Learned Temporal Models of Image Motion

An approach for learning and estimating temporalflow models from image sequences is proposed. The temporal-flow models are represented as a set of orthogonal temporal-flow bases that are learned using principal component analysis of instantaneous flow measurements. Spatial constraints on the temporal-flow are also developed for modeling the motion of regions in rigid and coordinated motion. The performance of these models is demonstrated on several long image sequences of rigid and articulated bodies in motion. 1 Introduction Tracking the image motion of a human body in action is an exceptionally challenging computer vision problem. Even ignoring the fine structure of the hands, and assuming that the feet are rigidly connected to the calves and the hands to the forearms, a human body is composed of ten basic parts, many of which can move in quite independent ways. Natural human motions, such as walking, kicking, etc., are, of course, very constrained by factors including motion symmet..