Motion segmentation methods are effective for tracking video objects. However, objects segmentation methods based on motion need to know the global motion of the video in order to back-compensate it before computing the segmentation. In this paper, we propose a method which estimates the global motion of a High Definition (HD) video shot and then segments it using the remaining motion information. First, we develop a fast method for multi-resolution motion estimation based on spatio-temporal tubes. So we get a homogeneous motion vectors field (one vector per tube). From this motion field, we use a robust approach to estimate the parameters of the affine model that characterizes the global motion of the shot. After back-compensation of the video shot global motion, the remaining motion vectors are used to achieve the motion segmentation and extract the video objects.