Text this: Model-based encoding of video images /