My usual astrophotographic process

I use short enough multiple exposures that drive errors will not be a problem in the final combined image (GRIP gives me a record of the drive errors between frames so I can check). I capture the photos on a (large, fast) compact flash (CF) card in the camera. After one of my observing sessions there will typically be 300 to 400 RAW (.cr2) files of about 26 Mbytes each (my camera has over 21 megapixels at 14 bits per channel, RAW files use lossless compression and the Bayer filter pattern saves us 67% anyway - there is really only 1 channel in a RAW file). I may sometimes have to swap to a second CF card in the camera. All subsequent steps are done indoors the next day.
I transfer the RAW images to my PC using the standard free software from Canon - EOS Utility. Plugging the camera in with its USB cable kicks that off automatically and puts the images in a standard place on the PC, in an automatically created subdirectory named by the date when the photos were taken (eg, 2010_12_27). It takes about 10 to 15 minutes to transfer the few hundred files that I would typically have captured.
(If jrawio were up to date, so GRIP could read the RAW files from my newish type of camera directly, I would not need this step.) I open the new directory in another free application that comes with Canon cameras: DPP (Digital Photo Professional). I select all the thumbnails and batch process them to convert them to 16-bit TIFF format at 300dpi, without embedded colour profiles (so they will be simple RGB images). This step takes getting on for an hour on my 4-processor Windows 7 (64-bit) laptop. I set DPP to put the files in a subdirectory called RAW.
(This step could also be done in DPP but I find it more convenient to use Adobe Bridge - part of PhotoShop, of which I have version CS4.) I look through all of the images in filmstrip mode (stepping through thumbnails I see each one larger above the strip). This enables me to do 2 things. First I can find which frames are the beginning and end of a multi-exposure sequence that will be combined to make one photo. Secondly, it enables me to identify certain frames that should be omitted from the combination process for various reasons. Typical reasons include: obvious meteor or satellite tracks, blurring due to telescope movement (eg, on a windy night), fleeting cloud fragments (often seen as brighter than usual background). I record frame numbers and relevant metadata (exposure settings, start and end times) in a spreadsheet so I can label the photos accurately on my web site after processing.
I move each multi-exposure set of TIFF files, apart from any visually rejected ones, into a subdirectory of its own, named after the subject (eg, M45 or CasAlpha). This makes for easier management for the subsequent steps and for archiving later.
I run my own GRIP program. In the batch menu I select "Astro combine into 1 image" (or sometimes the comet option). From the dialogue I use the "Browse" button to select the subdirectory containing the set of TIFF files, Ctrl-A to select all of them and proceed. The "Proceed" button of the main dialogue then starts the process. It takes about a minute per frame to combine them all into one, at the end of which 2 files are saved: a .accum file (my own format, that only GRIP can read) containing the full 32-bit-per-channel accumulator data, plus a linearly scaled 16-bit-per-channel TIFF file (any photo application can read). At this stage I do NOT use the optional look-up curves to read out from the accumulator - I want to keep the result of the image warping and accumulating process so I can always go back to that.
At this stage I can delete the input TIFF files and just keep the 2 result files (.accum and .tif) in the subdirectory. I will later archive the RAW files but there is no point in also archiving the much larger TIFF versions of the same images (124 Mbytes each).
If the purpose of the photo is to do photometry then that's it. We must not alter the contrast in any way before opening the TIFF image in GRIP again to estimate magnitudes. However, if we want the best presentation of deep-sky nebulae then there is more processing to be done, as in the following steps.
Reopen the .accum file in GRIP. Doing this also creates a window showing a histogram of the contents of the 32-bits-per-channel accumulator. It is usually possible to see quite clearly where the background peak is in each channel (red, green, blue). Then it is possible to set a look-up curve that extracts the most from the data. Save the result as a new TIFF file (not overwriting the other one from the combining). I usually name this file xxxx_acc.tif to show its separate extraction from the accumulator.
Open the new TIFF file in Photoshop.
Use curves to reduce the red and make the background as neutral as possible. There is more need for this step in the more light-polluted observing sites.
On the filter menu select Sharpen/Unsharp mask. In most cases select the largest possible radius (250px in Photoshop CS4) and experiment with the percentage somewhere around 150%. The result of this operation is often quite dramatic. NB: Do not use the other end of the scale, with just a few pixels radius, as you would for everyday (non-astro) photos.
That is usually enough. Save the result as another TIFF file, in case a print (at 300dpi) is ever required. Also convert to 8 bits per channel, crop/resize to something smaller at 72dpi, and save as JPEG for use on the web. (There is no point in having an image more than about 800px across for display in a browser.)