Authors:
Annette Stahl
and
Ole Morten Aamo
Affiliation:
Norwegian University of Science and Technology (NTNU), Norway
Keyword(s):
Motion estimation, Optimal control, Physical prior, Optimisation.
Related
Ontology
Subjects/Areas/Topics:
Adaptive Signal Processing and Control
;
Artificial Intelligence and Decision Support Systems
;
Enterprise Information Systems
;
Image Processing
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Knowledge-Based Systems Applications
;
Machine Learning in Control Applications
;
Optimization Algorithms
;
Optimization Problems in Signal Processing
;
Robotics and Automation
;
Signal Processing, Sensors, Systems Modeling and Control
;
Vision, Recognition and Reconstruction
Abstract:
In this paper we present an optimal control approach for image motion estimation in an explorative and novel way. The variational formulation incorporates physical prior knowledge by giving preference to motion fields that satisfy appropriate equations of motion. Although the framework presented is flexible, we employ the Burgers equation from fluid mechanics as physical prior knowledge in this study. Our control based formulation evaluates entire spatio-temporal image sequences of moving objects. In order to explore the capability of the algorithm to obtain desired image motion estimations, we perform numerical experiments on synthetic and real image sequences. The comparison of our results with other well-known methods demonstrates the ability of the optical control formulation to determine image motion from video and image sequences, and indicates improved performance.