Joint Restoration Method Based on Super-resolution Network and Diffusion Model

Mengran Xin; Chenghan Li; Mengyao Xin

doi:10.62051/dwx7ap48

Authors

Mengran Xin
Chenghan Li
Mengyao Xin

DOI:

https://doi.org/10.62051/dwx7ap48

Keywords:

Super-resolution network; diffusion model; image restoration; underwater robot; degradation model.

Abstract

To address the challenges of low image resolution obtained by underwater robots and poor mask inpainting effect, a joint restoration method based on super-resolution network and diffusion model is proposed. Through the degradation model and deep learning, the mapping between high and low resolution images is achieved, and combined with the reverse optimization ability of the diffusion model, the image clarity and physical authenticity are significantly improved. Experiments show that the proposed method outperforms existing technologies in terms of PSNR, SSIM, UIQM and LPIPS, especially in the restoration effect under complex occlusion scenarios. This research effectively enhances the visual perception ability of underwater robots and provides new ideas for image processing in complex water areas.

Downloads

Download data is not yet available.

References

[1] Tang J ,He K ,Tian M , et al.Serialized attention and masked residual network structure for free-form image inpainting[J].Knowledge-Based Systems,2025,316113385-113385.

[2] Kodati P ,Puli K V ,Chiplunkar R , et al.Robust to outlier image inpainting for interface detection in primary separation vessel[J].Journal of Process Control,2025,150103426-103426.

[3] Shao J ,Zhang H ,Miao J .Region gradient-guided diffusion model for underwater image enhancement[J].Machine Vision and Applications,2025,36(2):38-38.

[4] Wang Y, Zhang J, Cao Y, et al. A deep CNN method for underwater image enh ancement[C]//2017 IEEE International Conference on Image Processing (ICIP). IEEE, 2017: 1382-1386.

[5] Yunting L ,Zhuang Z ,Binghua S , et al.Super resolution of underwater image based on generating generative adversarial networks[C]//Beijing Institute of Technology (China),2023:

[6] Qiao Lukuan, Liu Aimin, Wang Zijun, et al. Vision inspection System of underwater vehicle based on FPGA [J]. Microprocessor, 2019,45(03):47-51. (in Chinese)

[7] Fei Y ,Lihui Z ,Panwang P , et al.Low bit-rate compression of underwater image based on human visual system[J].Signal Processing: Image Communication,2021,91116082-.

[8] Li J ,Liu Y ,Wu X , et al.Fault diagnosis in open circuit of inverters on electrical discharge milling machines using adaptive Gaussian wavelet convolutional network[J].Measurement,2025,248116856-116856.

[9] Baudouin L ,Imba A ,Mercado A , et al.Lipschitz stability of an inverse problem of transmission waves with variable jumps[J].Inverse Problems,2025,41(4):045007-045007.

[10] Cui SH, Lu Bo, ZHANG Mingyue, et al. 360° image quality and aesthetic evaluation method based on multi-modal fusion [J/OL]. Computer engineering, 1-9 [2025-04-14]. http://kns.cnki.net/kcms/detail/31.128 9.tp.20250314.1347.003.html.

[11] Zhang F ,Yuan Q ,Zhang X . Mamba-DDPM-BSA: Diffusion model based boundary sampling algorithm for imbalanced classification [J]. Expert Systems With Applications, 2025, 274 126926-126926.

[12] Huang Shaozong. Based on the depth study of image restoration method research [D]. Jiangxi university of science and technology, 2024. The DOI: 10.27176 /, dc nki. Gnfyc. 2024.000195.

[13] Huang Shaozong. Based on the depth study of image restoration method research [D]. Jiangxi university of science and technology, 2024. The DOI: 10.27176 /, dc nki. Gnfyc. 2024.000195.

[14] Zhao Xiaojie. Based on attribute augmented zero sample study [D]. Nanjing university of science and technology, 2022. The DOI: 10.27241 /, dc nki. Gnjgu. 2022.001434.

[15] Zhang Xukun. Research on cloud removal algorithm of H-alpha full-day solar image based on image conversion model [D]. The central university for nationalities, 2023. DOI: 10.27667 /, dc nki. Gzymu. 2023.000173