Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network
The single image super‐resolution (SISR) methods based on the deep convolutional neural network (CNN) have recently achieved significant improvements in accuracy, advancing the state of the art. However, these deeper models are computationally expensive and require a large number of parameters. Acco...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2019-05-01
|
| Series: | IET Image Processing |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/iet-ipr.2018.5907 |
| _version_ | 1818114857696755712 |
|---|---|
| author | Dan Guo Yanxiong Niu Pengyan Xie |
| author_facet | Dan Guo Yanxiong Niu Pengyan Xie |
| author_sort | Dan Guo |
| collection | DOAJ |
| description | The single image super‐resolution (SISR) methods based on the deep convolutional neural network (CNN) have recently achieved significant improvements in accuracy, advancing the state of the art. However, these deeper models are computationally expensive and require a large number of parameters. Accordingly, they demand more memory and are unsuitable for on‐chip devices. In this study, a novel SISR method using a deeply recursive CNN with skip connections and a network in network structure is proposed. The deeply recursive CNN with skip connections is adopted for the image feature extraction at both local and global levels. Parallelised 1 × 1 CNNs, usually called a network in network structure, are adopted for image reconstruction. Specifically, recursive learning is utilised to control the number of model parameters needed and residual learning is used to ease the difficulty of training. The proposed method performs favourably against the state‐of‐the‐art methods in terms of computational speed and accuracy. It significantly outperforms the previous methods by a large margin, while demanding far fewer parameters. This model requires less memory and is friendly to on‐chip devices. |
| first_indexed | 2024-12-11T03:57:24Z |
| format | Article |
| id | doaj.art-33c1970e28754b52bb22792ed8a3a143 |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-12-11T03:57:24Z |
| publishDate | 2019-05-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-33c1970e28754b52bb22792ed8a3a1432022-12-22T01:21:45ZengWileyIET Image Processing1751-96591751-96672019-05-011371201120910.1049/iet-ipr.2018.5907Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in networkDan Guo0Yanxiong Niu1Pengyan Xie2School of Instrumentation and OptoelectronicEngineering, Beihang UniversityBeijing100191People's Republic of ChinaSchool of Instrumentation and OptoelectronicEngineering, Beihang UniversityBeijing100191People's Republic of ChinaSchool of Instrumentation and OptoelectronicEngineering, Beihang UniversityBeijing100191People's Republic of ChinaThe single image super‐resolution (SISR) methods based on the deep convolutional neural network (CNN) have recently achieved significant improvements in accuracy, advancing the state of the art. However, these deeper models are computationally expensive and require a large number of parameters. Accordingly, they demand more memory and are unsuitable for on‐chip devices. In this study, a novel SISR method using a deeply recursive CNN with skip connections and a network in network structure is proposed. The deeply recursive CNN with skip connections is adopted for the image feature extraction at both local and global levels. Parallelised 1 × 1 CNNs, usually called a network in network structure, are adopted for image reconstruction. Specifically, recursive learning is utilised to control the number of model parameters needed and residual learning is used to ease the difficulty of training. The proposed method performs favourably against the state‐of‐the‐art methods in terms of computational speed and accuracy. It significantly outperforms the previous methods by a large margin, while demanding far fewer parameters. This model requires less memory and is friendly to on‐chip devices.https://doi.org/10.1049/iet-ipr.2018.5907accurate image super‐resolutiondeeply recursive CNNskip connectionsingle image super‐resolution methodsdeep convolutional neural networkon‐chip devices |
| spellingShingle | Dan Guo Yanxiong Niu Pengyan Xie Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network IET Image Processing accurate image super‐resolution deeply recursive CNN skip connection single image super‐resolution methods deep convolutional neural network on‐chip devices |
| title | Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network |
| title_full | Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network |
| title_fullStr | Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network |
| title_full_unstemmed | Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network |
| title_short | Speedy and accurate image super‐resolution via deeply recursive CNN with skip connection and network in network |
| title_sort | speedy and accurate image super resolution via deeply recursive cnn with skip connection and network in network |
| topic | accurate image super‐resolution deeply recursive CNN skip connection single image super‐resolution methods deep convolutional neural network on‐chip devices |
| url | https://doi.org/10.1049/iet-ipr.2018.5907 |
| work_keys_str_mv | AT danguo speedyandaccurateimagesuperresolutionviadeeplyrecursivecnnwithskipconnectionandnetworkinnetwork AT yanxiongniu speedyandaccurateimagesuperresolutionviadeeplyrecursivecnnwithskipconnectionandnetworkinnetwork AT pengyanxie speedyandaccurateimagesuperresolutionviadeeplyrecursivecnnwithskipconnectionandnetworkinnetwork |