Beware of Aliases -- Signal Preservation is Crucial for Robust Image Restoration
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914831794176000 |
|---|---|
| author | Agnihotri, Shashank Grabinski, Julia Keuper, Janis Keuper, Margret |
| author_facet | Agnihotri, Shashank Grabinski, Julia Keuper, Janis Keuper, Margret |
| contents | Image restoration networks are usually comprised of an encoder and a decoder, responsible for aggregating image content from noisy, distorted data and to restore clean, undistorted images, respectively. Data aggregation as well as high-resolution image generation both usually come at the risk of involving aliases, i.e.~standard architectures put their ability to reconstruct the model input in jeopardy to reach high PSNR values on validation data. The price to be paid is low model robustness. In this work, we show that simply providing alias-free paths in state-of-the-art reconstruction transformers supports improved model robustness at low costs on the restoration performance. We do so by proposing BOA-Restormer, a transformer-based image restoration model that executes downsampling and upsampling operations partly in the frequency domain to ensure alias-free paths along the entire model while potentially preserving all relevant high-frequency information. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_07435 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Beware of Aliases -- Signal Preservation is Crucial for Robust Image Restoration Agnihotri, Shashank Grabinski, Julia Keuper, Janis Keuper, Margret Computer Vision and Pattern Recognition Machine Learning Image and Video Processing Image restoration networks are usually comprised of an encoder and a decoder, responsible for aggregating image content from noisy, distorted data and to restore clean, undistorted images, respectively. Data aggregation as well as high-resolution image generation both usually come at the risk of involving aliases, i.e.~standard architectures put their ability to reconstruct the model input in jeopardy to reach high PSNR values on validation data. The price to be paid is low model robustness. In this work, we show that simply providing alias-free paths in state-of-the-art reconstruction transformers supports improved model robustness at low costs on the restoration performance. We do so by proposing BOA-Restormer, a transformer-based image restoration model that executes downsampling and upsampling operations partly in the frequency domain to ensure alias-free paths along the entire model while potentially preserving all relevant high-frequency information. |
| title | Beware of Aliases -- Signal Preservation is Crucial for Robust Image Restoration |
| topic | Computer Vision and Pattern Recognition Machine Learning Image and Video Processing |
| url | https://arxiv.org/abs/2406.07435 |