@inproceedings{eaab972476b340b586291a5c4a028564,
title = "Combining multi-level loss for image denoising",
abstract = "The image processing has witnessed remarkable progress in image denoising. Nevertheless, restoring the visual quality of the image remains a great challenge. Existing methods might fail to obtain the denoised images with high visual quality since they ignore the potential connection with the high-level feature and result in over-smoothing results. Aiming to research whether high-level feature could influence the performance of denoising task, we propose an end-to-end multi-module neural network architecture, which introduces the combination of the high-level and low-level feature in the training process, for image denoising. In order to guide model preserve structural information efficiently, we introduce a hybrid loss, which is designed to restore details from both pixel and feature space. The experimental results show our method improves the visual quality of images and performs well compared with state-of-the-art methods on three benchmarks.",
keywords = "Edge loss, Hybrid loss function, Image denoising",
author = "Fei Li and Nailiang Kuang and Jiangbin Zheng and Qianru Wei and Yue Xi and Yanrong Guo",
note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020.; 10th International Conference on Brain Inspired Cognitive Systems, BICS 2019 ; Conference date: 13-07-2019 Through 14-07-2019",
year = "2020",
doi = "10.1007/978-3-030-39431-8_41",
language = "英语",
isbn = "9783030394301",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer",
pages = "423--432",
editor = "Jinchang Ren and Amir Hussain and Huimin Zhao and Jun Cai and Rongjun Chen and Yinyin Xiao and Kaizhu Huang and Jiangbin Zheng",
booktitle = "Advances in Brain Inspired Cognitive Systems - 10th International Conference, BICS 2019, Proceedings",
}