The computation of the optical flow field from an image sequences requires the definition of constraints on the temporal change of image features. In general, these constraints limit the motion of the body in space and/or of the features on the image plane. In this paper we consider the implications in using multiple constraints, both in terms of the increased complexity, but also for its advantages. Considering multiple image points not only allows to obtain a (locally) better estimate of the velocity field but also to characterize the temporal behaviour of the image features and to detect erroneous measurements due to occlusions, motion/depth discontinuities or shadows. Moreover, by hypothesizing a constant acceleration motion model, the derivatives of the optical flow are computed. Several experiments are presented demonstrating the feasibility of the algorithm in real applications.
Proceedings of International Symposium on Computer Vision - ISCV