The author proposes a diffusion-based approach for blind text image super-resolution, utilizing Image Diffusion Models (IDM) and Text Diffusion Models (TDM) to restore text images with high fidelity and style realness.
Diffusion models IDM, TDM, and MoM cooperate to restore text images with high fidelity and style realness.