Data-Driven Numerical Analysis Using Machine Learning for Robust Signal and Image Processing Systems

Sunitha Maram; Dr.Tejaswini Pradhan; Mannem Ram Reddy; Dr.G.V.V jagannadha Rao; Vijaya Krishna Gembali; Monira Reyaz

doi:10.29284/kc0cs017

Authors

Sunitha Maram Research Scholar Department of Mathematics, Kalinga University, Naya Raipur, Vill: Kotni, Chhattisgarh India
Dr.Tejaswini Pradhan Assistant Professor, Department of Mathematics, Kalinga University, Naya Raipur, vill: Kotni, Chhattisgarh India
Mannem Ram Reddy Secondary Grade Govt. Teacher, Mpps Newbanjara hills, Penpahad mandal, Suryapet district, Telangana State
Dr.G.V.V jagannadha Rao Associate Professor, Department of Mathematics, The ICFAI University, Raipur, Dist : Raipur, State Chhattisgarh, India
Vijaya Krishna Gembali Assistant Professor, Department of Mathematics. Sasi Institute of Technology and Engineering, Tadepalligudem, India
Monira Reyaz Department of Mathematics, Kalinga University ,naya Raipur, Vill: kotni, chhatisgarh india

DOI:

https://doi.org/10.29284/kc0cs017

Keywords:

Data-driven numerical analysis; machine learning; signal processing; image processing; robustness; hybrid models

Abstract

Background:
The increasing sophistication of contemporary signal and image processing problems has revealed shortcomings of the traditional numerical analysis, especially observed with nonlinear and high-dimensional nonlinear and noisy data. Although machine learning has a robust data-driven modeling ability, purely learning-based algorithms are not always numerically stable and interpretable.

Methods:
This paper compares the data-driven numerical analysis frameworks that combine machine learning and known numerical modeling principles together. Peer-reviewed studies published since the year 2019 were reviewed accordingly into a systematic narrative review, limiting itself to the applications in the areas of signal denoising, feature extraction, image reconstruction, and enhancement. The reported metrics of signal fidelity, image quality, noise resistance and computational efficiency were used as performance trends.

Results:
Always, hybrid numerical-learning methods have better robustness and reconstruction quality than classical numerical methods and are more stable than fully data-driven models. Even though end-to-end learning remains the most successful as per the peak performance, hybrid frameworks offer a better stability, accuracy, and interpretability balance.

Conclusion:
Numerical analysis provide a sound and scalable paradigm of next generation signal and image processing systems because they bind numerical structure, and data adaptivity, and therefore are applicable in real world and safety crucial systems.

References

1. Wang N, Sun J. Model Meets Deep Learning in Image Inverse Problems. CSIAM Trans Appl Math. 2020;1(3):365–386. Doi: https://doi.org/10.4208/csiam-am.2020-0016

2. Sun Y, Wu Z, Xu X, Wohlberg B, Kamilov US. Scalable Plug-and-Play ADMM with Convergence Guarantees. IEEE Trans Comput Imaging. 2021;7:849–863. Doi: https://doi.org/10.1109/TCI.2021.3094062

3. Nair P, Gavaskar RG, Chaudhury KN. Fixed-Point and Objective Convergence of Plug-and-Play Algorithms. IEEE Trans Comput Imaging. 2021;7:337–348. Doi: https://doi.org/10.1109/TCI.2021.3066053

4. Kamyab S, Azimifar Z, Sabzi R, Fieguth P. Deep learning methods for inverse problems. PeerJ Comput Sci. 2022;8:e951. Doi: https://doi.org/10.7717/peerj-cs.951

5. Benfenati A, Cascarano P. Constrained Plug-and-Play Priors for Image Restoration. J Imaging. 2024;10(2):50. Doi: https://doi.org/10.3390/jimaging10020050

6. Luijten B, Chennakeshava N, Eldar YC, Mischi M, van Sloun RJG. Ultrasound Signal Processing: From Models to Deep Learning. Ultrasound Med Biol. 2023;49(3):677–698. Doi: https://doi.org/10.1016/j.ultrasmedbio.2022.11.003

7. Mardani M, et al. Machine Learning for Medical Image Reconstruction. In: Proceedings/Book. Springer; 2022. Doi: https://doi.org/10.1007/978-3-031-17247-2

8. Shlezinger N, Eldar YC. Model-Based Deep Learning. Foundations Trends Signal Process. 2023. Doi: https://doi.org/10.1561/9781638282655

9. Elad M. Image Denoising: The Deep Learning Revolution and Beyond. SIAM Rev. 2023. Doi: https://doi.org/10.1137/23M1545859

10. Peng X, et al. Improving diffusion models for inverse problems using ... ACM Proc. 2024. Doi: https://doi.org/10.5555/3692070.3693707

11. Adler J, Öktem O. Learned primal–dual reconstruction. IEEE Trans Med Imaging. 2021.

Doi: https://doi.org/10.1109/TMI.2021.3055100

12. Monga V, Li Y, Eldar YC. Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing. IEEE Signal Process Mag. 2021. Doi: https://doi.org/10.1109/MSP.2020.3016905

13. Karniadakis GE, Kevrekidis IG, Lu L, Perdikaris P, Wang S, Yang L. Physics-informed machine learning. Nat Rev Phys. 2021. Doi: https://doi.org/10.1038/s42254-021-00314-5

14. Bora A, Jalal A, Price E, Dimakis AG. Compressed sensing using generative models. IEEE Trans Inf Theory. 2021. Doi: https://doi.org/10.1109/TIT.2020.3018789

15. Chung H, Kim J, Monga V, Ye JC. Diffusion posterior sampling for general noisy inverse problems. ICLR. 2023. Doi: https://doi.org/10.48550/arXiv.2209.14687

16. Ryu E, Liu J, Wang S, Chen X, Wang Z, Yin W. Plug-and-play methods provably converge. ICML. 2020. Doi: https://doi.org/10.48550/arXiv.1905.05406

17. Gilton D, Ongie G, Willett R. Neumann networks for linear inverse problems. IEEE Trans Comput Imaging. 2022. Doi: https://doi.org/10.1109/TCI.2022.3151484

18. Baraniuk RG, et al. Learning sparse signal models. Proc IEEE. 2021.

Doi: https://doi.org/10.1109/JPROC.2021.3051346

19. Mehranian A, et al. Deep learning-based image reconstruction. Phys Med Biol. 2022.

Doi: https://doi.org/10.1088/1361-6560/ac6b5a

20. Gong E, et al. Direct attenuation and scatter correction of whole-body PET images using deep learning. Phys Med Biol. 2023. Doi: https://doi.org/10.1088/1361-6560/aca87b

21. Chen Z, Liu Y, Sun Q. Efficient deep unfolding networks for inverse problems. IEEE Trans Neural Netw Learn Syst. 2024. Doi: https://doi.org/10.1109/TNNLS.2023.3320196

22. Lanusse F, et al. Probabilistic neural operators. ICML. 2021. Doi: https://doi.org/10.48550/arXiv.2106.06221

23. Koutini K, et al. Explainable deep learning for signal processing. IEEE Signal Process Mag. 2024. Doi: https://doi.org/10.1109/MSP.2024.3351127

24. Shen L, et al. Deep learning for inverse problems: A review. IEEE Trans Pattern Anal Mach Intell. 2024. Doi: https://doi.org/10.1109/TPAMI.2023.3314686

25. Ongie G, et al. Deep learning techniques for inverse problems in imaging. IEEE J Sel Areas Inf Theory. 2025. Doi: https://doi.org/10.1109/JSAIT.2024.3389021

26. Arridge S, Maass P, Öktem O, Schönlieb CB. Solving inverse problems using data-driven models. Acta Numer. 2021. Doi: https://doi.org/10.1017/S0962492921000039

27. Cohen T, et al. Learning data-driven signal priors for inverse problems. IEEE Trans Signal Process. 2021. Doi: https://doi.org/10.1109/TSP.2021.3058604

28. Kutyniok G, Petersen P, Raslan M, Schneider R. A theoretical analysis of deep neural networks and inverse problems. Found Comput Math. 2022. Doi: https://doi.org/10.1007/s10208-021-09504-8

29. Knoll F, et al. Deep-learning methods for parallel magnetic resonance imaging reconstruction. IEEE Signal Process Mag. 2021. Doi: https://doi.org/10.1109/MSP.2021.3066256

30. Heide F, et al. Fast and flexible computational imaging with learned priors. Proc IEEE. 2022. Doi: https://doi.org/10.1109/JPROC.2022.3149103

31. Bouche D, et al. Generalization guarantees for model-based deep learning. IEEE Trans Inf Theory. 2023. Doi: https://doi.org/10.1109/TIT.2023.3270914

32. Hurault S, Leclaire A, Papadakis N. Proximal plug-and-play methods: Stability and convergence. SIAM J Imaging Sci. 2022. Doi: https://doi.org/10.1137/21M140909X

33. Adler J, Öktem O. Uncertainty quantification in learned inverse problems. Inverse Problems. 2023. Doi: https://doi.org/10.1088/1361-6420/acb0c1

34. Maier A, et al. Learning with known operators reduces data requirements. Nat Mach Intell. 2024. Doi: https://doi.org/10.1038/s42256-024-00808-6

35. Chen Y, Pock T. Trainable nonlinear reaction diffusion. IEEE Trans Pattern Anal Mach Intell. 2021. Doi: https://doi.org/10.1109/TPAMI.2021.3064792

36. Li Y, et al. Deep unfolding for inverse problems: A survey. IEEE Trans Neural Netw Learn Syst. 2022. Doi: https://doi.org/10.1109/TNNLS.2022.3140510

37. Ravishankar S, Bresler Y. Learning sparsifying transforms. IEEE Trans Signal Process. 2020.

Doi: https://doi.org/10.1109/TSP.2020.2968923

38. Ongie G, et al. Deep learning for inverse problems: Stability and convergence. IEEE Signal Process Mag. 2022. Doi: https://doi.org/10.1109/MSP.2022.3145707

39. Bigdeli SA, et al. Robustness of deep learning for image restoration. IEEE Trans Image Process. 2023. Doi: https://doi.org/10.1109/TIP.2023.3261248

40. Wang G, Ye JC, Mueller K, Fessler JA. Image reconstruction is a new frontier of machine learning. IEEE Trans Med Imaging. 2025. Doi: https://doi.org/10.1109/TMI.2024.3412091