Learning Distortion Invariant Representation for Image Restoration from A Causality Perspective
This work addresses a critical limitation in image restoration for applications requiring robustness to diverse real-world distortions, though it appears incremental as it builds on existing causality and meta-learning concepts.
The paper tackles the problem of deep neural networks failing to generalize to real-world image degradations with varying types and degrees by proposing a novel training strategy from a causality perspective, resulting in improved generalization capability for unseen distortions as demonstrated in extensive experiments.
In recent years, we have witnessed the great advancement of Deep neural networks (DNNs) in image restoration. However, a critical limitation is that they cannot generalize well to real-world degradations with different degrees or types. In this paper, we are the first to propose a novel training strategy for image restoration from the causality perspective, to improve the generalization ability of DNNs for unknown degradations. Our method, termed Distortion Invariant representation Learning (DIL), treats each distortion type and degree as one specific confounder, and learns the distortion-invariant representation by eliminating the harmful confounding effect of each degradation. We derive our DIL with the back-door criterion in causality by modeling the interventions of different distortions from the optimization perspective. Particularly, we introduce counterfactual distortion augmentation to simulate the virtual distortion types and degrees as the confounders. Then, we instantiate the intervention of each distortion with a virtual model updating based on corresponding distorted images, and eliminate them from the meta-learning perspective. Extensive experiments demonstrate the effectiveness of our DIL on the generalization capability for unseen distortion types and degrees. Our code will be available at https://github.com/lixinustc/Causal-IR-DIL.