This is used to determine the region of the image that will be loaded in each time-point. The size of the region will be: Object size * Window size (see also the help for the next text field). The reason you can specify the object size separately is due to the "center of mass" tracking algorithm, which for each time-point iteratively computes the center of mass, starting in a region of size "Object size * Window size" and then 'zooming in' until the region is of Object size.