Patch-based learning of adaptive Total Variation parameter maps for blind image denoising
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909653550497792 |
|---|---|
| author | Fantasia, Claudio Calatroni, Luca Descombes, Xavier Rekik, Rim |
| author_facet | Fantasia, Claudio Calatroni, Luca Descombes, Xavier Rekik, Rim |
| contents | We consider a patch-based learning approach defined in terms of neural networks to estimate spatially adaptive regularisation parameter maps for image denoising with weighted Total Variation (TV) and test it to situations when the noise distribution is unknown. As an example, we consider situations where noise could be either Gaussian or Poisson and perform preliminary model selection by a standard binary classification network. Then, we define a patch-based approach where at each image pixel an optimal weighting between TV regularisation and the corresponding data fidelity is learned in a supervised way using reference natural image patches upon optimisation of SSIM and in a sliding window fashion. Extensive numerical results are reported for both noise models, showing significant improvement w.r.t. results obtained by means of optimal scalar regularisation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_16010 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Patch-based learning of adaptive Total Variation parameter maps for blind image denoising Fantasia, Claudio Calatroni, Luca Descombes, Xavier Rekik, Rim Image and Video Processing Machine Learning Numerical Analysis We consider a patch-based learning approach defined in terms of neural networks to estimate spatially adaptive regularisation parameter maps for image denoising with weighted Total Variation (TV) and test it to situations when the noise distribution is unknown. As an example, we consider situations where noise could be either Gaussian or Poisson and perform preliminary model selection by a standard binary classification network. Then, we define a patch-based approach where at each image pixel an optimal weighting between TV regularisation and the corresponding data fidelity is learned in a supervised way using reference natural image patches upon optimisation of SSIM and in a sliding window fashion. Extensive numerical results are reported for both noise models, showing significant improvement w.r.t. results obtained by means of optimal scalar regularisation. |
| title | Patch-based learning of adaptive Total Variation parameter maps for blind image denoising |
| topic | Image and Video Processing Machine Learning Numerical Analysis |
| url | https://arxiv.org/abs/2503.16010 |