Real-time stereo- and motion-based figure ground discrimination and tracking using LOG sign correlation

We can compute large kernel Laplacian-of-Gaussian convolutions and make robust stereo and motion disparity measurements at video rates. This allows to see much more vividly how algorithms behave on dynamic scenes. This in turn enables to better think about the kind of measurements which ought to be made and methods far applying them to accomplish perceptual tasks. The paper describes the PRISM-3 real-time image matching system and progress using it in the areas of figure-ground discrimination and active object tracking using stereo and motion