Manuscript received February 20, 2024; revised May 17, 2024; accepted June 18, 2024; published November 25, 2024.

Abstract—This review summarizes the objectives of applying deep learning in the inverse problem of superresolution reconstruction, including basic concepts, principles, and common model architectures. The rapid development of deep learning technology has brought a breakthrough for super-resolution reconstruction, which can model complex mapping relationships and automatically extract features. In addition, deep learning also has wide applicability and potential for image denoising and repair. This paper also discusses the challenges that deep learning faces in inverse problem imaging, such as data requirements and computational complexity, and suggests directions for future research, such as model complexity, cross-domain learning, and interpretability. To address these challenges, future research can improve the model performance by exploring more complex deep learning architectures to enhance model capacity while optimizing model complexity, leveraging external data to enhance model generalization, improving model interpretability by integrating expert knowledge with traditional methods, and enhancing the robustness of the model in complex conditions. With the above improvements, deep learning will provide more accurate, high-quality, and interpretable super-resolution reconstruction solutions, driving the field of inverse problem imaging.

Keywords—deep learning, inverse problem imaging, superresolution reconstruction, image denoising, image restoration

Cite: Rongfu Wang and Mary Jane C. Samonte, "A Review of Deep Learning in Inverse Problem Imaging Using Super-Resolution Reconstruction Algorithms," Journal of Image and Graphics, Vol. 12, No. 4, pp. 396-409, 2024.

Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.