On Adversarial Robustness of Deep Image Deblurring

Kanchana Vaishnavi Gandikota, Paramanand Chandramouli, Michael Moeller

Abstract

Recent approaches employ deep learning-based solutions for the recovery of a sharp image from its blurry observation. This paper introduces adversarial attacks against deep learning-based image deblurring methods and evaluates the robustness of these neural networks to untargeted and targeted attacks. We demonstrate that imperceptible distortion can significantly degrade the performance of state-of-the-art deblurring networks, even producing drastically different content in the output, indicating the strong need to include adversarially robust training not only in classification but also for image recovery.

Resources

[arxiv]

Bibtex

@inproceedings{gandikota2022adversarial, title={On Adversarial Robustness of Deep Image Deblurring}, author={Gandikota, Kanchana Vaishnavi and Chandramouli, Paramanand and Moeller, Michael}, booktitle={IEEE International Conference on Image Processing (ICIP)}, pages={3161–3165}, year={2022}, organization={IEEE} }