Author Topic: 32 Bit v 16 Bit, which one and when?  (Read 649 times)

Offline STAstro

  • Newcomer
  • Posts: 43
    • View Profile
32 Bit v 16 Bit, which one and when?
« on: 2018 February 23 12:41:25 »
Hi All

I understand the need to leave your calibration masters etc in 32 Bit Floating, but what abount once you have combined channels etc?  The reason I ask is a single RGB image is now taking up almost 1GB of space, and I get quite frequent "Hangs" in PI when doing things like MMT Previews etc, especially when using Drizzle!

So once the image is in RGB, can it be reduced to 16Bit for Processing?


Offline John_Gill

  • PixInsight Old Hand
  • ****
  • Posts: 345
    • View Profile
Re: 32 Bit v 16 Bit, which one and when?
« Reply #1 on: 2018 February 25 23:00:05 »

I would always work in the highest resolution possible.  The "Scale" setting of 2 in Drizzle Integration will double the size of the image.  I think your issue is caused by your PC having insufficient CPU power and not enuff RAM.  My laptop is an i5 with 8GB RAM and PI seems to "hang" when doing very memory/processor hungry routines like MMT and TGVD etc, but I just wait for the process to complete.    There are a few things you can do to speed up PixInsight, like edit "Edit" ---> "Global Preferences" ---> "Directories and Network"  and add swap files etc.   

When Pre-Processing, I create the master Bias, Darks and Flats and then to save disk space I delete those "RAW" files.  The Lights are then processed with the Bias, Darks and Flats to create the final Integration file.  Again to save disk space, I then deleted the "Calibrated, Cosmetic, Debayered and Registered" files.  Never delete the Masters, Lights and of course your Integrated file.

I hope this helps
Look up
APM 107/700 apo on CGX mount
ZWO Optics - Autoguiding
ZWO1600mm and filters
... when there are no clouds ...

Offline RickS

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1298
    • View Profile
Re: 32 Bit v 16 Bit, which one and when?
« Reply #2 on: 2018 February 26 01:41:50 »

You certainly can convert to 16-bit though you'll need to be careful what operations you do afterwards.  After all, a lot of image processing in other domains is done with JPG files which are 8-bits per channel.

I'd suggest that you at least stretch the data before converting to 16-bit.  That's an operation that will benefit from the additional precision.  Keep the 32-bit data and you'll have it if something goes awry later in your workflow.

As John suggested, updating your computer so it can handle the larger 32-bit images is a good way to go if it is possible.  I have done some large mosaics (> 100 megapixels) and not had problems on fairly old hardware.


Offline STAstro

  • Newcomer
  • Posts: 43
    • View Profile
Re: 32 Bit v 16 Bit, which one and when?
« Reply #3 on: 2018 February 28 06:30:39 »
Thanks Guys

I am using a Core i7 with 12GB Ram, I might consider upping that to 32GB RAM, I will test the same workflow on my MAC Pro

I also noticed a setting in PI for the number of threads, as a tip for you Rick, try dropping the number of cores, since my CPU has 4 cores, I have dropped this to 3 in PI, which leaves 1 core free which prevents the hangs, but takes a bit longer ot process