32 bit vs 16 bit - What's this all about?

[Scorpius]

I’m very new to image processing so trying to learn what I can through trial and error. One thing I’ve noticed is most capture software produces 32 bit images (Fit’s or Tiff’s) and it seems that Photoshop does a better job processing 32 bit Tiff’s once subs have been calibrated, aligned and stacked. However, when converting 32 to 16 bit - the difference is immediately noticeable in terms of noise, etc. and PS automatically defaults to HDR toning. From that point forward - no additional processing I’ve tried can bring the image quality back to where it was. Of course this makes sense - and I’d expect there to be some loss of quality – I’ve just been surprised how noticeable the difference really is. PS won’t let me create a JPEG directly from a 32 bit Tiff so I’ve been converting it to a 16 bit Tiff prior to saving it as a JPEG. Is this the recommended way to convert an original 32 bit image into a web ready JPEG and if so - are there any tips for maintaining image quality during the conversion?

Again - I’m a complete newbie so please don’t assume I know something most astro-imagers already would...

[D4N]

JPEG files can only be 8 bit per channel for a total of 24bit, this is all you need for display purposes as the human eye cannot detect the tonal differences in higher bit depth than that.

I don't use PS myself but the software I use has a distinction between saving and exporting, when you save you are saving what is on the screen and it won't let you destroy data by using the wrong file type. Exporting you can do whatever you want.

Sent from my iPad using Tapatalk

[NickK]

Additionally most laptop screens are 10-12 bit split between R,G&B too.

The largest image I've processed is 64bits per channel (around 6GB for the image). In this situation you use software to provide you with a perspective rather than see the full image.

Once you're stacking - it also depends on the form of stacking. Basic average stacking will only require the same sized bit plane however accumulative or speculative post-processing often requires expanding to prevent under or over run of data values within the dynamic range. HDR is one main form of processing where stacking requires expansion of the dynamic range of values.

RAW large image -> perspective (compression/stretch or other) -> render as JPG.

It's more fun to look at large images in 3D rather than colour.. you can see more in the range.

Edited May 7, 2015 by NickK

[Ruud]

An 8 bits channel allows for 2 ^ 8 = 256 brightness levels

A 16 bits channel allows for 2 ^ 16 = 65536 brightness levels

A 32 bits channel allows for 2 ^ 32 = 4294967296 brightness levels

Many monitors show three 8 bit channels which amounts to 256 ^3 = 16777216 colours.

[Scorpius]

Thanks to all for the replies but I must admit this is all a little bit (no pun intended ) over my head. Here are a few screen shots which show the results of trying to start in PS with a 32 bit tiff and then convert that image to a 16 bit tiff so I can end up with a jpeg. Although I can certainly do this – the image quality appears to suffer when converting from 32 to 16 bit and that loss of quality would logically carry forward to the final JPEG. My question was how to convert a 32 bit tiff to a jpeg with the least possible negative effect on image quality.

Here’s a screen shot of the original 32 bit file opened in PS.

Then here’s the dropdown save list but JPEG isn’t an option when starting with a 32 bit tiff.

If I try to rename the file and save as - I have to choose a bit mode and when choosing 16 - I get a pop-up informing me some tiff readers don’t support 16 bit floating point values.

If I try a different approach and change the mode directly from 32 to 16 bit – it allows me to do that but PS automatically defaults to HDR toning and the image quality is obviously (at least to me) affected in a negative way.

It would seem a JPEG created in this way could never be as good as a JPEG created directly from the original 32 bit tiff. Maybe it’s just that my understanding of bit modes and file types is so poor that I’m asking how to do something that’s simply not possible or even necessary?

[D4N]

Sounds odd, have you tried - file > export assets?

Sent from my iPad using Tapatalk

[Scorpius]

Sounds odd, have you tried - file > export assets? Sent from my iPad using Tapatalk

I appreciate the clarification and no – I had not tried that. Turns out my version of PS doesn’t offer the option to (file > export assets).

But I tried (file > scripts > export layers to files) which works but I need to create a duplicate layer since that option requires multiple layers to work.

This lets me convert the 32 bit image directly to a JPEG (without converting to 16 bit first) but it creates a separate JPEG for each layer. It seems odd it’s necessary to have two layers to convert the file – and also that it creates two separate JPEGS (one for each layer?) - but it seems to work so guess that’s all that matters...

Thanks for the help and clear skies!  

[D4N]

Well I'm glad you have a workflow that works, I'm surprised no one that actually uses PS has chimed in.

Sent from my iPad using Tapatalk

[melsky]

On your fifth image where it says "Local Adaption" click and try either Equalise Histogram or Highlight Compression, I actually use a combination of both for images stacked in DSS?

Mel