Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rate Thread
  #1  
Old 03-08-2011, 12:38 PM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,828
FITS, bit depths, signed/unsigned, let's talk

Trying to get my head around bit depths and flicking my FITS files between different software. I have a basic understanding of what it all means, I know the higher bit depth, the better for stretching and maintaining data. I understand there's no point converting an 8 bit image to 16 bit because you don't gain anything, and I understand that dropping a 16 bit to 8 bit could result in data loss if you try to stretch afterwards etc.

My understanding is that my CCD camera (QSI583, KAF-8300 sensor) captures 16 bit data, and I assume that's unsigned (0-65535). So that's what my individual subs will be.

Different software may handle FITS files with different bit depths internally, I think MaxIm and PixInsight are 32 bit (unsigned?). Apparently once you stack your 16 bit images, you can increase the bit depth, presumably because you now have much higher signal to noise ratio?

It seems that the documentation in both MaxIm and PixInsight recommend once stacked you could/should save as 32 bit IEEE floating point FITS, particularly for big stretches etc.

Then for output to, say, Photoshop, you'd save as 16 bit (signed/unsigned?) TIFF because that's the maximum it can handle. And after that, saving for web it'd go to 8 bit JPG for example.

So, am I on the money above? What bit depths do you operate at? Do you find 16 is enough, or is 32 worth it? 32 bit signed, unsigned integer, or IEEE floating point? Benefits of one over the other? I notice PI does 64 bit IEEE as well, anyone go that high?
Reply With Quote
  #2  
Old 03-08-2011, 04:50 PM
lepton3's Avatar
lepton3 (Ivan)
Registered User

lepton3 is offline
 
Join Date: Jan 2007
Location: Adelaide, Australia
Posts: 670
Troy,

I think you are pretty much on the money.

I would just add that the IEEE 32 bit format has a 23+1 bit mantissa, so for data with limited dynamic range, (eg a stack of unsigned 16-bit integer images) it is like an int-23 for quantisation noise, so potentially not as good as int-32, but in reality they may be pretty similar.

Definitely, a 16-bit result of stacking 16-bit subs has way more quantisation noise than either an int-32 result or a float-32 result, so I would avoid converting to int-16 until after the curves have been adjusted.

And be sure to save in a 32-bit format in case you want to come back and tweak the curves later!

-Ivan
Reply With Quote
  #3  
Old 03-08-2011, 07:14 PM
jase (Jason)
Registered User

jase is offline
 
Join Date: Sep 2006
Location: Melbourne, Victoria
Posts: 3,916
Good questions Troy

You're right, most CCD cameras will out FITS in 16bit unsigned format.

The 16bit space doesn't last long either as most image processing software will manipulate it to a higher bit space depending on functions utilised. A simple example of this would be using a sum combine function where the value of each pixel in each sub is an addition to the final pixel value. Clearly if the software left the data in a 16bit format it would easily clip values over 65k, hence the software automatically increases the space giving the user two options when it comes to saving the image, either save at a higher bit space such as 32bit or IEEE floating point format or scale the data down back into the 16bit space.

I personally don't sum data often, but even with other functions such as median combine its possible to have masters higher than the 16bit 65k limit. For this reason, if working with FITS along I typically leave the data in IEEE floating point format.

The challenge of course is interoperability between other software processing tools. FITSLiberator will allow 32bit FITS to be loaded into Photoshop, but many of the advanced functions can't be applied in 32bit mode, thus the file needs to be converted to 16bit. There are a variety of ways to scale data down, most tools automatically scale, where others you'll get better results manually performing the task with pixel math or log stretching. Depends on how much control you want over the data.
Reply With Quote
  #4  
Old 04-08-2011, 11:06 AM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,828
Thanks Ivan and Jase. Think I've got it right now. Cheers.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 07:47 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Celestron Australia
Advertisement
OzScopes Authorised Dealer
Advertisement
Lunatico Astronomical
Advertisement
NexDome Observatories
Advertisement
Meade Australia
Advertisement
SkyWatcher Australia
Advertisement
Astronomy and Electronics Centre
Advertisement