Bit-depth?

[InskiP]

Please keep in mind that I'm new to digital photography...
Can I select the bit-depth on my camera(20D) like I would on my scanner?
Does anyone know how much bit-depth the 20D captures?
Curious and confused about this. I couldn't find any info about this in the manual so maybe it doesn't even apply???

[nitsch]

Hi there, you can't select a different bit depth as such but JPEG's are 8-bit and RAW files are 12-bit. :)

[tim]

Like Nitsch says, you can't select the bit depth on the 20D, it's fixed at maximum quality. JPG are 8 bit but it's not a linear scale so it's a little misleading. RAW is 12 linear bits. If you get the shot perfect and save it as a JPG it will look as good or very close to the quality of a RAW image converted to JPG in photoshop. If you don't get the exposure correct RAW gives you more leeway to correct errors later.

[nitsch]

Well explained Tim! :D

[tim]

:)

[Cactus]

Well, to be exact 8bit per channel. Three channels (Red, Green and Blue) equals 3*8bits = 24bits. Raw files are 12bits per channel or 36bits. Regarding quality, raw files are better, not only because of the bit-depth but will be more flexable when "developing".

There is something one should think about regarding bit-depth. When working with the image in photoshop or whatever tool one uses, you're probably using 24bit color in windows/mac. That's 24 for the three channels, or 8bit per channel. However, if one's using 32bit it's still 24bits of colour and 8bits of padding. That's the reason why we can't see difference between 8bit and 16bit files in photoshop or what have you. Still working in 16bit is better to avoid quality loss when "developing".

My 2cents. I might be mistaking, if so, correct me. ^_^

[hmhm]

It's worth noting that you're not likely to be able to view an image in anything greater than 8 bits/channel precision, nor are your eyes likely to be capable of making that distinction any way. The image is highly likely to be converted to 8-bits per channel before display on a computer screen or being sent to a printer, or even being saved in a file format that doesn't allow higher than 8 bits/channel precision (e.g. jpg).

We capture and edit images in higher bit-depth because many editing operations necessarily "lose information", so preserving precision while editing helps to maximize the "diversity of values" possible in the final result. When editing in lower precision, it's possible for a pair of pixels that had different values in the capture to end up with the same values after some editing operation, whereas maintaining higher precision in the editing process would yield a final result where those pixels retained distinct values (even after the final result is converted to 8 bit/channel precision).

If we're not doing any significant editing, then the added precision doesn't buy us much in the final product, since our last step is to throw away that added precision any way.

Some versions of photoshop are capable of editing images in 16 bits per channel. In some cases, certain operations are restricted to 8 bit per channel operation. Depends on the function and the software version.
-harry

[robertwgross]

Raw files are 12bits per channel or 36bits. ...

My 2cents. I might be mistaking, if so, correct me. ^_^

RAW files are 12 bits total, so you are mistaken.

---Bob Gross---

[hmhm]

RAW files are 12 bits total, so you are mistaken.

Well, to be clear, the Canon sensors capture 12 bits per channel, but each pixel only captures one channel (red, green, or blue), so each pixel in a RAW file has only 12 bits, and represents only one color. When the RAW file is converted to a more conventional format, for each pixel the values for the 2 "other" colors are calculated from the values of neighboring pixels of that color, a process known as Bayer interpolation. This yields an image with 36 bits per pixel, i.e. 12 bits per channel and 3 channels per pixel.

So, yeah, a RAW file has only 12 bits per pixel, though that's still 12 bits per channel, it's just that there's only 1 channel per pixel until the Bayer interpolation is done.
-harry