Study on Wavelet Convolution-Based Underwater Image Denoising

Wei Li, Haiyan Xie, Jiaxi Li
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:491-497, 2025.

Abstract

The processing of underwater images is critical for marine science, seafloor mapping, and underwater rescue operations. However, underwater optical images often suffer from poor quality due to light absorption and scattering caused by suspended particles. Additionally, due to technical limitations and environmental interference, underwater robots often capture images where light has been reflected and refracted multiple times before reaching the camera, further exacerbating noise. To address these challenges, this paper proposes an innovative underwater image processing model that combines wavelet convolution and dilated convolution for noise reduction. The model employs wavelet transformation to decompose images into high- and low-frequency components for preliminary processing, followed by the use of dilated convolution to extract noise and image features. This approach effectively removes noise from underwater images. Experimental results demonstrate that this method can adaptively handle illumination and detail information across different scales, addressing challenges such as uneven lighting, low contrast color distortion, and suspended particle noise. The processed images exhibit significantly improved clarity and contrast, even in complex underwater environments.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-li25h, title = {Study on Wavelet Convolution-Based Underwater Image Denoising}, author = {Li, Wei and Xie, Haiyan and Li, Jiaxi}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {491--497}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/li25h/li25h.pdf}, url = {https://proceedings.mlr.press/v278/li25h.html}, abstract = {The processing of underwater images is critical for marine science, seafloor mapping, and underwater rescue operations. However, underwater optical images often suffer from poor quality due to light absorption and scattering caused by suspended particles. Additionally, due to technical limitations and environmental interference, underwater robots often capture images where light has been reflected and refracted multiple times before reaching the camera, further exacerbating noise. To address these challenges, this paper proposes an innovative underwater image processing model that combines wavelet convolution and dilated convolution for noise reduction. The model employs wavelet transformation to decompose images into high- and low-frequency components for preliminary processing, followed by the use of dilated convolution to extract noise and image features. This approach effectively removes noise from underwater images. Experimental results demonstrate that this method can adaptively handle illumination and detail information across different scales, addressing challenges such as uneven lighting, low contrast color distortion, and suspended particle noise. The processed images exhibit significantly improved clarity and contrast, even in complex underwater environments.} }
Endnote
%0 Conference Paper %T Study on Wavelet Convolution-Based Underwater Image Denoising %A Wei Li %A Haiyan Xie %A Jiaxi Li %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-li25h %I PMLR %P 491--497 %U https://proceedings.mlr.press/v278/li25h.html %V 278 %X The processing of underwater images is critical for marine science, seafloor mapping, and underwater rescue operations. However, underwater optical images often suffer from poor quality due to light absorption and scattering caused by suspended particles. Additionally, due to technical limitations and environmental interference, underwater robots often capture images where light has been reflected and refracted multiple times before reaching the camera, further exacerbating noise. To address these challenges, this paper proposes an innovative underwater image processing model that combines wavelet convolution and dilated convolution for noise reduction. The model employs wavelet transformation to decompose images into high- and low-frequency components for preliminary processing, followed by the use of dilated convolution to extract noise and image features. This approach effectively removes noise from underwater images. Experimental results demonstrate that this method can adaptively handle illumination and detail information across different scales, addressing challenges such as uneven lighting, low contrast color distortion, and suspended particle noise. The processed images exhibit significantly improved clarity and contrast, even in complex underwater environments.
APA
Li, W., Xie, H. & Li, J.. (2025). Study on Wavelet Convolution-Based Underwater Image Denoising. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:491-497 Available from https://proceedings.mlr.press/v278/li25h.html.

Related Material