In this paper, we propose a nonparametric background modeling for background subtraction using exponentially weighted histograms. Our background model is constructed by using exponentially increasing weights. We express our model by using recurrence formula. In our model, recently observed pixels have a bigger influence on the background model than older ones. The proposed model need not hold past pixel values in order to remove an old value from the model for updating. We confirmed that the proposed method is processed in real time experimentally and the accuracy of the background subtraction using our background model is comparable to that of conventional methods.
|Translated title of the contribution||Background Modeling using Exponentially Weighted Histogram|
|Number of pages||2|
|Journal||電子情報通信学会技術研究報告. PRMU, パターン認識・メディア理解|
|Publication status||Published - Feb 6 2014|