Development of an Image Enhancement and Denoising System for Decommissioning Operations Using Deep Learning

Abstract

This research developed a GUI-based image enhancement and denoising system that enables practical application of deep learning–based methods in field operations by allowing flexible switching between multiple models. The proposed system integrates four networks—Noise2Noise，SRCNN，FSRCNN， and RED-Net—enabling users to select the most suitable method depending on operating conditions. In evaluation experiments using low-quality images obtained from the Fukushima Daiichi Nuclear Power Station，Noise2Noise demonstrated particularly strong denoising performance，significantly improving image quality. Furthermore，the GUI design reduced operation steps by 50% and execution time by 50% compared with conventional CUI-based workflows ， achieving substantial improvements in work efficiency. Comparative experiments among the models also confirmed that the proposed system combines both practicality and flexibility.