Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted l1-l1 minimization; this unfolding leads to a different proximal operator (aka different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.
Van Luong, H, Joukovsky, B, Eldar, Y & Deligiannis, N 2020, A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation. in 28th European Signal Processing Conference. pp. 1432-1436, 28th European Signal Processing Conference, Amsterdam, Netherlands, 24/08/20.
Van Luong, H., Joukovsky, B., Eldar, Y., & Deligiannis, N. (2020). A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation. In 28th European Signal Processing Conference (pp. 1432-1436)
@inproceedings{18e9c090388346d4abfdf23ebec0ca36,
title = "A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation",
abstract = "Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted l1-l1 minimization; this unfolding leads to a different proximal operator (aka different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.",
author = "{Van Luong}, Huynh and Boris Joukovsky and Yonina Eldar and Nikolaos Deligiannis",
year = "2020",
month = oct,
day = "2",
language = "English",
pages = "1432--1436",
booktitle = "28th European Signal Processing Conference",
note = "28th European Signal Processing Conference, EUSIPCO2020 ; Conference date: 24-08-2020 Through 28-08-2020",
url = "https://eusipco2020.org/",
}