CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing
Deep learning based methods hold state-of-the-art results in low-level image processing tasks, but remain difficult to interpret due to their black-box construction. Unrolled optimization networks present an interpretable alternative to constructing deep neural networks by deriving their architectur...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Open Journal of Signal Processing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9769957/ |
| _version_ | 1818231448696520704 |
|---|---|
| author | Nikola Janjusevic Amirhossein Khalilian-Gourtani Yao Wang |
| author_facet | Nikola Janjusevic Amirhossein Khalilian-Gourtani Yao Wang |
| author_sort | Nikola Janjusevic |
| collection | DOAJ |
| description | Deep learning based methods hold state-of-the-art results in low-level image processing tasks, but remain difficult to interpret due to their black-box construction. Unrolled optimization networks present an interpretable alternative to constructing deep neural networks by deriving their architecture from classical iterative optimization methods without use of tricks from the standard deep learning tool-box. So far, such methods have demonstrated performance close to that of state-of-the-art models while using their interpretable construction to achieve a comparably low learned parameter count. In this work, we propose an unrolled convolutional dictionary learning network (CDLNet) and demonstrate its competitive denoising and joint denoising and demosaicing (JDD) performance both in low and high parameter count regimes. Specifically, we show that the proposed model outperforms state-of-the-art fully convolutional denoising and JDD models when scaled to a similar parameter count. In addition, we leverage the model’s interpretable construction to propose a noise-adaptive parameterization of thresholds in the network that enables state-of-the-art blind denoising performance, and near-perfect generalization on noise-levels unseen during training. Furthermore, we show that such performance extends to the JDD task and unsupervised learning. |
| first_indexed | 2024-12-12T10:50:33Z |
| format | Article |
| id | doaj.art-507cd472c1fe4278bf09f750a997a4c2 |
| institution | Directory Open Access Journal |
| issn | 2644-1322 |
| language | English |
| last_indexed | 2024-12-12T10:50:33Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Signal Processing |
| spelling | doaj.art-507cd472c1fe4278bf09f750a997a4c22022-12-22T00:26:46ZengIEEEIEEE Open Journal of Signal Processing2644-13222022-01-01319621110.1109/OJSP.2022.31728429769957CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and DemosaicingNikola Janjusevic0https://orcid.org/0000-0002-8277-3175Amirhossein Khalilian-Gourtani1https://orcid.org/0000-0003-1376-9583Yao Wang2https://orcid.org/0000-0003-3199-3802New York University Tandon School of Engineering, Electrical and Computer Engineering Department, Brooklyn, NY, USANew York University Tandon School of Engineering, Electrical and Computer Engineering Department, Brooklyn, NY, USANew York University Tandon School of Engineering, Electrical and Computer Engineering Department, Brooklyn, NY, USADeep learning based methods hold state-of-the-art results in low-level image processing tasks, but remain difficult to interpret due to their black-box construction. Unrolled optimization networks present an interpretable alternative to constructing deep neural networks by deriving their architecture from classical iterative optimization methods without use of tricks from the standard deep learning tool-box. So far, such methods have demonstrated performance close to that of state-of-the-art models while using their interpretable construction to achieve a comparably low learned parameter count. In this work, we propose an unrolled convolutional dictionary learning network (CDLNet) and demonstrate its competitive denoising and joint denoising and demosaicing (JDD) performance both in low and high parameter count regimes. Specifically, we show that the proposed model outperforms state-of-the-art fully convolutional denoising and JDD models when scaled to a similar parameter count. In addition, we leverage the model’s interpretable construction to propose a noise-adaptive parameterization of thresholds in the network that enables state-of-the-art blind denoising performance, and near-perfect generalization on noise-levels unseen during training. Furthermore, we show that such performance extends to the JDD task and unsupervised learning.https://ieeexplore.ieee.org/document/9769957/Interpretable deep learningunrolled networksblind denoisingjoint demosaicing and denoisingdictionary learningsparse coding |
| spellingShingle | Nikola Janjusevic Amirhossein Khalilian-Gourtani Yao Wang CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing IEEE Open Journal of Signal Processing Interpretable deep learning unrolled networks blind denoising joint demosaicing and denoising dictionary learning sparse coding |
| title | CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing |
| title_full | CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing |
| title_fullStr | CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing |
| title_full_unstemmed | CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing |
| title_short | CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing |
| title_sort | cdlnet noise adaptive convolutional dictionary learning network for blind denoising and demosaicing |
| topic | Interpretable deep learning unrolled networks blind denoising joint demosaicing and denoising dictionary learning sparse coding |
| url | https://ieeexplore.ieee.org/document/9769957/ |
| work_keys_str_mv | AT nikolajanjusevic cdlnetnoiseadaptiveconvolutionaldictionarylearningnetworkforblinddenoisinganddemosaicing AT amirhosseinkhaliliangourtani cdlnetnoiseadaptiveconvolutionaldictionarylearningnetworkforblinddenoisinganddemosaicing AT yaowang cdlnetnoiseadaptiveconvolutionaldictionarylearningnetworkforblinddenoisinganddemosaicing |