Nikon Coolscan V, 12bit or 16bit?

[dano1]

Hey all,

 

Was just reading in a photography magazine how today's DSLRs are only

12bit, then upscale to 16bit. Seems like an awful waste of bits to

me. On the other hand, it allows for 69 Billion Colors, as opposed to

8bit's 17 Million.

 

What I'm wondering though is if the Nikon Coolscan V makes use of all

16 bits. It would be a great way to increase quality in images

without increasing file size.

 

BTW, do any of the TIFF lossless compression algorithims bring a 12Bit

image that is scaled to 16bit back down to 12bit file sizes?

 

I have noticed that using LZW Compression I can get file sizes down to

2/3 thier original file size. This is remarkably close to the 12bit

vs 16bit file size difference.

 

Then again, does anyone really need 281 Trillion Colors?

 

 

Thanks,

 

Dan O.

[dano1]

Well, I guess I found the answer to the 12bit/16bit question myself.

 

It's actually 14bit, which allows for 4.4 Trillion Colors.

 

That's really cool, my scanner can output almost 4.4 Trillion more colors than a DSLR, and about 2x-3x the resolution.

 

That's a lot more information about the image!

 

Anyway, I'd still like to know if LZW or any other TIFF compressions make use of the extra 0's in the file that are put in when upscaling to 16bit.

 

Thanks,

Dan O.

[robert_m3]

The Coolscan V scans at either 8 bit or 14 bit, not 12 bit. It saves 14 bit scans in an

uncompressed 16 bit tiff.

[Ed_Ingold]

Dan, my DSLR produces 12-bit files. Many medium-format digital backs produce true 16-bit files. The advantage of higher bit depth in scanners is that you can make more adjustments without causing posterization. It has little to do with dynamic range or color display.

[dano1]

"Dan, my DSLR produces 12-bit files. Many medium-format digital backs produce true 16-bit files. The advantage of higher bit depth in scanners is that you can make more adjustments without causing posterization. It has little to do with dynamic range or color display."

 

Edward, I'm full aware of the benifets of higher bit-depths.

 

It's these extra colors that prevent posterization during heavy adjustments.

 

Dan O.

[peterblaise]

.

 

There are at least two benefits to additional bits:

 

- Smoother continuous tone transitions.

 

- Greater recovery of shadow and highlight detail during subsequent curves adjustments.

 

8-bit files have neither, but if the subject and the exposure are right on, it doesn't matter, much.

 

16-bit files permit a greater flexibility on revisiting a captured image and making enhancements according to the photographer's re-vision.

 

32-bit files are a reality in Photoshop CS2 - nice!

 

It's NOT the total number of colors that matters, it's the total number of shades or tones available from which to make subsequent image detail recovery adjustments.

 

A 16-bit file has 256 additional tone information points for every SINGLE point in an 8-bit file. Think about it. When s-t-r-e-a-t-c-h-i-n-g an 8-bit file to recover shadow or highlight detail, what's gonna fill in the resulting gaps? NOTHING! With a 16-bit file, you've got a buffer of 256 additional data points in-between each 8-bit point to smooth out the results.

 

However, you da boss. Your negatives are your masters, so you can always re-scan as your grow your photographic skill set. Why not scan at low-bit low-rez now, and only rescan at hi-bit hi-rez when you understand and want the features and benefits of all that extra data later in life? The negatives will be waiting, patiently, for you to revisit them anytime you want!

 

Let us know what you do, and why not share some samples?

 

Click!

 

Love and hugs,

 

Peter Blaise peterblaise@yahoo.com http://www.peterblaisephotography.com/

 

PS - My digital capture is 14-bits. ;-)

[ellis_vener_photography]

This answer will not directly address your question about file sizes and compression. A a

quick partial answer is that once you reduce an image from 16-bit to 12 or 8 bits bit per

channel, those other 4 or 8 buts of information is gone forever. That is a different thing

entirely from a data compression scheme to reduce overall file size.

 

 

Photoshop is in reality a 15 bit + 1 system. So what we are talking about is 32, 769

gradations of shade in each of the red, green, and blue channels. If you want to look at it

as total gradations of color you multiply the three channels together and get

35,187,593,412,609 total color gradations. Can anyone actually see the subtle and minute

differences between each of these shade? No. The best estimates I've seen is that the

human visual perception system (eyes + brain) at its peak (a young very healthy adult) can

distinguish roughly 603,979,776 different shades --or about 9.5 bits per red. green and

blue channel in digital terms. I am digressing for a reason.

 

 

What Photoshop does when it "upscales" a 10, 12, or 14 bit per channel to 16 bits per

channel is to add zeros as place holders. Colors are definitely not being added.

 

Yes the Nikon Coolscan film scanners (as do the scanners from Epson, Canon Microtek,

Fuji, Imacon, etc.) make full use of all 14 bits of information per color channel but only to

the extent that there is that much information in the original piece of film to start with.

 

Back to my earlier digression. while humans cannot see all of those incredibly subtle

differences in shading, Photoshop and other digital image processing systems don't have

the limitations we humans do and can make use of the differences. Some of that data is

destroyed by rounding errors in processing and some is destroyed simply by the steps we

take to refine an image-- apply a filter, sharpen or use USM, edit color, rotate an image,

adjust exposure, use curves or levels, etc. but the amount of data destroyed per step is

fixed, think of this as different size buckets and the range of information (color) as a

pool. If we start with a large pool of information the percentage scooped out compared

to the size of the pool ends up not being significant. Of course if you started with a

smaller pool -- say an 8 bit per channel image (16,777,216 total gradations of color --

already significantly below what a young, healthy human eye can perceive) the percentage

taken out by each bucket scoop will be more potentially more noticeable. How much more

noticeable will depend on the visual nature of what is image in the image. Where you have

large areas of smoothly graduating color the damage will show up first, Banding in a blue

sky for example.

 

Most output devices are 8 bit per channel capable. the goal with use high bit depths to

start with is to end up with what JeffSchewe calls "the best 8 bits".

[peterblaise]

.

 

Here's a bit, so to speak, of cleanup on our discussion so far:

 

Earlier: "... I have noticed that using LZW Compression I can get file sizes down to 2/3 their original file size. This is remarkably close to the 12bit vs. 16bit file size difference..."

 

Peter Blaise responds: BUT totally unrelated. TIF compression does not throw away bits and is a lossless compression.

 

-----

 

Earlier: "...the Nikon Coolscan film scanners (as do the scanners from Epson, Canon Microtek, Fuji, Imacon, etc.) make full use of all 14 bits of information per color channel but only to the extent that there is that much information in the original piece of film to start with..."

 

Peter Blaise responds: WRONG! The number of bits has nothing to do with image content, as if there were such a thing as less-than a certain "bit" content in one or another analogue image. All the bits in the scanner do is resolve the inbetweenies or smoothness across any image. If an image has a range of 10:1 or 1,000:1 contrast, THAT contrast whatever it is will be recorded, if accurately scanned, with the number of bits the scanner can resolve in it's AD analogue to digital converter.

 

-----

 

Earlier: "... Photoshop is in reality a 15 bit + 1 system..."

 

Peter Blaise responds: Yes, pre-CS, but good point. For some reason, Adobe previously limited itself to 2^15+1 (32,769) shades, not true 16-bit 2^16 (65,536) shades. 8-bit is 2^8 (256) shades.

 

-----

 

Earlier: "... What Photoshop does when it "upscales" a 10, 12, or 14 bit per channel to 16 bits per channel is to add zeros as place holders..."

 

Peter Blaise responds: We're talkin' about original CAPTURE, not "up-bitting" (nor "up-scaling", either, which would be "up-rezzing", right?).

 

When Photoshop "up-bits", it converts 0-255 shades to their proportional position across the broader 0-32,769, NOT inserting 0s anywhere! I use sub-pixel gausian blur to smooth out the resulting up-bit, but that's rare when I need to do that.

 

"Inserting 0s" implies that Photoshop inserts additional pixels with 0 value, which would be up-rezzing with black pixels inserted between existing pixels -- huh? I can't imagine HOW to do that, let alone why!

 

Here's up-bitting: Suppose we have an 8-bit file with 3 tone values - 12, 128 and 243 - and convert it to 16-bit, that file would still have 3 tone values, now up-bitted to 1,536, 16,384 and 31,226. No 0s! The two images, by the way, would look identical.

 

-----

 

Earlier: "... say an 8 bit per channel image...16,777,216 total gradations of color..."

 

Peter Blaise responds: 8-bit is 256 gradations of tone, times 3 colors when RGB (but NOT 16 million gradations of anything!).

 

Let's not confuse tone (bit depth per channel) with chrome (total bit depth per channel times number of channels).

 

-----

 

Carry on!

 

Click!

 

Peter Blaise peterblaise@yahoo.com http://www.peterblaisephotography.com/

[robert lee]

The Coolscan V has a 14bit ADC. This is an awkward bit size for the computer to work with (where 32, 16, and 8 bit quantities are "natural.") So, when scanned data is stored into a 16bit TIFF file, each 14 bit value is scaled by 4, i.e., shifted left by 2 bits. The scanner isn't making up new colors.

 

Now, as it turns out, each doubling of scanning passes gives you one additional bit of color resolution. So, if you multipass scan twice, this gives 15 bits color resolution. If you multipass scan 4 times, this gives you a full 16 bits.

[hakon_soreide]

Peter Blaise said:

<blockquote><i><p>

Here's up-bitting: Suppose we have an 8-bit file with 3 tone values - 12, 128 and 243 - and convert it to 16-bit, that file would still have 3 tone values, now up-bitted to 1,536, 16,384 and 31,226. No 0s! The two images, by the way, would look identical.</i></blockquote>

<p>

The computer only reads 0's and 1's, so presumably it just adds 0's to the string of 0's and 1's already there, which is the same thing you're talking about, only you're referring to it in the decimal system.<p>

I assume that whenever the computer upscales to 16-bit from 8, the logical thing, as Spock would say, is to add 8 empty bits at the end, so bin 1100, or dec 12, becomes bin 1100 0000 0000, or dec 3,072 as my computer calculates it, or - in the case of the actual 15-bit conversion used by PS, one would have to assume it added 7 bits and used bin 110 0000 0000, or dec 1,536, as Peter suggested.<p>

In both cases, lots of zeroes!

[peterblaise]

.

 

Let's not get confused about the meaning of zeros.

 

Zeros to Photoshop are BLACK pixels in an image file - NOT what up-bitting or up-rezzing does.

 

Zeros to a computer have different meanings dependent upon the base system (2 or 10 or hex) and if they are before or after other numbers and /or any "decimal" point.

 

I'd like MORE zeros in my paycheck, but please put them BEFORE the decimal point and after the most siginficant digit, but not AFTER the decimal point - $0100.00 does not equal $1000.00 does not equal $1.00000!

 

To paraphrase Gertrude Stein, "A zero is not a zero is naught a zero is knot a zero!"

 

Click!

 

Love and hugs,

 

Peter Blaise peterblaise@yahoo.com http://www.peterblaisephotography.com/

[mr._smith]

Dan,

 

The Nikon Coolscan 5000 has 16 bits and digs a little bit better in the shadows then the Coolscan V. Not by much, just a little bit. Otherwise, you'd get about the same scans from both scanners.

 

If your budget allows it go for the Coolscan 5000. If it doesnt, don't give it another thought. You won't miss much. The rest is all semantics.

[Ed_Ingold]

Dan, then you must also know that computer files are saved in multiples of 8-bit bytes. Since a 12-bit color file is more than 8 and less than 16, it is stored as a 16 bit file. The unused (low, on a PC) bits are stored as zeros.

 

LZW-compressed TIFF files have limited compatability with various software packages. I guess you knew that too. Why compress when disk memory is so cheap? Is your time worth nothing?

 

I guess I don't know.

[roger_smith4]

"LZW-compressed TIFF files have limited compatability with various software packages. I guess you knew that too. Why compress when disk memory is so cheap?"

 

Processing power is also cheap and for me, cheaper than hard drive space (I have a Pentium IV laptop). I barely notice a difference opening and saving compressed tiffs. I output to jpeg.

[hakon_soreide]

I'm quite on board with Peter's suggestion for adding zeroes in the right place on my paycheck. Perhaps there is some sort of 8 to 16-bit conversion that might add them in the right place? Well, not to be too greedy, just one extra bit might do the trick for starters.<p>

Anyway, as Peter said, there are lots of kinds of zeroes out there, and my previous post was meant to point out that he'd taken for granted a certain meaning of zero in his earlier response to an even earlier post, which I'm sure the poster of said earlier post didn't intend. <p>And of course, as long as we're talking bits, it's 1's and 0's first of all. Can't have a bit with another value than those. Bytes are a whole other story.<p>

Click!

[ellis_vener_photography]

<I>"Inserting 0s" implies that Photoshop inserts additional pixels with 0 value, which

would be up-rezzing with black pixels inserted between existing pixels </I><P>It seems I

wrote badly. While that is one way to read what was written Peter I was not intending to

imply or say that. the zeros are added as place holders t otel lthe program that it is

dealign wth a 16 bit (per channel) file. In much the same way as settling on a 4 place file

numbering system where the 12th file in a series is represented by 0012.