An Online Unsupervised Feature Selection and its Application for Background Suppression

Background suppression in video sequences has recently received great attention. While there exist many algorithms for background suppression, an important and challenging issue arising from these studies concerns that for which attributes of the data should be used for background modelling. It is interesting and difficult because there is no knowledge about the data to guide the search. Also, in real application for background suppression, the video length is unknown and the video frames are generated dynamically in a streaming fashion and arrive one at a time. Thus, it is impractical to wait until all data have been generated before feature learning begins. In this paper, we present an online unsupervised feature selection for background suppression. The advantage of our method is that it avoids any combinatorial search, is intuitively appealing, and allows us to prune the feature set. Moreover, our method, based on the self-adaptive model, has an ability to adapt and change through complex scenes. Experiments on real-world datasets are conducted. The performance of the proposed model is compared to that of other background modelling techniques, demonstrating the robustness and accuracy of our method.