An Improved Background Reconstruction Algorithm Based on Online Clustering
Xiao Mei, Long Liu · 2008
Based on the assumption that background appears with large appearance frequency, a new background reconstruction algorithm based on mend basic sequential clustering is proposed in this paper. First, pixel intensity in period of time are classified based on mend basic sequential clustering. Second, merging procedure and reassignment procedure are run to classified classes. Finally, pixel intensity classes, whose appearance frequency are higher than a threshold, are selected as the background pixel intensity value. So the improved algorithm can rebuilt the background images of various scenes. Compared with the background reconstruction method based on online clustering, the simulation results show that the formed clusters, which are very closely located, are merged into a single one. And our method overcomes the effect of input order of data.