Deep Learning for Face Deblurring: A Survey

Face image deblurring has been widely used in various fields. However, existing deblurring methods still face challenges such as low robustness and low precision. This work summarizes these challenges by investigating and analyzing the current state of face image deblurring methods. This survey explores three key aspects: adversarial-network-based generative approaches, auto-encoder-based variational approaches, and diffusion-model-based approaches, with an emphasis on the diffusion-model-based ones. Moreover, this work evaluates experimental results over widely-compared metrics to analyze the performance of the current diffusion models systematically. The findings of this survey provide reference and future research directions for face image deblurring.