Background division, a suitable technique for moving object detection

Abstract

Nowadays, background model does not have any robust solution and constitutes one of the main problems in surveillance systems. Researchers are working in several approaches in order to get better background pixel models. This is a previous step to apply the background subtraction technique and results are not as good as expected. We concentrate our efforts on the second step for segmentation of moving objects and we propose background division to substitute background subtraction technique.This approach allows us to obtain clusters with lower intraclass variability and higher inter-class variability, this diminishes confusion between background and foreground,pixels.We compared results using our background division approach versus wallflowers algorithm [1] as the baseline to compare. © 2010 Springer-Verlag.