that means the user marks the boundary of the object to be manipulated, the rest of the image (including the object's shadow) will be treated as the background. This separation is called a mask in image processing lingo.