A New Optical Flow Estimation Based on Dynamical System Approach with Intensity Variation Violating the Conservation Law

摘要

Optical flow is a vector that represents the motion of objects in images. The dynamical system approach to optical flow estimation [3] was proposed as one of the optical flow estimation methods. Most of the optical flow estimation methods are based on the intensity conservation law. Therefore, these estimation methods might not be able to estimate correctly if the intensity conservation law is violated by various factors. This paper proposes a new estimation model for images in which the image intensity is not maintained after moving objects in images. The proposed method can estimate not only the optical flow and its spatial gradients but also the intensity variation and its spatial gradients that do not follow the intensity conservation law. In addition, this paper focuses on the possibility of the optical flow estimation from the perspective of the observability. The effectiveness of the proposed model is demonstrated through numerical examples.