Deblurred Image Quality Improvement by Learning-based Deblurring Method Utilizing ConvNeXt-V2
Masaki Aoi, Tomio Goto · 2024
Image restoration is a field that has been studied for a long time, especially blurred image restoration is difficult to restore and is still being studied. Therefore, this study aims to improve the accuracy of learning-based blurred image reconstruction. In recent years, various neural networks for image processing have been developed. Among them, we focus on a learning method in a model which performs well in image classification and segmentation tasks. We apply the learning method to a deblurring network to enable more generic blurred image restoration. Experimental results show that pre-trained models record higher PSNR values than models without pre-training, indicating that pre-training with the learning method also works well for the deblurring network.