A Review of Super-Resolution Reconstruction Technology Based on Convolutional Neural Networks and Generative Adversarial Networks

DOI: https://doi.org/10.62517/jbdc.202601107

Author(s)

Yao Chen

Affiliation(s)

Huaiyin Institute of Technology, Data Science and Big Data Technology, Huai'an, Jiangsu, China

Abstract

As a key technology in the field of computer vision, image super-resolution reconstruction aims to recover high-fidelity details from low-resolution images and has been widely applied in various fields in recent years. This paper mainly introduces two mainstream deep learning models for super-resolution reconstruction: Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN), and systematically compares their technical characteristics and applicable scenarios. Relying on local feature extraction and parameter sharing mechanisms, CNN models excel in objective indicators such as Peak Signal-to-Noise Ratio (PSNR), with high computational efficiency and stable training. Represented by models like SRCNN and EDSR, they are suitable for scenarios with strict accuracy requirements, such as medical imaging and remote sensing monitoring. Through the adversarial training between generators and discriminators, GAN models introduce perceptual loss to improve visual realism. Models such as SRGAN and ESRGAN can generate rich texture details, but they have issues such as reliance on high-quality data for training and susceptibility to artifact generation. They are more applicable to fields that prioritize subjective experience, such as film and television restoration and game image quality enhancement. This paper further analyzes the differences between the two types of models in terms of computational cost and generalization ability, clarifies the basis for model selection by combining typical application cases, and finally looks forward to the development directions such as lightweight fusion architectures, providing a reference for the practical implementation of the technology.

Keywords

Super-Resolution Reconstruction; Deep Learning; Convolutional Neural Network (CNN); Generative Adversarial Network (GAN); Computer Vision

References

[1] Ponney M V ,Velammal L B ,Kulothungan K . DB-DSCN: Image Tampering Detection Using Dual Branch Deep Stacked Convolutional Network[J].International Journal of Software Engineering and Knowledge Engineering,2025,35(09). [2] Burçak C K . An efficient deep convolutional network model using mask images for multiclass classification of breast cancer ultrasound images[J].Neural Computing and Applications,2025,(prepublish):1-20. [3] Jaincy M E D ,Pattabiraman V . BCDCNN: breast cancer deep convolutional neural network for breast cancer detection using MRI images[J].Scientific Reports,2025,15(1):29014-29014. [4] Singh K ,Sudershan A ,Kumar S , et al. BC-MBINet: A Novel Architecture for Accurate Classification of Breast Cancer with Microscopic Biopsy Images Using Deep Convolutional Neural Networks[J].International Journal of Pattern Recognition and Artificial Intelligence,2025,39(13). [5] Tolgyesi A ,Whyne C ,Hardisty M . Predicting rat lumbar vertebral failure patterns as synthetic μCT images using a deep convolutional generative adversarial network[J].Journal of the Mechanical Behavior of Biomedical Materials,2025,170107116-107116. [6] Zhang J D. Research on Super-Resolution Reconstruction of Single 3D-MRI Image Based on Deep Convolutional Neural Network[D]. Chongqing Medical University, 2024. [7] Zhao Z W. Image Super-Resolution Reconstruction Based on Deep Convolutional Neural Network[J]. Science and Technology Innovation, 2024, (07): 92-95. [8] Yuan X L, Zhang B H, Zhang Q, et al. Infrared Image Super-Resolution Reconstruction Technology Based on Deep Convolutional Neural Network[J]. Infrared Technology, 2023, 45(05): 498-505. [9] Huang Z Y. Research on Single Image Super-Resolution Based on Very Deep Super-Resolution Convolutional Neural Network (in English)[J]. Journal of Measurement Science and Instrumentation, 2022, 13(03): 276-283. [10] Wang X H, Zhao X Y, Wang X Y, et al. Single Image Super-Resolution Reconstruction Using Deep Residual Network with Undecimated Wavelet Edge Learning[J]. Acta Electronica Sinica, 2022, 50(07): 1753-1765. [11] Xu W. Research on Hyperspectral Image Super-Resolution Method Based on Deep Convolutional Neural Network[D]. Tianjin University, 2022. [12] Bai Z Y, Tao J Y. Super-Resolution Reconstruction of Brain MRI Images Using Pseudo-3D Convolutional Network[J]. Journal of Computer-Aided Design & Computer Graphics, 2022, 34(02): 208-216. [13] Yang Y, Xie X, He L, et al. Super-Resolution Reconstruction Combining Continued Fraction Interpolation and Convolutional Neural Network[J]. Journal of Hefei University of Technology (Natural Science Edition), 2021, 44(08): 1146-1152. [14] Cha T B. Research on Super-Resolution Method for Train Detection Images Based on Deep Convolutional Neural Network[D]. Southwest Jiaotong University, 2021. [15] Chen X, Luo B. Multi-Angle Remote Sensing Image Super-Resolution Reconstruction Using Deep Network with Dynamic Upsampling Filter[J]. Geomatics and Information Science of Wuhan University, 2021, 46(11): 1716-1726. [16] Shi J P. Research on Image Super-Resolution Reconstruction Technology Based on Deep Convolutional Network[D]. Nanjing University of Aeronautics and Astronautics, 2021. [17] Zhang J T. Research on Image Super-Resolution Based on Attention Convolutional Neural Network[D]. Harbin Engineering University, 2021. [18] Zou C Z, Huang X S. Hyperspectral Image Super-Resolution Method Based on Deep Residual Convolutional Neural Network[J]. Journal of Fuzhou University (Natural Science Edition), 2020, 48(05): 545-550. [19] Xie C, Zhu H Y. Image Super-Resolution Reconstruction Method Based on Deep Convolutional Neural Network[J]. Transducer and Microsystem Technologies, 2020, 39(09): 142-145.