Robust foreground segmentation using improved Gaussian Mixture Model and optical flow

Abstract

In automatic video surveillance applications, one of the most popular topics consists of separating the moving objects from the static part of the scene. In this context, Gaussian Mixture Model (GMM) background subtraction has been widely employed. It is based on a probabilistic approach that achieves satisfactory performance thanks to its ability to handle complex background scenes. However, the background model estimation step is still problematic; the main difficulty is to decide which distributions of the mixture belong to the background. To achieve an improved overall performance, motion cue could provide a rich source of information about the scene. Therefore, in this paper, we propose a new approach based on incorporating an uniform motion model into GMM background subtraction. By considering these both cues, high accuracy of foreground segmentation is obtained. Our approach has been experimentally validated showing better segmentation performance by comparisons with other approaches published in the literature.