A gaussian-median filter for moving objects segmentation applied for static scenarios

Abstract

Background subtraction or also called foreground detection is an approach normally used for moving object segmentation in video sequences captured from a fixed camera. Most of the methods under this approach are not able to segment or require the strict absence of objects during their training or learning period in the first frames. In this document, a method capable of segmenting moving objects from the beginning of a video sequence and at the same time constructing a reference background image is proposed. The segmentation results show that the foreground and background regions of the scene are not affected during this stage compared to other methods.