The benefits of 12-bit colour over 8-bit or jpeg files are clear; it avoids the limited colour levels that exist in 8-bit when reshaping the exposure with curves to bring up shadow detail. I was considering converting some 8-bit files to 16-bit to avoid this posterization problem. However, what happens when one goes to print the file?
If the printer and driver only support 8-bit per colour channel, then the lowest 4 bits will be discarded either explictly or in a conversion to an 8-bit print file and it would seem posterization would re-appear in the final print.
Are there 12-bit per colour channel printer drivers included in tools such as CS2? Do high end printers support 12-bit colour resolution?
Andy
Full Version: 16-bit printer driver?
> converting some 8-bit files to 16-bit to avoid this posterization problem
Absolutely. Even a file that either started life 8-bit per channel or was dumbed down to 8-bit per channel should be bumped up to 16-bit per channel for radical editing. But then you (or the app) have to dumb down to 8-bit per channel again for printing or monitor display.
> If the printer and driver only support 8-bit per colour channel, then the lowest 4 bits will be discarded either explictly or in a conversion
Not quite. As I understand it, no complete bit range is discarded; it's a matter of rounding and averaging. The gamut is not clipped; it's just made somewhat coarser. The result is not visible to the human eye; it's just that 8-bit editing is more likely to lead to visible changes. Think of a piano: it has 88 keys covering just over 7 octaves. A quarter tone piano would have 176 keys; there would still be a 7 1/3 octave range and the lowest note would still be the same growling bass note and the highest key would still be the same tinkling high note. What would be different is that the same gamut would be divided into smaller chunks.
> Are there 12-bit per colour channel printer drivers included in tools such as CS2? Do high end printers support 12-bit colour resolution?
To the best of my knowledge: no and no. Yes; 8-bits sounds rather lame; but we're not really dealing with 8-bit files; instead, we're dealing with 24-bit files (8x3 channels) for something like an 6.7 million colour palette. Inkjets ultimately have to deal in discrete droplet sizes, dot gain, etc. So far as I know, if there is a qualitative difference between low and high end printers it's a matter of the low end printers not being able to reproduce all the subtlety of 24-bit colour. And of course different inksets and paper combinations have smaller or larger overall gamuts.
Absolutely. Even a file that either started life 8-bit per channel or was dumbed down to 8-bit per channel should be bumped up to 16-bit per channel for radical editing. But then you (or the app) have to dumb down to 8-bit per channel again for printing or monitor display.
> If the printer and driver only support 8-bit per colour channel, then the lowest 4 bits will be discarded either explictly or in a conversion
Not quite. As I understand it, no complete bit range is discarded; it's a matter of rounding and averaging. The gamut is not clipped; it's just made somewhat coarser. The result is not visible to the human eye; it's just that 8-bit editing is more likely to lead to visible changes. Think of a piano: it has 88 keys covering just over 7 octaves. A quarter tone piano would have 176 keys; there would still be a 7 1/3 octave range and the lowest note would still be the same growling bass note and the highest key would still be the same tinkling high note. What would be different is that the same gamut would be divided into smaller chunks.
> Are there 12-bit per colour channel printer drivers included in tools such as CS2? Do high end printers support 12-bit colour resolution?
To the best of my knowledge: no and no. Yes; 8-bits sounds rather lame; but we're not really dealing with 8-bit files; instead, we're dealing with 24-bit files (8x3 channels) for something like an 6.7 million colour palette. Inkjets ultimately have to deal in discrete droplet sizes, dot gain, etc. So far as I know, if there is a qualitative difference between low and high end printers it's a matter of the low end printers not being able to reproduce all the subtlety of 24-bit colour. And of course different inksets and paper combinations have smaller or larger overall gamuts.
QUOTE(AndyF @ Jan 10 2006, 05:53 PM)
If the printer and driver only support 8-bit per colour channel, then the lowest 4 bits will be discarded either explictly or in a conversion to an 8-bit print file and it would seem posterization would re-appear in the final print.
No, this is wrong.
The colours are, depending on your choices in preparing for print, re-mapped to the printer's colour space.
There is no one-to-one conversion from the image you work with to the image you get on paper.
Also keep in mind that your monitor or monitor driver most likely isn't able to support more than 8-bit per colour channel. Do you see any particular posterization because of this?
I found the hereunder included comment made by Jonathan in another thread ( here ) very useful and clarifying.
( [bold] is mine)
( [bold] is mine)
QUOTE(Jonathan Wienke @ Dec 26 2005, 07:36 AM)
....
First of all, a 7600/9600 has no trouble printing color from a 16-bit image file. I just sold my 7600, but have printed thousands of square feet of color images from 16-bit files. Others here have done the same. With the default Epson driver, the printer data is indeed downsampled to 8-bit before being sent to the printer, but since this happens after the color space conversion from the editing space to the printer profile, one still derives a benefit from a 16-bit source file over an 8-bit. Every possible 8-bit printer output value can be used when the source is 16-bit, but when converting from 8-bit to 8-bit, this is not the case.
There are some rather compelling advantages to editiing in 16-bit mode as much as possible prior to printing. One can do much more aggressive level, curve, and other adjustments to an image before banding or posterization will become evident in the image. For a practical example, see 16-Bit Vs. 8-Bit Workflow and download the TIFF version of the test image, and run the action on the image in 8-bit mode and 16-bit mode if you doubt the results shown. That you even ask the question is an exhibit of considerable naiveté on your part.
First of all, a 7600/9600 has no trouble printing color from a 16-bit image file. I just sold my 7600, but have printed thousands of square feet of color images from 16-bit files. Others here have done the same. With the default Epson driver, the printer data is indeed downsampled to 8-bit before being sent to the printer, but since this happens after the color space conversion from the editing space to the printer profile, one still derives a benefit from a 16-bit source file over an 8-bit. Every possible 8-bit printer output value can be used when the source is 16-bit, but when converting from 8-bit to 8-bit, this is not the case.
There are some rather compelling advantages to editiing in 16-bit mode as much as possible prior to printing. One can do much more aggressive level, curve, and other adjustments to an image before banding or posterization will become evident in the image. For a practical example, see 16-Bit Vs. 8-Bit Workflow and download the TIFF version of the test image, and run the action on the image in 8-bit mode and 16-bit mode if you doubt the results shown. That you even ask the question is an exhibit of considerable naiveté on your part.
Few drivers can support and use more than 8-bits per color. One example is ImagePrint from ColorByte.
The interesting point that Jonathan made, if true, is that there is some benefit to be had in sending a 16 bit file to the printer because the downsampling takes place after conversion to the printer profile.
The question is, would such differences be visible with the naked eye?
The question is, would such differences be visible with the naked eye?
Here is another way to think about this,
If the typical photographic print needs about 400-500 distinct brightness levels to seem continous for a smooth black to white gradient, the 256 brightness levels of 8bit would seem as if not being enough.
However,
Photoshop CS2 does have a dither option for color conversions to 8-bit color, so I would assume that if Photoshop is doing the color conversion for the printer, it would be dithering the 16-bit data of the RGB working colorspace into the 8-bit data of the paper colorspace.
If the dithering algorithm is good, and source material is at a high enough pixels per inch, then sending 16-bit to the printer should not be necessary.
---
> Do you see any particular posterization because of this?
Jani, I definatly do, even with a perceptually color calibrated monitor. Perhaps when Intel puts out a 10 Ghz chip, Adobe with think about dithering the 16bit data onto the 8bit screen for soft proofing...
If the typical photographic print needs about 400-500 distinct brightness levels to seem continous for a smooth black to white gradient, the 256 brightness levels of 8bit would seem as if not being enough.
However,
Photoshop CS2 does have a dither option for color conversions to 8-bit color, so I would assume that if Photoshop is doing the color conversion for the printer, it would be dithering the 16-bit data of the RGB working colorspace into the 8-bit data of the paper colorspace.
If the dithering algorithm is good, and source material is at a high enough pixels per inch, then sending 16-bit to the printer should not be necessary.
---
> Do you see any particular posterization because of this?
Jani, I definatly do, even with a perceptually color calibrated monitor. Perhaps when Intel puts out a 10 Ghz chip, Adobe with think about dithering the 16bit data onto the 8bit screen for soft proofing...
That point had occurred to me after posting - that the monitor is also only 8bit/channel, yet smoother transitions are visible. Since the colour spaces of the image file and display (or printer) do not directly map to each other, having a larger source file should allow more precise mapping of colours to the 8bit/channel file. The 8bit/channel colour space should be large enough except where an image contains graduations of a single colour channel, then it would show it's 256-level limitation, if that's discernible?
The quote from Jonathan is interesting. Delaying the 16-bit to 8-bit conversion until after the change to the printer colour space instead of converting to 8-bit then printing avoids another mapping operation, especially one that occurs in 8bit/channel space where correct mappings will be more difficult. This leads back to needing a 16-bit printer driver, if as Ray says those differences would be visible. Now, if the image editor sent the file already converted to the printer's space, no remapping at all should be required in the printer driver, and this has been solved.
I'll have a look at ImagePrint to learn a bit more from them, and your website Andrew.
Andy
The quote from Jonathan is interesting. Delaying the 16-bit to 8-bit conversion until after the change to the printer colour space instead of converting to 8-bit then printing avoids another mapping operation, especially one that occurs in 8bit/channel space where correct mappings will be more difficult. This leads back to needing a 16-bit printer driver, if as Ray says those differences would be visible. Now, if the image editor sent the file already converted to the printer's space, no remapping at all should be required in the printer driver, and this has been solved.
I'll have a look at ImagePrint to learn a bit more from them, and your website Andrew.
Andy
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.