What pixel depth to use?

[nickdu]

I've got a canon 30D. From the specs it appears it captures 12 bits per pixel. If I convert to 8 bits per pixel, say TIFF 8, am I losing much information?

Based on how I stated the above it appears a likely answer would be that I'm losing 4 bits per pixel or 33% of the information which seems like a lot. However, I don't know what those 12 bits comprise. Maybe it's 8 bits for the color and 4 bits for something else.

Thanks,
Nick

[DavidB]

I don't know what those 12 bits comprise. Maybe it's 8 bits for the color and 4 bits for something else
In the image produced by the camera, which has already been processed to get to this point, each pixel has a given level or intensity of each of the component colours, red, green and blue. If this intensity is represented by an 8-bit number, there are 256 possible intensities for each colour, and each pixel is represented by 24 (3x8) data bits. Each time you add a bit, you double the number of available steps. 12-bit color has 16 times as many steps as 8 bit (4096).

Whether this additional data actually represents additional information is a complex issue; there are a range of factors to consider which are way beyond the scope of a post such as this. It is worth getting hold of a good book such as Martin Evening's 'Adobe Photoshop for photographers' (I use the CS2 edition, and don't know whether there is an update for CS3 yet). On pages 164 and 165, there is an explanation of the advantages of 16 bit editing (Photoshop gives you the choice between 8- and 16-bit, so you have to use 16-bit to make use of all the data captured by your 30D).

Those of us who cannot aspire to Evening's almost fanatical attention to detail in quality control and workflow probably have a way to go before the difference between 8- and 16-bit becomes a major issue. LCD screens, for instance, are still 8-bit devices. It's very much a matter of personal judgement as to whether storing originals in 16-bit format is a sensible piece of future-proofing, or not. For what it's worth, I'm happy to stick with 8-bit at present.

[Chris Breeze]

The 12-bit per channel raw data does contain more data than an 8-bit per channel JPEG image. This allows you to adjust white balance and other settings post exposure without loss of quality but also gives slightly more dynamic range. In practice this means you can probably retain between 1/2 and 1 stop extra highlight detail in a raw file vs a JPEG. There is probably very little difference between saving the image as an 8-bit or a 16-bit TIFF (other than file size of course) if you use the exposure compensation or adjust the levels when converting the image to make the most of the image data that is available.
There is also some benefit in editing images in Photoshop in 16-bit mode as there is less chance that image data will be lost due to rounding/quantisation errors. These benefits are probably very small for most images where only a small amount of editing in Photoshop is required.