期刊
NEUROCOMPUTING
卷 524, 期 -, 页码 59-68出版社
ELSEVIER
DOI: 10.1016/j.neucom.2022.12.050
关键词
Single image super-resolution; Neural architecture search; Internal learning
Driven by the learning ability of deep networks, generalized models have been proposed for single-image super-resolution tasks using external datasets. However, models trained solely on external data may struggle to super-resolve images in unfamiliar domains. To address this limitation, internal learning approaches have been proposed, but they often suffer from poor performance due to a fixed architecture. Therefore, we propose a novel training process that combines external and internal learning, allowing the network to adapt to specific test images. Our approach outperforms state-of-the-art super-resolution algorithms in various image domains and produces impressive results in terms of texture and color accuracy.
Fueled by the powerful learning ability of deep networks, generalized models have been proposed that use external datasets for single-image super-resolution tasks. However, a model trained only with exter-nal data may have difficulty in super-resolving images in a domain that differs from the training data. To solve this drawback, several methods have been proposed for internal learning approaches that learn the weights of the network accordance with the test image. Despite these attempts to adapt to specific images using internal learning, they suffer from poor performance due to lack of flexibility that comes from using a fixed architecture regardless of image domain. We thus propose a novel training process that includes external and internal learning. Our internal learning process finds a suitable network archi-tecture and trains the weights for each unseen test image. The overall training process allows the net-work to learn obtain knowledge from external data and internal learning in a balanced manner. In blind and non-blind experiments, our proposed method outperforms state-of-the-art super-resolution algorithms in various image domains with different kernels. Our proposed approach obtains impressive results in terms of expressing detailed texture and accurate color in images from various domains.(c) 2022 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据