On the use of trajectory information to assist stereopsis in a dynamic environment

If stereopsis is to be used in a dynamic environment, it makes little sense to re-compute the entire representation of disparity space from scratch at each time step. One simple approach would be to use the results from the current solution to “prime” the algorithm for the next solution. If three dimensional trajectory information was available, this information could be used to first update the previous solution, and then this updated solution could be used to “prime” the algorithm for the following stereo pair. Recent work[4, 5] has demonstrated that it is possible to measure such trajectory information very quickly without complex token or feature extraction. This paper demonstrates how raw disparity measurement made by this earlier technique can be integrated into a single trajectory measurement at each image point. A mechanism is then proposed that updates a stereopsis algorithm operating in a dynamic environment using this trajectory information.