Re: 10 bit, higher bitrate, How much better? by Tim Kolb on Sep 25, 2008 at 5:36:58 pm
The dynamic range of the camera is a separate issue from the codec or bitrate... detail in the darker areas can be affected by compression, but digital clip is digital clip, and the brightest areas of a conventional video image simply don't carry as much saturation as midtones and shadows.
As far as quality...higher bitrates always make a difference, but when you add bits to a really low bitrate, it makes quite a difference. Lighting a match doesn't throw a ton of light...unless you're standing in the dark.
10 bit vs 8 bit. Many people confuse this with having a wider dynamic range (brighter whites and more detailed blacks), but this isn't the case. In the codec world (which I dabble in) we would use the word "precision" after these figures. Every 'bit' has an 'on' and 'off' state if you will...the zeros and ones that make up data. One bit would be the equivalent of ancient terminal CRTs...each pixel is lit...or not. Adding a bit gives us another on and off switch, which when you look at the 2 states of the first bit and multiple it by the two states of the second bit...you have four possible states...add another bit for a total of 3 bits and double that giving you eight states, or 'shades' if you're referring to imaging...
4 bit=16 shades or states
10 bit =1024
So, an RGB '8 bit file' is usually being referred to as shorthand for '8 bits per channel'...some would refer to it as a 24 bit file with the combined channels, and adding an 8 bit alpha channel gives you a 32 bit file in the case of a Targa or similar file format.
Now...notice the difference between 8 bits per channel and 10 bits per channel. The possible values inside each channel quadruple. You have the same color range subdivided into four times smaller increments. The subtlety of shades within the palette is increased significantly. How significantly? Let's do the math:
256 x 256 x 256 = 16,777,216 total colors, 8 bit palette
1024 x 1024 x 1024 = 1,073,741,824 total colors, 10 bit palette
Pretty good jump.
This would make a difference in a situation where you have subtle gradient and you need to color correct. Let's say you have a circle of light under streetlight that falls off as you get more distant from the source...that gradient will 'posterize' or 'band' much quicker if you adjust the picture in an 8 bit environment vs a 10 bit environment.
Does everybody need 10 bit? I suppose not, but if you have FX work or color correction to do (and color correction is becoming more and more popular at lower and lower budgets) it does hold an obvious advantage.