2000 character limit reached
SDIP: Self-Reinforcement Deep Image Prior Framework for Image Processing (2404.12142v1)
Published 17 Apr 2024 in cs.CV, cs.LG, and eess.IV
Abstract: Deep image prior (DIP) proposed in recent research has revealed the inherent trait of convolutional neural networks (CNN) for capturing substantial low-level image statistics priors. This framework efficiently addresses the inverse problems in image processing and has induced extensive applications in various domains. However, as the whole algorithm is initialized randomly, the DIP algorithm often lacks stability. Thus, this method still has space for further improvement. In this paper, we propose the self-reinforcement deep image prior (SDIP) as an improved version of the original DIP. We observed that the changes in the DIP networks' input and output are highly correlated during each iteration. SDIP efficiently utilizes this trait in a reinforcement learning manner, where the current iteration's output is utilized by a steering algorithm to update the network input for the next iteration, guiding the algorithm toward improved results. Experimental results across multiple applications demonstrate that our proposed SDIP framework offers improvement compared to the original DIP method and other state-of-the-art methods.
- Lose the views: Limited angle ct reconstruction via implicit sinogram completion. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6343–6352, 2018.
- On instabilities of deep learning in image reconstruction and the potential costs of ai. Proceedings of the National Academy of Sciences, 117(48):30088–30095, 2020.
- Computed tomography reconstruction using deep image prior and learned reconstruction methods. Inverse Problems, 36(9):094004, 2020.
- Low-complexity single-image super-resolution based on nonnegative neighbor embedding. 2012.
- On hallucinations in tomographic image reconstruction. IEEE transactions on medical imaging, 40(11):3249–3260, 2021.
- Ct-guided pet parametric image reconstruction using deep neural network without prior training data. In Medical Imaging 2019: Physics of Medical Imaging, volume 10948, pages 238–243. SPIE, 2019.
- Unsupervised pet logan parametric image estimation using conditional deep image prior. Medical Image Analysis, 80:102519, 2022.
- Regularization by architecture: A deep prior approach for inverse problems. Journal of Mathematical Imaging and Vision, 62:456–470, 2020.
- Nonlocally centralized sparse representation for image restoration. IEEE transactions on Image Processing, 22(4):1620–1630, 2012.
- Inverse and ill-posed problems, volume 4. Elsevier, 2014.
- Denoising and regularization via exploiting the structural bias of convolutional generators. arXiv preprint arXiv:1910.14634, 2019.
- Dior: Deep iterative optimization-based residual-learning for limited-angle ct reconstruction. IEEE Transactions on Medical Imaging, 41(7):1778–1790, 2022.
- Deep convolutional neural network for inverse problems in imaging. IEEE transactions on image processing, 26(9):4509–4522, 2017.
- Anisotropic total variation for limited-angle ct reconstruction. In IEEE Nuclear Science Symposuim & Medical Imaging Conference, pages 2232–2238. IEEE, 2010.
- Rethinking deep image prior for denoising. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 5087–5096, 2021.
- Sergei Igorevich Kabanikhin. Definitions and examples of inverse and ill-posed problems. 2008.
- Image super-resolution: A comprehensive review, recent trends, challenges and applications. Information Fusion, 91:230–260, 2023.
- ChuMiao Li. A survey on image deblurring. arXiv preprint arXiv:2202.07456, 2022.
- Enhanced pet imaging using progressive conditional deep image prior. 68(17):175047, 2023.
- Image restoration using total variation regularized deep image prior. In ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7715–7719. Ieee, 2019.
- Deepred: Deep image prior powered by red. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, pages 0–0, 2019.
- The little engine that could: Regularization by denoising (red). SIAM Journal on Imaging Sciences, 10(4):1804–1844, 2017.
- Jonathan Richard Shewchuk et al. An introduction to the conjugate gradient method without the agonizing pain. 1994.
- A global constraint to improve ct reconstruction under non-ideal conditions, 2022.
- Rbp-dip: High-quality ct reconstruction using an untrained neural network with residual back projection and deep image prior. arXiv preprint arXiv:2210.14416, 2022.
- Sparse-view and limited-angle ct reconstruction with untrained networks and deep image prior. Computer Methods and Programs in Biomedicine, 226:107167, 2022.
- Deep image prior. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 9446–9454, 2018.
- Limited-angle ct reconstruction via the l1/l2 minimization, 2021.
- Image reconstruction is a new frontier of machine learning. IEEE transactions on medical imaging, 37(6):1289–1296, 2018.
- The evolution of image reconstruction for ct—from filtered back projection to artificial intelligence. European radiology, 29:2185–2195, 2019.
- Deep radon prior: A fully unsupervised framework for sparse-view ct reconstruction. arXiv preprint arXiv:2401.00135, 2023.
- Multi-scale weighted nuclear norm image restoration. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3165–3174, 2018.
- Simulation tools for two-dimensional experiments in x-ray computed tomography using the forbild head phantom. Physics in Medicine & Biology, 57(13):N237, 2012.
- On single image scale-up using sparse-representations. In Curves and Surfaces: 7th International Conference, Avignon, France, June 24-30, 2010, Revised Selected Papers 7, pages 711–730. Springer, 2012.
- Reference-driven compressed sensing mr image reconstruction using deep convolutional neural networks without pre-training. Sensors, 20(1):308, 2020.