RAW files and 12-bit pictures ?

[Yvon_Rey]

I shoot RAW images since a few months and I’m really convinced to have made the right choise, in particular for the possibility of intervention at “early stages“ of post-processing. Howerer, the option of saving an originally 12-bit file in 16-bit or 8-bit format – for further treatment in Photoshop for instance – let me somewhat confused. Indeed, I don’t really see the differences in histograms of the (same) picture saved in 8-bit (256 tone levels) ou in 16-bit format (of these, only 12 bits used or 4096 levels). With 12/16-bit format, I have 16 more information (or tone levels) than in 8-bit file. Where are these supplementary informations ? Is’t possible to visualize ant to work work with these (always 256 levels on my histograms…) ? In concrete situations, how can I exploit these ressources ? Do I have to produce 2 or 3 pictures in RAW transformation module (each with different “expositions“), move artfificially the histograms bodies along X-axis (from shadows to highlights), and blend the images produced to obtain a full 8-10 dynamic range stops picture ?

I’m in a blur since a few weeks and I would be very happy if anyone can help me 1) to get out this mist, 2) to understand the principles of treatement 12/16 bits, and 3) to give concrete recipes and formulas to work efficiently with the supplementary informations. Thanks in advance. Have a nice day !

[danbachmann]

I can understand your confusion, but it is easy to explain away.

The reason why only 12 bits of data are used in RAW output from cameras into a 16 bit image is that the sensors only report back 12 bits of data; however, standard image formats go from 8 bit to 16 bit because computers deal it bytes (8 bits at a time) more easily than bits, so you have 1 byte of color detail (8 bit) or 2 bytes (16 bit). Working with 16 bit takes the same amount of work, if not less, than working with 12 bit data.

Now the advantage of 16 bit... if you make adjustments such as curves or levels, you may start to see a fragmented histogram, where in 16 bit, it will remain smooth. I feel the largest advantage of RAW is in the initial conversion where you can set the white balance and a few other adjustments on your computer rather than having the camera do it.

[AdrianW]

The final piece of the puzzle is that we can only distinguish roughly 16.7 million or so unique colours, which equates nicely to a 24-bit computer display (8-bits per channel) - since 2^24 = 16777216. Since our perception is limited we design graphics cards/displays around those limits.

In other words we can only percieve a maximum of 8-bits per channel, and the computer display will only show 8-bits per channel anyway. As a result you'll only ever really notice it when you need to edit the images - as Dan has already said :)

[philip_coggan]

So you mean that I should convert to 16 bit before doing adjustments such as curves etc?

[Luko]

So you mean that I should convert to 16 bit before doing adjustments such as curves etc?

Definitely yes, Philip.

I don't remember who gave a link to a site showing the differences between two processes :

1- 16 bit file, convert to 8 bits, autolevels, save, close and open again

2- 16 bit file, autolevels, convert to 8 bits, save, close and open again

The difference was impressive, the last file to be converted kept that 3D feeling that 16 bits gives.

[AdrianW]

Convert from RAW to 16-bit, and work in 16-bit, yes - assuming your computer can take the stress ;)

Converting 8-bit to a 16-bit would be fairly pointless though, unless you were going to resize afterwards...

[philip_coggan]

Adrian - this the best reason I've ever seen for shooting RAW - possibly the only reason I've seen that's convincing.

Damn! - now I'll have to go buy some 2G cards to hold all those RAW images... :).

[AdrianW]

I've been shooting RAW for a couple of months, and I'm not planning to switch back to JPEG any time soon. The only reason I might change back to JPEG is if I run out of storage space whilst travelling...

I haven't managed to fill a 1Gb card in a days shooting yet though, even with RAW ;)

[philip_coggan]

This how many frames per day is an interesting question. Someone shooting landscapes, buildings, and other things that stand still, as you do, probably doesn't need all that many shots of a single subject. As I like to do people, and as people move, I tend to take lots of frames so that I can pick the one I like best - each moment of changing expression or movement will be different. This brings us back to a drawback of RAW - it takes longer to write to the card. Ah well, maybe I have to upgrade to the Canon Mark II... :). (Tho I've been half-considering going for a Mamiya 7II - on the basis that the manufacturers of digital can be gurenteed to supercede your latest top of the range state of the art model within 12 manonths, so why play their game?)

[greg]

I disagree, Adrian. It's worth converting 8-bit to 16-bit if you are going to do any processing that involves *more than one step*. It's a pity Luko didn't find the link he refers to, but what he says surprises me. If the software is written well (I suppose we're talking anout PS), there should be no difference between those two procedures (in proc. 1 the degradation occurs when you do autolevels; in proc. 2 when you make the 8-bit conversion - but the degration should be identical). If you do anything more, though, e.g. any dodging and burning, then it's much better to be in 16-bit, because you don't suffer from the degradation at every step, but only at the final conversion. Once two close shades have been mapped onto the same shade in 8-bit, there's no way to separate them again - this doesn't happen in 16-bit.

By the way, for Philip's comment - how about one of these in your camera?