{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T20:03:28Z","timestamp":1776888208956,"version":"3.51.2"},"reference-count":27,"publisher":"MDPI AG","issue":"7","license":[{"start":{"date-parts":[[2023,6,30]],"date-time":"2023-06-30T00:00:00Z","timestamp":1688083200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"US National Science Foundation","award":["DMS 2038118"],"award-info":[{"award-number":["DMS 2038118"]}]},{"name":"US National Science Foundation","award":["DMS 2208294"],"award-info":[{"award-number":["DMS 2208294"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["J. Imaging"],"abstract":"<jats:p>In recent years, large convolutional neural networks have been widely used as tools for image deblurring, because of their ability in restoring images very precisely. It is well known that image deblurring is mathematically modeled as an ill-posed inverse problem and its solution is difficult to approximate when noise affects the data. Really, one limitation of neural networks for deblurring is their sensitivity to noise and other perturbations, which can lead to instability and produce poor reconstructions. In addition, networks do not necessarily take into account the numerical formulation of the underlying imaging problem when trained end-to-end. In this paper, we propose some strategies to improve stability without losing too much accuracy to deblur images with deep-learning-based methods. First, we suggest a very small neural architecture, which reduces the execution time for training, satisfying a green AI need, and does not extremely amplify noise in the computed image. Second, we introduce a unified framework where a pre-processing step balances the lack of stability of the following neural-network-based step. Two different pre-processors are presented. The former implements a strong parameter-free denoiser, and the latter is a variational-model-based regularized formulation of the latent imaging problem. This framework is also formally characterized by mathematical analysis. Numerical experiments are performed to verify the accuracy and stability of the proposed approaches for image deblurring when unknown or not-quantified noise is present; the results confirm that they improve the network stability with respect to noise. In particular, the model-based framework represents the most reliable trade-off between visual precision and robustness.<\/jats:p>","DOI":"10.3390\/jimaging9070133","type":"journal-article","created":{"date-parts":[[2023,6,30]],"date-time":"2023-06-30T16:06:37Z","timestamp":1688141197000},"page":"133","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":7,"title":["Ambiguity in Solving Imaging Inverse Problems with Deep-Learning-Based Operators"],"prefix":"10.3390","volume":"9","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6261-7717","authenticated-orcid":false,"given":"Davide","family":"Evangelista","sequence":"first","affiliation":[{"name":"Department of Mathematics, University of Bologna, 40126 Bologna, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4283-2994","authenticated-orcid":false,"given":"Elena","family":"Morotti","sequence":"additional","affiliation":[{"name":"Department of Political and Social Sciences, University of Bologna, 40125 Bologna, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9951-3564","authenticated-orcid":false,"given":"Elena Loli","family":"Piccolomini","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Engineering, University of Bologna, 40126 Bologna, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"James","family":"Nagy","sequence":"additional","affiliation":[{"name":"Department of Mathematics, Emory University, Atlanta, GA 30322, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1968","published-online":{"date-parts":[[2023,6,30]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","unstructured":"Hansen, P.C., Nagy, J.G., and O\u2019leary, D.P. (2006). Deblurring Images: Matrices, Spectra, and Filtering, SIAM.","DOI":"10.1137\/1.9780898718874"},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"168","DOI":"10.1007\/s11263-011-0502-7","article-title":"Non-uniform deblurring for shaken images","volume":"98","author":"Whyte","year":"2012","journal-title":"Int. J. Comput. Vis."},{"key":"ref_3","unstructured":"Raskar, R., Agrawal, A., and Tumblin, J. (2006). ACM Siggraph 2006 Papers, ACM."},{"key":"ref_4","doi-asserted-by":"crossref","unstructured":"Bertero, M., Boccacci, P., and De Mol, C. (2021). Introduction to Inverse Problems in Imaging, CRC Press.","DOI":"10.1201\/9781003032755"},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"2103","DOI":"10.1007\/s11263-022-01633-5","article-title":"Deep image deblurring: A survey","volume":"130","author":"Zhang","year":"2022","journal-title":"Int. J. Comput. Vis."},{"key":"ref_6","doi-asserted-by":"crossref","unstructured":"Engl, H.W., Hanke, M., and Neubauer, A. (1996). Regularization of Inverse Problems, Springer Science & Business Media.","DOI":"10.1007\/978-94-009-1740-8"},{"key":"ref_7","unstructured":"Scherzer, O., Grasmair, M., Grossauer, H., Haltmeier, M., and Lenzen, F. (2009). Variational Methods in Imaging, Springer."},{"key":"ref_8","doi-asserted-by":"crossref","unstructured":"Venkatakrishnan, S.V., Bouman, C.A., and Wohlberg, B. (2013, January 3\u20135). Plug-and-play priors for model based reconstruction. Proceedings of the 2013 IEEE Global Conference on Signal and Information Processing, Austin, TX, USA.","DOI":"10.1109\/GlobalSIP.2013.6737048"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"1872","DOI":"10.1109\/LSP.2017.2763583","article-title":"A plug-and-play priors approach for solving nonlinear imaging inverse problems","volume":"24","author":"Kamilov","year":"2017","journal-title":"IEEE Signal Process. Lett."},{"key":"ref_10","first-page":"126967","article-title":"Plug-and-Play gradient-based denoisers applied to CT image enhancement","volume":"422","author":"Cascarano","year":"2022","journal-title":"Appl. Math. Comput."},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Hansen, P.C. (2010). Discrete Inverse Problems: Insight and Algorithms, SIAM.","DOI":"10.1137\/1.9780898718836"},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"084002","DOI":"10.1088\/1361-6420\/ab1c6b","article-title":"A nonconvex penalization algorithm with automatic choice of the regularization parameter in sparse imaging","volume":"35","author":"Lazzaro","year":"2019","journal-title":"Inverse Probl."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Hradi\u0161, M., Kotera, J., Zemc\u0131k, P., and \u0160roubek, F. (2015, January 7\u201310). Convolutional neural networks for direct text deblurring. Proceedings of the BMVC, Swansea, UK.","DOI":"10.5244\/C.29.6"},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"103134","DOI":"10.1016\/j.cviu.2020.103134","article-title":"Single-image deblurring with neural networks: A comparative survey","volume":"203","author":"Koh","year":"2021","journal-title":"Comput. Vis. Image Underst."},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Sun, J., Cao, W., Xu, Z., and Ponce, J. (2015, January 7\u201312). Learning a convolutional neural network for non-uniform motion blur removal. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA.","DOI":"10.1109\/CVPR.2015.7298677"},{"key":"ref_16","doi-asserted-by":"crossref","unstructured":"Zhang, K., Luo, W., Zhong, Y., Ma, L., Stenger, B., Liu, W., and Li, H. (2020, January 13\u201319). Deblurring by realistic blurring. Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA.","DOI":"10.1109\/CVPR42600.2020.00281"},{"key":"ref_17","unstructured":"Gottschling, N.M., Antun, V., Hansen, A.C., and Adcock, B. (2020). The Troublesome Kernel\u2014On Hallucinations, No Free Lunches and the Accuracy-Stability Trade-Off in Inverse Problems. arXiv."},{"key":"ref_18","doi-asserted-by":"crossref","unstructured":"Amjad, J., Sokoli\u0107, J., and Rodrigues, M.R. (2018, January 3\u20137). On deep learning for inverse problems. Proceedings of the 2018 26th European Signal Processing Conference (EUSIPCO), Roma, Italy.","DOI":"10.23919\/EUSIPCO.2018.8553376"},{"key":"ref_19","unstructured":"Bastounis, A., Hansen, A.C., and Vla\u010di\u0107, V. (2021). The mathematics of adversarial attacks in AI\u2013Why deep learning is unstable despite the existence of stable neural networks. arXiv."},{"key":"ref_20","unstructured":"Evangelista, D., Nagy, J., Morotti, E., and Piccolomini, E.L. (2022). To be or not to be stable, that is the question: Understanding neural networks for inverse problems. arXiv."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Ronneberger, O., Fischer, P., and Brox, T. (2015, January 5\u20139). U-net: Convolutional networks for biomedical image segmentation. Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Germany.","DOI":"10.1007\/978-3-319-24574-4_28"},{"key":"ref_22","doi-asserted-by":"crossref","unstructured":"Chen, L., Chu, X., Zhang, X., and Sun, J. (2022, January 23\u201327). Simple baselines for image restoration. Proceedings of the Computer Vision\u2014ECCV 2022: 17th European Conference, Tel Aviv, Israel. Proceedings, Part VII.","DOI":"10.1007\/978-3-031-20071-7_2"},{"key":"ref_23","doi-asserted-by":"crossref","unstructured":"Morotti, E., Evangelista, D., and Loli Piccolomini, E. (2021). A green prospective for learned post-processing in sparse-view tomographic reconstruction. J. Imaging, 7.","DOI":"10.20944\/preprints202107.0265.v1"},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Nah, S., Kim, T.H., and Lee, K.M. (2017, January 21\u201326). Deep Multi-Scale Convolutional Neural Network for Dynamic Scene Deblurring. Proceedings of the CVPR, Honolulu, HI, USA.","DOI":"10.1109\/CVPR.2017.35"},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"108","DOI":"10.1162\/neco.1995.7.1.108","article-title":"Training with noise is equivalent to Tikhonov regularization","volume":"7","author":"Bishop","year":"1995","journal-title":"Neural Comput."},{"key":"ref_26","doi-asserted-by":"crossref","unstructured":"Gonzalez, R.C. (2009). Digital Image Processing, Pearson Education India.","DOI":"10.1117\/1.3115362"},{"key":"ref_27","unstructured":"Wang, Z., Simoncelli, E.P., and Bovik, A.C. (November, January 29). Multiscale structural similarity for image quality assessment. Proceedings of the Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, Pacific Grove, CA, USA."}],"container-title":["Journal of Imaging"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2313-433X\/9\/7\/133\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,10]],"date-time":"2025-10-10T20:03:43Z","timestamp":1760126623000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2313-433X\/9\/7\/133"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,6,30]]},"references-count":27,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2023,7]]}},"alternative-id":["jimaging9070133"],"URL":"https:\/\/doi.org\/10.3390\/jimaging9070133","relation":{},"ISSN":["2313-433X"],"issn-type":[{"value":"2313-433X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,6,30]]}}}