Evaluation of background subtraction techniques for video surveillance

Abstract

Background subtraction is one of the key techniques for automatic video analysis, especially in the domain of video surveillance. Although its importance, evaluations of recent background subtraction methods with respect to the challenges of video surveillance suffer from various shortcomings. To address this issue, we first identify the main challenges of background subtraction in the field of video surveillance. We then compare the performance of nine background subtraction methods with post-processing according to their ability to meet those challenges. Therefore, we introduce a new evaluation data set with accurate ground truth annotations and shadow masks. This enables us to provide precise in-depth evaluation of the strengths and drawbacks of background subtraction methods.

Background Subtraction on Embedded Hardware

“…It is thus recommended to include postprocessing operations, even when comparing techniques. This was also the conclusion of Brutzer et al [24] and Benezeth et al [14]. Note that other filters can be used such as temporal filters, shadow filters [24], and complex spatio-temporal filtering techniques to relabel the classification results.…”

Dynamic tree-structured sparse RPCA via column subset selection for background modeling and foreground detection

“…The average processing time on a sequence of 100 RGB frames with resolution 600 × 800 with image alignment and background motion estimation is about 665 seconds, which is decreased to 195 seconds with CSSP, meaning a time-saving of more than 3.4 times. We perform extensive tests using four datasets [20], [21], [22], [23], comprised of 49 videos with various challenges. This allows us to compare our method to a large number of alternative methods.…”

A Perfect Estimation of a Background Image Does Not Lead to a Perfect Background Subtraction: Analysis of the Upper Bound on the Performance

“…Both techniques can also be combined. Note that post-processing techniques are known to always increase the performance [3,12]. In this paper, we study the performance of the BGS without any post-processing.…”