I’ve taking about 50 pictures of the Orion nebula, each exposure 20 seconds long at base iso. It’s only a 135mm lens so there is nothing exceptional but it’s the best I can do. I intend to stack them using another program, but this program needs tif files as input.
Rawtherapee easily batch convert my rawfiles to tifs, no problem there, but the question is if there is any thing I should do, or not do to preserve every little bit of information that might hide in the dark parts of the rawfiles. For instance color temperature, is daylight usually the best as to not cut/push any information out? Should I move the black point? Colorspace?
Any hints appreciated.
Edit, after posting this I saw the “Save unclipped images” post, that might just be what I was asking for…
Save to 32-bit floating-point TIFF files using the new “Unclipped” processing profile. See Save unclipped images
… and now I just read your last sentence :]
If your other program doesn’t support 32-bit floating-point TIFFs, then apply the “Neutral” profile, perhaps apply raw CA correction, and no tone curve to keep things flat and dull. Adjust exposure compensation to keep data away from the edges of the histogram. Save to 16-bit integer TIFFs. And if your other program doesn’t support 16-bit integer TIFFs, then delete it.
Thanks, it does support 16 bit TIFF and it can output 32 bit floating-point tiff (but not read it for unknown reasons). Also it can read 32/64 bit floating point FITS files, but I guess that is not an option in this case.
You can convert the 32-bit fp TIFFs to FITS using some program which retains all data, e.g. ImageMagick compiled with quantum depth set to Q32. Maybe LuminanceHDR could do that as well. Just verify that clipping does not occur, else use the 16-bit TIFF workflow.
Photoshop can read both 16 bit integer and 32bit floating point TIFF so I saved two versions of the same image and compared them, and as far as I can tell the 16 bit version held the same information as the 32 bit version. So I will stick with 16bit at the moment.
However, color balance and demosaicing method might need a little bit more testing. LMMSE seemed surprisingly good at a first glance.
This is likely a pedantic point, as 16-bit data will probably allow sufficient precision to pass between programs without visible impact. But, what you’re really trying to do is preserve the precision of the edits done with the first program, and the internal 32-bit float image will definitely capture more precision in those edits than a 16-bit internal image. Saving the internal 32-bit image to a 16-bit TIFF will lose some of that precision. And, even if the downstream program uses 32-bit images internally, the precision of the previous edits is lost.
Between 16-bit integer and 32-bit float data, the outcome of single edit operation will likely not show visible difference. It’s in the cumulative effect of edits that the final product stands more of a chance of showing difference that you care about. Me, I’d prefer to:
Translate my 16-bit integer raw data to 32-bit floating point at the earliest opportunity (that is, choose a raw processor that works raw images in 32-bit float) ,
Keep that data in 32-bit float for the entire processing pipeline, through whatever programs will edit it, and then,
Convert to whatever integer output is desired/needed for rendering to the medium to be viewed - JPEG, display, etc.
In principle I would like to do that too, but at the moment I guess that the choice of demosaicing method will have a much greater impact on the final result.
Not to mention my photoshop skills, which I thought where good enough until I saw an ADVANCED tutorial. Apparently I knew nothing about astro editing.
