Understanding 16-bit versus 8-bit processing (Software)

[Macavity]

As an 8-bit ("Two-bit"?) Video Astronomer, I am well familiar with the limitations 

of 8-bit camera output (As opposed to 16-bits) when processing image data! But

one of the main obstacles is that 16-bit software (Photoshop) cost big money.

But as an avid reader of Imaging tutorials in magazines, I am now more aware

that there are very many ways in which (you) classical imagers process data.

The most crucial element - Notably in 16-bit processing, *seems* to that initial

STRETCH of data. And I see quite a few folk use (eminently free) IRIS for that! 

Provided you save the data in suitable exchange format the "worst bit" is over?

Once you have constrained your data into a *reasonable* length histogram, it

may be that additional adjustments - (minor) "levels", "curves" are fairly safe?

I am fairly adept with GIMP and now use it for me (new!) Photography hobby.

Of course, much of the above is just me talking through my "cat flap", but...

I guess the big advantage of Photoshop is that magazine examples use it! 

Moreover, those who write useful *macros* tend to do likewise. Which brings 

to a final question... What is / isn't 8-bit / 16-bit? I sense *standard* GIMP is 

8-bit (with future / beta 16-bit versions?). But what of Photoshop Elements?

OR maybe eight bit (whatever) is all you need for final "photo tweakings"?

Comments (Even flames!) - Just so I can better *understand* these things! 

P.S. I note you can "mess around for hours" for indifferent results. Then

you discover something like "contrast masking", "contrast blending" etc.

Sometimes creating "brilliant" results - even with humble 8-bit images...

[michael.h.f.wilkinson]

I only do conversion to 8 bits right at the end when I want to print stuff. Any modification of the 8-bit histogram after quantization from 16 down to 18 bits can give nasty artefacts, especially in the low-brightness areas. BTW, ImageJ is free and handles 16 bits or even 32-bit float nicely.

[Alien 13]

When I started I used DSS to stretch the Image and equalise the RGB levels then saved to 8 bit for processing with PS elements and to be honest cant see any difference to a 16 bit workflow.

Alan

[Macavity]

Thank you Michael for taking this question (waffle) on! I see your point. Maybe I will experiment

with my 8-bit video data - There is *less* to lose? I will indeed have a look at ImageJ - A new toy

to play with is never bad? Sometimes I do think I made a very big mistake a LONG time ago tho:

Had I cloned the exact setup of (most?) classical imagers, I would be doing far better by now -

Rather than still "wondering" such stuff. I could have save the requisite money over time too.

No bitterness, just a wry comment. I suppose it's all been "good for the soul" or something?

I sense we will all (at least in Northern Europe) remain prey to the prevailing weather though...

[Macavity]

When I started I used DSS to stretch the Image and equalise the RGB levels then saved to 8 bit for processing with PS elements and to be honest cant see any difference to a 16 bit workflow.

Alan

One the other hand, that's rather interesting too...

(And maybe just what I was thinking about)

[michael.h.f.wilkinson]

I often capture my solar data at 8 bit, but stacking gives me 16bit results. If I first quantize to 8 bits, and then perform any look-up-table operation or sharpening I get results that are not as good as doing the sharpening and look-up table manipulations in ImPPG at 16 bits per pixel, and then quantize. ImPPG is quite specialised, but does a good job (in grey scale) of processing 16-bit data

[ollypenrice]

I'm going to be honest here. I was a late convert to 16 bit CS3 and previously posted images processed in an older 8 bit Photoshop. I don't find the difference as radical as others seem to. Recently I must have made a false click and done a lot of processing in 8 bit before spotting the error. When I repeated the steps I was hoping to see something special, but it didn't happen. Obviously I use 16 bit but I'm a little bemused by this.

Olly

[michael.h.f.wilkinson]

In many cases the differences will not be obvious to the eye, but I can get more of the fine detail in prominences in 16 bit than in 8

[glowingturnip]

I've got Photoshop Elements (never felt the need to go for the full thing) - it handles some things in 16 bit, but insists you drop down to 8 bit for other things (like adding a layer - so most things then).  I only use it for very limited parts of my processing, like making a specific mask, and do the rest in PI with 32 bit or 64 bit floating 

If you've got a lot of data that's compressed into a narrow range of values, as you have in a raw file before stretching, then if you stretch that with too low a bit depth, you'll get posterization start to happen which can lead to ugly backgrounds

[dph1nm]

I stretch in DSS, then process in 8-bit GIMP. Works fine (although there is now a development 16/32 bit GIMP version).

NigelM

[riklaunim]

Planetary cameras record at 8-bit and can't even take advantage of more than 8-bits as the signal to noise ratio for such exposures isn't high enough. But 8-bit AVI is stacked to 16-bit stack (like a TIFF) so that it gets most of the SNR increase possible and much more problem free processing.

For DS imaging I use Atik 314L+. It captures 16-bit FITS files but it's worth to know that it can't give more than around 14-bit of data range on those single frames (sensor isn't capable). Yet for stacking is much better to have it as 16 bit.

[Macavity]

Thanks for the great responses here! I feared this might be a "daft" question.

Perhaps [Olly] you're onto something?! I sense, if you have good *initial* signal

to noise ratio, that is much of the battle. Much as I love Video Astronomy, if you

want to post-process, there is very little leeway with 8-bits. Moreover, cramming 

a lot of electronics into a small volume (like my Watec camera) doesn't help.

Those (few) who do experiment (and document results!) with Video cameras 

seem to find keeping "gain" settings low, switching "gamma" to OFF is best!

Though as testament to Watec, I have almost burnt my finger on the camera

during a white light solar session! I now do use a scope mounted sun shade.

I am tempted by Photoshop Elements - Thanks for info Stuart (glowingturnip)

Intrigued by Pro Digital's "Astronomy Tools" (PS macros?) processing suite.

But I sense I can do many such things myself... Given sufficient application!

My biggest problem (Video astronomy again!) is [heat] "Gradient Removal".

For that, many people seem to use "GradientXterminator"... The dialogue

boxes therein do look remarkably similar to those within the IRIS software?

(No aspersions cast. Maybe such things are general / licensed plugins etc.)

But thanks ALL for answering some questions I *often* "wonder about"...  

And do keep the random thoughts coming! Such things are a joy here.

[ollypenrice]

Yes, I tend to have very high SNR becaue 1) I have a dark site and 2) I'm a nutter and take barking amounts of data. It could be that.

Olly

[Davey-T]

Even in P'Shop some things only work if you first convert to 8 bit.

Dave

[Macavity]

(To Olly):

1. I unashamedly envy you that one. lol  

2. I genuinely admire such dedication...  

[glowingturnip]

another annoying omission from Photoshop Elements is a proper curves tool btw, and any kind of channels manipulation