Bit depth query

PhotoGramly · May 31, 2017

Hey people,

This is more of a question for the sake of my own understanding rather than because I'm trying to do anything in particular.

If I'm understanding what I've read in various articles online, from a RAW source file you can create either an 8-bit or 16-bit image to edit as you see fit. According to an interview with a Canon representative (I shoot with Canon), Canon RAW files are recorded in 14 bits per channel. So my question is how can a 14-bit file produce a 16-bit image?

snowbear · May 31, 2017

I think it just leaves high-bit zeros - like going from 100 to 1,000.

Ysarex · Jun 1, 2017

PhotoGramly said:
Hey people,

This is more of a question for the sake of my own understanding rather than because I'm trying to do anything in particular.

If I'm understanding what I've read in various articles online, from a RAW source file you can create either an 8-bit or 16-bit image to edit as you see fit. According to an interview with a Canon representative (I shoot with Canon), Canon RAW files are recorded in 14 bits per channel. So my question is how can a 14-bit file produce a 16-bit image?

First off, some Canon cameras save 14 bit raw files, but not all. Depending on model and age the CR2 file you get from a Canon camera can be 10, 12 or 14 bit. This applies to other brands as well. What we've seen over the past decades is a steady increase in raw file bit depth so that today most pro-level cameras save 14 bit raw files while most consumer grade cameras are saving 12 bit raw files with an upward trend.

The formats you're dealing with here are not analogous such that the structure of a X bit raw file is similar to an X bit RGB photo -- they're not. So to answer your question in simple terms, you're trying to compare apples and oranges because they're both called fruit but the structures are really very different.

A 16 bit RGB photo has three channels, Red, Green, and Blue with 16 bits per channel mapped to a specific color space. A raw file is not mapped to any color space. So when we look at a numeric value in a 16 bit RGB file in say the red channel we're specifying a color -- a specific color. When we look at a numeric value in a 14 bit raw file in say the red channel we are not specifying a color -- apples and oranges. Think instead in a raw file that a numeric value in any of the channels is specifying a tone value -- light to dark. During the conversion (demosaicing) process the tone values in the raw file are interpolated and assigned color values. As part of this process the application of a tone curve is included that re-maps the raw file tone values to new tone/color values in the RGB file. In this process there's no conflict between a 10, 12, 14 bit raw file and 16 bit RGB file.

Joe