Foreground Segmentation via Dynamic Tree-Structured Sparse RPCA
Video analysis often begins with background subtraction which consists of creation of a background model, followed by a regularization scheme. Recent evaluation of representative background subtraction techniques demonstrated that there are still consider
- PDF / 1,554,818 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 9 Downloads / 211 Views
Abstract. Video analysis often begins with background subtraction which consists of creation of a background model, followed by a regularization scheme. Recent evaluation of representative background subtraction techniques demonstrated that there are still considerable challenges facing these methods. We present a new method in which we regard the image sequence as being made up of the sum of a low-rank background matrix and a dynamic tree-structured sparse outlier matrix and solve the decomposition using our approximated Robust Principal Component Analysis method extended to handle camera motion. Our contribution lies in dynamically estimating the support of the foreground regions via a superpixel generation step, so as to impose spatial coherence on these regions, and to obtain crisp and meaningful foreground regions. These advantages enable our method to outperform state-of-the-art alternatives in three benchmark datasets.
1
Introduction
Foreground segmentation plays a critical role in applications such as automated surveillance, action recognition, and motion analysis. Despite the efforts in this field, recent evaluation of state-of-the-art techniques [1,2] showed that there are still shortcomings in addressing all challenges in foreground segmentation. Addressing these challenges, leads to a number of considerations in designing a background model, as well as expected behavior from foreground objects, which in complex real-life applications remains an open problem. The background model can undergo sudden or gradual illumination changes, as well as background motions such as trees swaying or water rippling in a lake. In addition, global motion caused by camera movement or jitter can affect detection of genuine foreground objects. Noise is another problematic factor which is interleaved with challenges of camouflage. In most cases noise can increase the range of values considered to belong to the background, allowing camouflaged objects to remain undetected. A desirable background model must be able to learn a variety of modes from the video feed, such that it handles variations in the background, moved objects, and noise without compromising its ability to detect camouflaged regions. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part I, LNCS 9905, pp. 314–329, 2016. DOI: 10.1007/978-3-319-46448-0 19
Foreground Segmentation via Dynamic Tree-Structured Sparse RPCA
315
In this paper, we handle all these challenges using an approximated Robust Principal Component Analysis (RPCA) based method for background modeling. Given a data matrix containing the frames of a video sequence stacked as its columns, A ∈ Rm×n , RPCA [3] solves the matrix decomposition problem min L∗ + λS1 L,S
s.t.
A = L + S,
(1)
as a surrogate for the actual problem min rank(L) + λS0 L,S
s.t.
A = L + S,
(2)
where L is the low-rank component corresponding to the background and S is the sparse component containing the foreground outliers. We are interested in a case where we can decompose the matrix
Data Loading...
