Computational Models for Video Understanding
Abstract
Tremendous progress has been made within the past decade in the
collection, storage and transmission of video. However, the tools for
analyzing the video and obtaining mathematical descriptions of the
underlying content
lag behind. In this presentation, I will focus on computational dynamical
models that we have developed in order to represent the temporal evolution
of a video sequence. Together with static image models that describe the
object appearance, this provides a mathematical description for
representing the video sequence. We will consider particular examples in
human activity inference and human motion modeling to demonstrate the
applicability of dynamical models in shape space using Kendall's
statistical shape theory.