I agree with Marko. In RGB you can put a tone curve on a B&W print.
I agree with Marko. In RGB you can put a tone curve on a B&W print.
Actually, the information is triple. The number of potential colors is much greater than triple, but the information is triple.
However, for B&W, your numerical rationalization is not correct. The increase in meaningful information held in a RGB grayscale image is only about 1.5 bits greater than a comparable grayscale file.
This is again not entirely true. You have slightly more meaningful grayscale information in an RGB, and for the most part it equates to more dithering in the file, but since there is a huge amount of overage when working at the 16 bit level, I seriously doubt it makes any beneficial difference in the final print.
I do think this would be a very good approach to get a higher bit depth if you had 8 bit files to work with, but at 16 bits, I am doubtful it will have any benefit.
This again, is not correct. Think of it this way. You're saying that you get more information in a grayscale image out of 3 channels of an 8 bit file than you will get out of a single 16 bit grayscale file.
You can't think of a B&W image as unique colors, you have to think of it in terms of unique levels. In this context, a single channel 8 bit image will have 256 levels possible. A 9 bit image will have 512 levels, possible. That will coincide exactly to that number of possible levels from a file that has two channels and are 8 bit. So a three channel B&W 8 bit image will have essentially 1.5 more bit depth than an 8 bit file.
---Michael
Actually, this is not correct either.
One bit is the the most basic, discrete piece of information, therefore, any amount of information can only be a full multiple of one bit. There is simply no such thing as "1.5 times the bit depth" or "1.5 bits greater", as it makes no sense.
Yes, a file with a single 16-bit channel carries much smaller number of discrete pieces of information - bits - than a file with three 8-bit channels. It's a matter of simple math - the former is a 16-bith file, while the latter is a 24-bit file. The information is cumulative, not additive, so a file containing one 9-bit channel from your example will indeed contain 512 discrete pieces of informaton (2^9), but the file with 2 8-bit channels carries the same amount of information as if it had a single 16-bit channel - (2^8)^2 = 2^16.
There is also no such thing as more or less "meaningful" information. Information on this level is binary, it either IS or it IS NOT.
What I was saying is that using three channels is effectively like using a file with a bit depth 1.5 higher than not having three channels. While a bit is the least increment in a single channel, as soon as you have three channels, you effectively can produce partial bits of information when compared to a single channel file, so my statement is correct.
You are misunderstanding meaningful differences in a file with raw data permutations. When you equate all the possible color combinations in a RGB file with the meaningful information in a grayscale file.
What do I mean by this? Several things. First, if you scan a B&W negative in color and invert it to make a positive, you will have a color image of a black and white source, so the image will probably look more or less B&W.
Take a particular level in that image (lets use 8 bit for the example). Let's take level 128. If you look at a pixel that has a value of 128 (luminance or in a converted grayscale file), you might see a RGB or something like 125,128,131, or so. The three channels will not be permitted to deviate too much from the neutral (nominal) value of 128. So even though you have the potential for a large variety of RGB combinations that equate to a value of 128, there are really only a limited number that will actually occur due to the neutrality of the source negative and the way the the source is mapped to individual RGB values during the scanning.
In other words, if you have a composite value of 128, it is not going to be possible to have a B of 0, and R of 255 and a G of 128, because the neutrality of the negative won't have that kind of deviation happening within it.
The other issue has to do with the compression of all the colors in a color file into levels in a grayscale file. Even if you had the ability to get a scan that had the kinds of variations I mention above, the color engines effectively make many of the possible
color permutations in a color file numerically equivalent in terms of luminance or K values. So while it's possible to have a RGB of 127,128,129, and while this is a different color than another like 129,128,127, the converted grayscale levels of these two numbers could be identical within the bit depth of the file. There are many such permutations that are rendered nemerically equivalent in grayscale.
In other words, while there is a color difference in the file, there is no meaningful levels difference between these two numbers. That's what I mean by meaningful information differences.
(Note that I am not equating these values, I pulled them out of the air as examples.)
---Michael
OK, I see what you mean now and it does make sense.
What I have in mind talking about bits of information is the actual amount of data loss in processing as well as in the translation process during printing. In other words, even though some, or even a good part of it may be redundant, it still provides a much bigger pool to lose - and keep - data from, so that the actual amount of "keeper" data we end up with remains as close to the maximum possible to have in a final 8-bit file.
Marko,
You're absolutely right, and I think that for 8 bit scans or files, that approach is going to go a considerable way toward knocking out the posterization problems that can be experienced. Its a great approach for a person with an old drum scanner, for example.
For high bit depth images, I think there is enough overhead in the images toe avoid having to go to these kinds of heroic measures.
That said, printing color images of B&W files when there is a particular toning that is desired is a very effective approach for some people. However, the file does not have to be RGB throughout for this approach if it is a high bit depth file. Probably can't hurt, though. It simply means working with larger image files.
---Michael
I scan and manipulate in PS in 16-bit, then convert to 8-bit before sending to the printer.
Hi all,
Great advice guys... thanks for all the terrific information. It's much appreciated.
Bruce,
I'll get ahold of the lab and make the necessary inquiries.
Again, thanks for all the great advice.
Cheers
Life in the fast lane!
Capocheny,
In terms of B&W it breaks down like this...
8 bit B&W has 256 levels of grey (including black and white) which is about what the human eye and brain can work out. 16 bit B&W has a lot more than that (were talking tens of thousands here).
If you plan on printing a scan without manipulating the file then there is no reason to go 16 bit. If however you want to move tonal values around in Photoshop before you print it is better to have a 16 bit scan done. This way you have less of a risk of posterization.
Once you are happy with how your 16 bit file looks then convert it down to 8-bit and save a version to send to the lab.
By the way. Who is scanning and printing your work these days?
Regards,
DL
Last edited by Dominique Labrosse; 26-Feb-2007 at 23:26. Reason: clarity
Hi Dominique,
Thanks for the info...
One of the guys from last years workshop is scanning some of my negs until I pick up my own V750, which will be sooner than later.
G. King Photo is printing my stuff these days. I really need to get into doing my own printing and contact printing though. Soon!
Cheers
Life in the fast lane!
Bookmarks