DAIU-Net: Distributed Attention-Enabled Deep Learning Model for Medical and Satellite Image Denoising

View at publisher Summarise with VIRAT
Authors: Sharanabasav Hosamani, Shashidhar Sonnad
Cited by: 0
References: 18

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Image denoising aids in the accurate reconstruction of the image from the noisy observation through the accurate elimination of the noisy components. Despite the advantages leveraged by the conventional image-denoising methods, there exist certain drawbacks, including the manual parameter tuning requirement, local convergence with increased computational complexity, and overfitting problems. Hence, to mitigate such drawbacks, an effective Distributed Attention-enabled Imperialistic U-Net (DAIU-Net) model is proposed in this research. Specifically, the Hybrid Distributed Attention (HDA) enhanced the denoising performance by selectively focusing on the most informative features, thereby reducing the computational complexity and improving the denoising performance. Additionally, the incorporation of Residual Flow (ResFl) descriptor-based feature extraction enhances the image denoising performance and reduces the data complexity challenges. Moreover, the utilization of the Imperialistic Learning Optimization (ILO) effectively fine-tunes the hyperparameters of DAIU-Net to address the local convergence issues. Besides, the modified U-Net aids in the better generation of the relevant features responsible for accurate denoising. Nevertheless, the performance evaluation in terms of performance metrics visualized the superiority of DAIU-Net by attaining 54.99dB PSNR, 0.81 SSIM, and 52.21dB SDME for medical images using the Alzheimer's dataset.

Fields of Study & Subjects

Image and Signal Denoising Methods Image Enhancement Techniques Advanced Image Processing Techniques

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
01–09
Publication date
21 January 2026
Language
EN