In the future, I plan to extend the probabilistic recognition paradigm to use
contextual indicators (e.g., the environmental conditions)
and goals to drive the prior probabilities (i.e., using expectation to improve efficiency).
I believe this direction of research is a key to achieving efficient
recognition over a wide range of scenarios with large model bases.
My thesis research only considers rigid objects (i.e., 6 degrees of freedom).
Extending the techniques to apply to articulated objects such as human forms is
another line of research I plan to pursue.
I am familiar with machine learning and neural networks and
am interested in exploring the use of techniques from these fields
to aid or enhance solutions to computer vision problems.
Recently, the graphics and vision communities have demonstrated simple techniques
that process visual data to produce astonishing graphical effects (e.g., image
mosaicing and morphing).
The door is open for applications of computer vision techniques to
automate these effects, and I am very anxious to collaborate with graphics
researchers on these problems.