The Coding Ecology:
Image Coding via Competition among Experts
William Butera and V. Michael Bove, Jr.
MIT Media Laboratory
to appear in IEEE Transactions on Circuits and Systems for Video
Technology
We consider the image coding problem as a competitive ecology of
specialists, each vying for the task of coding a portion of an
image. Each specialist, or expert, derives its competitive advantage
from its ability to concisely describe an underlying visual event
( e.g. shadows, motion, occluding objects). In this paper, the
metaphor of an auction informs the design of a predictive
coder. Experts locate candidate regions of support, characterize the
activity within and submit their proposals in the form of bids passed
along to an auctioneer. We describe the designs for six such experts, and
examine candidate strategies for the decision unit (the auctioneer).
We define protocols for comparing the bids, and compare overall coding
performance using several examples.