Spatiotemporal Removal of Text in Image Sequences