Master calibration frames

[JGMoreau]

Quite a basic question about master calibration frames :

Should they be .fit or .tif ?

Can i use a combination of both ?

Thanks,

Jean Guy

[Josh Lake]

Your best bet is to use 32-bit FITS files for everything...

Unless you're shooting with a DSLR? And then the raw images will be TIFs.

[JGMoreau]

Thanks for the reply.

I am using a ST2000XCM camera wich is one shot color with bayer matrix
and acquisition with CCDSoft which gives either SBIG proprietary images,
that i carefully avoid   
or .fit that i will stick with from now.

Jean Guy

[Harry page]

Hi

Sorry I disagree , use 16 bit fits for calibration it is only when you start processing do you need 32 bit , all you will get with 32 bit file at calibration stage is very large files which will add nothing 

Harry

[Nocturnal]

Let's keep the terminology straight Source files (light, dark, flat, bias, flat dark etc) should be in native format. Lacking that a 16b TIFF or FIT is fine. The camera outputs integer data and you do not gain anything from sticking that into a much larger float.

*Master* frames should be stored as 32b Float images. I recommend TIFF for that as it's a less ambiguous format but you can be sure that if you use PI to save as FIT in 32b float PI will be able to read it again fine and use it as a master frame.

[Harry page]

Hi

Why do you need to store master frames as 32 bit ( i Presume you mean darks bias flats ) there is no gain here

Harry

[Nocturnal]

Sure there is Harry. The average of 4500 and 4501 is 4500.5, right? Would you really want to throw away that 'painfully' obtained accuracy in your master?

[Harry page]

Hi

Why not save in 64 bit then for the extra millionth    I know what you say but I see no real world differance 

Regards Harry

[Nocturnal]

Hi,

It's all about the minute brightness differences, especially on the very dim and very bright ends of your image. You start truncating your masters at whole integers and those small differences all disappear. Goodbye HDRMT, goodbye aggressive stretch in HT without posterization.

But, if you don't see a difference, by all means use 16b masters.

[Harry page]

Hi

Would be interested in others input on this  me calibration 16 bit , processing 32 bit 

Harry

[troypiggo]

Personally, I use 32 bit as well.  Totally agree with Sander.  Everything is 32 bit FITS right up until final output when I convert to a 16 bit TIFF ProPhoto RGB master finished image (old habit from PhotoShop).  Then for web display it's converted to sRGB JPG.

Keep the higher quality files up front, then once you've done all the heavy stretching and number-crunching, right at the end of the process you can start reducing the bits.  Less chance of losing quality.

Keep in mind that just because you can't see a difference, doesn't mean it isn't there.  When I first started astro, I wasn't even using PhotoShop, I was using GIMP.  8 bits.  Stretching the histogram in that really highlighted the necessity of at least 16 bits that PS offered.

I know you're talking about master calibration files, not the actual images here.  I think the philosophy still stands.  As Sander mentioned, it's all about the numbers in between.  Start with the highest quality that is reasonable up front, for all files, then right at the end you can start reducing it down when it's less likely to impact quality.

[Juan Conejero]

Definitely, the 32-bit floating point format is advisable for all calibrated frames, and is really necessary for all master calibration frames. A master calibration frame is the result of an integration - it simply cannot be represented correctly with 16 bits per pixel sample.

[Harry page]

Hi

Ok lord Juan has spoken and I am wrong     , well there had to be a first time 

Harry

[Nocturnal]

Hi,

I'm glad I was able to help make it happen 

[Harry page]

Hi,

I'm glad I was able to help make it happen 

I feel your love 

Harry