Spatial mixture of Gaussians for dynamic background modelling

Abstract

Modelling pixels using mixture of Gaussian distributions is a popular approach for removing background in video sequences. This approach works well for static backgrounds because the pixels are assumed to be independent of each other. However, when the background is dynamic, this is not very effective. In this paper, we propose a generalisation of the algorithm where the spatial relationship between pixels is taken into account. In essence, we model regions as mixture distributions rather than individual pixels. Using experimental verification on various video sequences, we show that our method is able to model and subtract backgrounds effectively in scenes with complex dynamic textures.

Regularised region-based Mixture of Gaussians for dynamic background modelling

“…In addition to the simulated data, we compared the performance of the regularised RMoG with the baseline RMoG algorithm from [14] on four well known video sequences. These datasets are the bottle sequence [17], the beach sequence [7], the waving trees sequence [12] and a CCTV sequence that was captured by us on board a moving bus.The ROC curves for each of the sequences are shown in Figure 2.…”

“…The RMoG algorithm takes into consideration this spatial relationship between pixels in a region and builds the model accordingly. It was shown in [14] that this framework models dynamic motion in the scene quite effectively.…”

“…Here, we start by outlining the online RMoG algorithm for dynamic background that was introduced in [14]. This algorithm is a complete framework for MoG modelling extending the standard per-pixel approach to larger regions thus enabling it to model highly dynamic regions in the scene.…”

“…In this paper, we apply momentum as a regularisation term to the region-based Mixture of Gaussians approach proposed by Varadarajan et al [14] to improve the performance of the algorithm.…”

Region-based Mixture of Gaussians modelling for foreground detection in dynamic scenes

“…In [20], we gave a brief overview of a region-based Mixture of Gaussians (RMoG) algorithm without any rigorous theoretical treatment and presented some preliminary results. In the current work, we show the detailed derivation of both batch and online mode update equations for the region based mixture modelling approach by using EM theory in section 4.…”