Floating point?

[Joshua Jackson]

I'm attempting to learn more about digital color correction and I'm currently mastering my images in linear and vid spaces. I've had success converting vid spaces to linear and have had good results in color mastering, but I'm having difficulty finding out what "floating point" means exactly. Any clarity on this would be appreciated.

[Paul Bruening]

Instead of being useful, I'm dumping a Wikipedia link on ya'.

 

http://en.wikipedia.org/wiki/Floating_point

[wolfgang haak]

Hi Joshua,

 

You will have heard the term "bit depth". This refers to the value range available to encode each colour channel.

A 8 bit image contains 8 bits per channel, so 2⁸ = 256 values. This allows for 256³ = 16m colours.

 

To store and organize this data on a computer, you will choose an unisgned integer format, nice and simple.

For Bit depth, which are larger, i.e 32 bit = 2³² values per channel. To store that amout of data, you need a better container, as an integer runs out pretty quick (in programming variables can not take infinetly large values) so you need something a bit better suited for the job.

The variable type used here is a Float (Float32). or half float (Float16) with 32 bits and 16 bits respectively.

 

The terms used in the post packages here refer to the data types used to store the information and in a float you store greater numbers.

Unless a specific package makes ambigous use of terms, the word float has no bearing on colourspace, linear/encoded gamma etc. Use it like "massive" vs half float that means "big" and integer that means "standard"

 

Hope it helps,

regards,

Wolfgang

[David Mullen ASC]

This may also shed a little light on why floating-point is used:

http://www.openexr.com/about.html

[Joshua Jackson]

Excellent. Thank you both. That answers it perfectly.