Video Segmentation Algorithm Design Based On The Background Reconstruction

Video also called dynamic images, the activity image or image motion,is a group of images in time the shaft. Not only it contains a static image contains contents in the scene, but also contains target motion information and the object Video objects segmentation is always one of the difficult and hot research at home and abroad in video monitoring field, the purpose of which is to extract video objects from video scene which people are interested with or something have important properties of one or more. And the region which people are interested in usually has movement characteristic, so video division normally refers to the moving object detection and tracking. The effective segmentation of moving targets is vital to target identification, classification, trace etc. late operation.Background Difference value method is a fast and accurately method getting moving targets,but it's tend to be impacted by the environment like sunshine,which results in a not very good measuring result. So how to get the real-time background image will become the difficulty and the focus. This paper gives a new method of background reconstruction algorithm based on the shortage of various reconstruction algorithm, taking a good effect.This paper is to use background difference method along with the method of analytic frame differential, difference lies in adopting three frames difference-method to detect the change region, after getting the moving targets, completing background image by reconstructing the background. Background-reconstruction is by way of accumulative atlas mapping through change detection results,constantly replacing initial pixels with a new background pixels, finally getting relatively complete update background image.At last through background difference method to get more ideal target motion. Experiment shows that the algorithm can better remove background occluded area, effectively realize video object division, get a accurate and complete moving targets, and the method is very easy to realize, have good segmentation stability.