Research >> Branch and bound techniques
Branch and bound techniques
Contour tracking can be made
robust by posing it as a mixed
continuous-combinatorial optimization
problem. The combinatorial
part derives from the very large number of
curves that can be formed
from edge-points in the image. The continuous
piece enforces the
constraint that the curve must lie on a low-dimensional
manifold, which
describes the object of interest. We find efficient, tree-like
algorithms whose output are provably close to the global optimum.
Faculty: Daniel Freedman 
Publication:
D. Freedman. Effective tracking through tree-search. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 25(5):604-615, 2003. [ pdf ] [ bib ]
Home | People | Research | Publications | Contact | News/Events | Download | Links