Abstract:Low-rank and sparse decomposition is a powerful tool in background extraction and moving object detection problems since the static background usually lies in a low dimensional subspace while the rapidly changing object maintains a sparse property. Conventional techniques for low-rank and sparse decomposition can handle noise-free or Gaussian noise scenarios. However, their performance significantly degrades in challenging scenarios contaminated by impulsive noise or outliers. In this paper, a new model with a separable noise component, which we call unstructured l(p)-regularized low-rank representation (ULLR), is proposed. Instead of incorporating the Frobenius norm constraint on the noise component, the proposed ULLR incorporates flexible l(p)-regularized schemes with 0 < p <= 2 to handle varied scenarios. For the moving object detection problem, we further incorporate the local group structure and homogeneous group structure to encode the underlying object. Then, a new model called the structured l(p)-regularized low-rank representation (SLLR) is proposed to enhance the detected foreground while suppressing the noise and dynamic background. Especially, when 0 < p <= 1, the SLLR achieves robustness against outliers and hence handles challenging scenarios well. The resultant optimization problem is solved using the framework of alternating direction method of multipliers (ADMM). Each ADMM subproblem can be efficiently solved. More importantly, the solutions for the noise-based subproblem are rigorously derived with respect to different l(p)-regularized schemes. Simulation results demonstrate the efficiency, accuracy, and robustness to outliers of the proposed algorithms.

Moving Object Detection by Patch Dividing Through Low-Rank Framework

Masked-RPCA: Sparse and Low-rank Decomposition Under Overlaying Model and Application to Moving Object Detection

Moving Object Detection Via Low-Rank Total Variation Regularization

Non-Local Image Inpainting Using Low-Rank Matrix Completion: Non-Local Image Inpainting Using Low-Rank Matrix Completion

Moving Object Detection in Dynamic Environment Via Weighted Low-Rank Structured Sparse RPCA and Kalman Filtering

An Improved Total Variation Regularized RPCA for Moving Object Detection with Dynamic Background

Moving Object Detection Using Tensor-Based Low-Rank and Saliently Fused-Sparse Decomposition.

Moving Object Detection Based on On-Line Block-Robust Principal Component Analysis Decomposition

Outlier-Robust Moving Object and Background Decomposition via Structured <i>l<sub>p</sub></i>-Regularized Low-Rank Representation

Salient object detection with low-rank approximation and ℓ2,1-norm minimization

Moving Object Detection Based on Incremental Learning Low Rank Representation and Spatial Constraint.

Moving Object Detection by Detecting Contiguous Outliers in the Low-Rank Representation

Moving Target Detection by Robust PCA in the Topological Space of Low-Rank Matrices

Joint Video Frame Set Division and Low-Rank Decomposition for Background Subtraction

Robust Tracking with Discriminative Ranking Middle-Level Patches

Infrared Small Target Detection with Patch Tensor Collaborative Sparse and Total Variation Constraint

Tracking in Multimedia Data Via Robust Reweighted Local Multi-Task Sparse Representation for Transportation Surveillance

Detecting Moving Objects from Dynamic Background Combining Subspace Learning with Mixed Norm Approach

Learning to Rank Patches for Unbiased Image Redundancy Reduction

Background subtraction based on low-rank and structured sparse decomposition.

Low-Rank Matrix Recovery Via Continuation-Based Approximate Low-Rank Minimization