Streaming Video Completion using a Tensor-Networked Kalman Filter

Master thesis (2020)

Authors

S.J.S. de Rooij Mechanical Engineering

Contributors

K. Batselier Team Jan-Willem van Wingerden - Mechanical, Maritime and Materials Engineering (supervisor 1)
J.W. van Wingerden Team Jan-Willem van Wingerden - Mechanical, Maritime and Materials Engineering (supervisor 2)
J.F.P. Kooij Intelligent Vehicles - Mechanical, Maritime and Materials Engineering (supervisor 2)
Faculty
Mechanical Engineering
Copyright: © 2020 S.J.S. de Rooij
Kalman Filter Tensor decomposition Tensor-trains Streaming video Video completion
More Info
expand_more

Published Date

13-07-2020

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Mechanical Engineering

Abstract

In streaming video completion one aims to fill in missing pixels in streaming video data. This is a problem that naturally arises in the context of surveillance videos. Since these are streaming videos, they must be completed online and in real-time. This makes the streaming video completion problem significantly more difficult than the related video completion problem. State-of-the-art streaming video completion methods based on adaptive matrix completion, do not work well when the number of missing pixels is high (~95%). Therefore, in this report a new streaming video completion method will be introduced based on a tensor-networked Kalman filter. The results in this report will show that this Kalman filter method performs better than the state-of-the-art methods when the percentage of missing pixels is high (~95%).