STEPHEN M. BENOIT, PhD.

Computer Vision - Artificial Perception Lab


Where to Reach Me:

Centre for Intelligent Machines
3480 University Street
Montréal, Québec
H3A 2A7
Canada

Telephone: +1 (514) 398-2185
Fax      : +1 (514) 398-7348
Email    : benoits@cim.mcgill.edu

What I have worked on


Towards Direct Recovery of Shape and Motion Parameters from Image Sequences

My more recent research has dealt with extracting structure from motion without optical flow. This process required the introduction of the Correspondence Matrix representation of image deformation, and a method to compute correspondenceless deformation detectors.
The different presentations and publications on this subject that I made in 2003 covered different aspects of the theory, case studies and results, but were given essentially the same title. The Correspondence Matrix representation has been applied to other problems. Steerable kernels are now fundamental tools in image processing and early vision. But can they be constructed for hexagonal pixels (common in biological vision systems) or even disorganized misshapen pixels (not necessarily in a grid)? The answer is yes. The specific problem of recovering Time To Collision (TTC) to various surfaces for a robot moving through a scene with only its camera to guide it has been studied as well.

Online Commercial Ventures

You can visit my favorite sites on the Internet that I have financed or built.

Special Events


Hobbies and Information


Usenet News Groups


Internet Search Engines