local paint store's paint chip kiosk and stare at them and see if you can differentiate the different colors and shades. Then take a picture of the kiosk and develope it in JPEG, TIFF 8 and TIFF 16 and print them out and see if you still can differentiate it.

[tim]

Are 8 bit files 8 bit per pixel, or 8 bits per channel? Not that it really matters. I have to wonder if it's worth moving to 16 bit images, which will be twice the size for the ones I edit...

Here is a test. Go to your local paint store's paint chip kiosk and stare at them and see if you can differentiate the different colors and shades. Then take a picture of the kiosk and develope it in JPEG, TIFF 8 and TIFF 16 and print them out and see if you still can differentiate it.

That's a good suggestion :)

[tim]

And don't you have to convert to an 8-bit file before you can save as JPG anyway?

[PhotosGuy]

And don't you have to convert to an 8-bit file before you can save as JPG anyway? Yes. Plus, there are some things you can't do with 16 bit files, so chances are you'll also convert to do those anyway.

[Hellashot]

That's not accurate, and it's somewhat misleading.

Canon Camera RAW is 12 bit, sure, if you choose 16 bit in CS RAW it does pad the last 4 bits with zeros. JPG is 8 bits, but it's meant to be logarithmic, not linear, so to the eye it will look like more than 8 bits.

Photoshop is proper 16 bit - perhaps some higher end cameras are 16 bit, or scanners perhaps. It gives you a much wider range of colors to work with. If you open a RAW image in Photoshop (not brigde) then make adjustments with levels or colors, the 16 bit images will stay more accurate than 8 bit images. How important it is I don't know. Personally I just use 8 bit images, and I will until I can see a difference myself, or need to do a lot of manipulations involving color.

It was accurate as to being a brief explanation of the differences.

[Mernya]

Our eyes have a wonderful ability to shift our own internal gamma. It's why we can see someone that is backlit just fine, but the camera needs a fill in flash. That same gamma comes into play when we look at color images. Fill your field of vision with something that is 'one' color, like a forest or a bowl of cherries. Give yourself a few moments and notice how many different shades you can see now that the color is the only one in your field of view. That's just my little note on the paint color test. It's not how many we can see, it's how many we process.


8 bits are per channel : RGB is 24 bit color (R8, G8, B8)
16 bit RGB would be 48 bit color.

In other trivia a 1200 DPI printer should be more then enough for the average human eyes resolution and 2400 DPI covers enough for just in case, with 4800 being a touch of overkill. Which is why you don't see significant improvement in printer DPI anymore..

Apples and Oranges.

DPI isn't pixels (since that is what our bits are about).

One company's DPI isn't comparable to another. A pixel (well, a cluster of three) on a screen can be made of X% of Red, Y% of Green, and Z% of Blue to make one of 16 million shades or so. A printer (ink jet) can make one color per well type (inkwells with light magenta, light cyan help) . To make red, it clusters magenta and yellow densely. To make pink, it spreads them out (which at that point, you see dots). There is no 'light red' color on a printer, just less dense red (which is magenta and yellow). Of course, the 6 color+ printers do change this somewhat.

That being the case, I believe you still will see technology improvements on inks, clustering, finer droplets (and they are very fine these days), and the like. We may see new print technologies or refinements of others (like dye sublimation).