What issues are there with printing 32-bit HDR images?
- Do printing companies simply convert them to 8-bit?
- Do some have printers with a wider gamut that can use additional color data?
- Is there a better-than-8-bit but not 32-bit format?
It is simply impossible to print a 32bit image.
For the sake of truth it is impossible to view it on a monitor too :) ...
As far as I know there is no device being able to render 32 bit per channel, neither the human eye ... so if you'd be able to print you'll not be able to see it :)
Of course you can choose how to tone map your huge 32bit information into a smaller range, this is what is achieved by the tone mapping process. This is a very liberal procedure, I mean that you are free to choose what you want to render, what to enhance, exasperate and so on ... HDR + tone mapping is an "artistic" tool, not a photorealistic one IMHO.
@msw: HDR tone mapping is necessary both for RGB and CMKY, none of them can represent 32 bit of information.
Generally speaking, I think the format you want your images in for printing is 16bit. Most of the time, when I've taken in photos to be printed, I've always had them in 16bit per pixel (48bit) TIFF files, and used the Adobe RGB color profile. That seems to be pretty industry standard, however my experience printing with a third-party company is limited, so I can't say that definitively.
I always save a final "print" copy of my HDR images as 16bit TIFF images. Usually, Photoshop does not even give you the option to fine-tune the tone curve when Merging to HDR until you downconvert from 32bit to 16bit or 8bit anyway.
There is simply no physical way to emulate the gamut of additive luminance (light, your monitor) with subtractive pigments (CMYK). Not even close.
The best use you could make of an HDR image for print is to compress the luminance selectively while preserving borders between light and dark regions. This is a matter of great art and you'd do well to learn of Ansel Adams' darkroom techniques as he achieved the illusion of huge tonal range on a medium that didn't support it by dodging and burning.
In some sense, you'd wind up doing the equivalent of unsharp mask, but on a regional - not pixel - level. This fools the perceiver into thinking the shadows are as detailed as the mid-tones while having equal mean luminosity.
I believe that by "32 bit" you are referring to CMYK colors, that is 8 bits per channel with 4 channels. This is a color space that is usually used for offset printing, not for continous tone photographic printing.
Almost all images captured by a digital cameras or scanners are RGB colors and sometimes known as 24 bit color, 8 bits per channel with three channels. Confusingly the same images might be called 8 bit images because of the number of bits per channel. I will use the number of bits per channel terminology.
Pardon me if this all review for you, it's hard for me to be sure so I'm covering all the bases.
If you truly are referring to CMYK images I would not submit them to a service for fulfillment in this layout as almost all photographic hardcopy output devices use RGB colors (like laser printing on silver halide paper, inkjet, or thermal) converting between CMYK and RGB can be tricky and it is likely the results won't match your expectations.
Perhaps you are really asking about 16 bit images which are RGB with 16 bits per channel? I'll assume this is the case. Almost all prints made from digital images at various fulfillment services are done using 8 bit RGB images. If you submit a 16 bit image and it is not rejected the service will sample the 16 bit image to an 8 bit image. Typically (except for specialized applications) a 16 bit image allows you to have values from 0 (darkest) to 1023 (lightest). An 8 bit image is 0 (darkest) to 255 (lightest) the difference between the two is the number of intermediate steps available to be manipulated, not on the maximum lightness or darkness. The gamut of an image printed on an output device is driven by many factors: input color space, printer calibration, media, ink, minimum and maximum density achievable, etc. The number of bits in your image is a very small component.
Testing has shown that the human eye can only differentiate about 266 discrete levels in the green part of the visible light spectrum, about 200 in red and 150 in blue. A properly prepared and well-printed 8 bit image is capable of producing results indistinguishable by humans from a higher bit depth image.
High level advice for the highest gamut possible: capture your image in RAW and render it to a wide gamut color space like "Adobe RGB" or "Wide Gamut" ("sRGB" the default for most cameras encompasses the gamut of CRT monitors and is not close to the gamut available in a good hardcopy printer), carefully set white and black points in the image submitted, use a device with a large gamut (most pro level inkjet printers and laser silver halide printers do very well).
All that said it is my anecdotal opinion that most non-professional services are likely to use uncalibrated or poorly calibrated and poorly color managed devices which will override any of the careful preparation outlined above. You get what you pay for.