|
Unless otherwise noted all contents and images on this site are copyright © 2001-2008 Roger Cavanagh
and may not be used without permission.
Labelled with ICRA
|
Unless otherwise noted all contents and images on this site are copyright © 2001-2008 Roger Cavanagh
and may not be used without permission.
Labelled with ICRA
Cameras like the 10D, 1Ds, 300D and many of the P&S digicams capture 12-bits of data for each pixel. (Of course, it's only luminance information that is turned into RGB data by the interpolation of data from multiple pixels, but I'm going to ignore that complication.) All 12-bits are saved in the raw file and can be converted into 16-bit TIFF by your favourite conversion program. Most image-processing PC programs use either 8- or 16-bit, so the 12-bit raw data is "padded" to fill 16-bits. The JPG created by the camera saves only 8-bit RGB.
What this means is that, for each colour channel, the 12(16)-bit file can take 4096 discrete values, whereas the 8-bit files only gives 256. Of course, this still means: 256 x 256 x 256 = 16,777,216 separate colours. It's a lot smaller than 68,719,476,736 (4096 x 4096 x 4096), but still a pretty big number. It's also a lot bigger than the number of different colours the human eye can distinguish. There seems to be some argument about this, as estimates vary from 1 million to 7 million. Things are even more complicated because there are colours not in the 16.7 million that the human eye can distinguish.
One argument often advanced in favour of 16-bit vs. 8-bit is that post-processing adjustments in Photoshop (or whatever is your preferred picture editor) will cause less posterisation. You can see this is happening when the histogram starts to get spiky and has gaps. In this article, Bruce Fraser demonstrates that there is an advantage to editing 8-bit images in 16-bit space. His argument is compelling, I have tried the same thing and proved it for myself. If it is an advantage to edit in 16-bit with 8-bit data, I must conclude that it is more of an advantage to edit 16-bit data in 16-bit space.
Conversely, 16-bit editing is much slower in Photoshop and uses more memory. There are facilities, such as layers and some of the tools, that can only be used on 8-bit images. Bruce Fraser has some workarounds and Photoshop 8 is expected to expand 16-bit processing options.