Compositing digital images, revisited
Digital image compositing is a core technique of computer graphics, used for combining separately-produced elements into a composite image.
The technique, which underlies Photoshop's layers and just about every other image manipulation system, augments the usual colour components of image pixels with an extra value, invariably called alpha, that specifies how much of the background image should shine through the foreground when you layer two images.
This talk will go over a few theoretical things we have learned about compositing since 1984 when the first public description of digital compositing appeared. Alpha-augmented colours turn out to be mathematically interesting, and the mathematics has interesting applications.
In particular:
- Alpha-augmented colours, at least the non-opaque, have inverses. This has surprising applications, for example to rendering clouds in a REYES-style shading–before-hiding renderer.
- We can define a power function on alpha-augmented colours that gives us the mathematical leverage we need to properly do volume combining and deep compositing.
- Alpha-augmented colours form a Lie Group, which has a corresponding Lie Algebra with log and exponential maps joining the two. By mapping to the Lie Algebra we can properly define natural interpolants and splines on alpha-augmented colours.
Reception to follow in William G. Davis Computer Research Centre (DC) room 1301 (Fishbowl).
Space is limited. Register today!