Properties of a variational model for video inpainting