Running average is a simple and effective background modeling method that generates adaptive background image for moving object detection. Fuzzy Running Average (FRA) improves the selectivity of Standard Running Average (SRA). However, its background restoration rate is slow. This leads to false object detection when a static object becomes dynamic. To overcome this problem, an Extended Fuzzy Running Average (EFRA) is proposed. The results show that the EFRA not only retains the selectivity benefit of FRA, but also improves the restoration rate significantly.