Robust moving object region extraction using a column and row projection algorithm

In intelligent surveillance field, the numerous methods have been proposed for foreground extraction from a stationary or dynamic background from a general video sequence. It is very difficult for the proposed methods to have both high accuracy, low computational complexity and less memory requirements. Unlike previous approaches to object detection which detect objects by building adaptive models of the background, this paper discusses projecting the pixels in the observed image to the bottom row or the last column and using the change of projection to find the moving object region. The algorithm is easy to be realized and the computational complexity is very low. As one of methods for moving object detection, it has high efficiency firstly. Additionally, because the special characteristics of pixels is considered, it has a stronger ability to enhance the moving object region information in comparison with the congener methods. It is tested for various indoor and outdoor video sequences which reliably deals with illumination changes and other small motions.