AutoDIR: Automatic All-in-One Image Restoration with Latent Diffusion
This provides an all-in-one solution for users needing to restore images with various unknown issues, though it appears incremental as it builds on latent diffusion methods.
AutoDIR tackles the problem of automatically restoring images with unknown degradations by using a two-stage system that identifies issues and applies restoration, achieving state-of-the-art performance across a wider range of tasks.
We present AutoDIR, an innovative all-in-one image restoration system incorporating latent diffusion. AutoDIR excels in its ability to automatically identify and restore images suffering from a range of unknown degradations. AutoDIR offers intuitive open-vocabulary image editing, empowering users to customize and enhance images according to their preferences. Specifically, AutoDIR consists of two key stages: a Blind Image Quality Assessment (BIQA) stage based on a semantic-agnostic vision-language model which automatically detects unknown image degradations for input images, an All-in-One Image Restoration (AIR) stage utilizes structural-corrected latent diffusion which handles multiple types of image degradations. Extensive experimental evaluation demonstrates that AutoDIR outperforms state-of-the-art approaches for a wider range of image restoration tasks. The design of AutoDIR also enables flexible user control (via text prompt) and generalization to new tasks as a foundation model of image restoration. Project is available at: \url{https://jiangyitong.github.io/AutoDIR_webpage/}.