International Conference on Information Complexity and Statistical Modeling in High Dimensions with Applications (IC-SMHD-2016), Nevşehir, Turkey, 18 - 21 May 2016, pp.1-15
This paper addresses Statistical Background Modeling focusing on Information Complexity guided Gaussian Mixture Models (GMMs). We successfully employ GMMs, which are optimal with respect to the information complexity criteria. Moving objects are segmented out through background subtraction which utilizes the computed background model. Here in this study, surveillance data content highlighting through background suppression is our main concern. Through utilizing an efficient background modeling the regions in motion are distinguished from background which has no significant importance. It is possible to evaluate the contribution weight of each pixel to form the background at any time in the image sequence. A GMM is accompanied to each pixel and an information complexity guided optimal GMM selection scheme is used. Regions highly occupied by moving objects are extracted optimally using parameter maps for component number and the shape of the components for each pixel. This new technique in background modeling field results in a stable moving target segmentation which reliably overcomes the demanding challenges of lighting changes, repetitive motions from clutter, and long-term scene changes.