Text this: Leveraging text semantics for enhanced scene text image super-resolution