A Study on the Super Resolution Combining Spatial Attention and Channel Attention
Existing CNN-based super resolution methods have low emphasis on high-frequency features, resulting in poor performance for contours and textures. To solve this problem, this paper proposes single image super resolution using an attention mechanism that emphasizes high-frequency features and a featu...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-03-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/13/6/3408 |
| _version_ | 1797613693331046400 |
|---|---|
| author | Dongwoo Lee Kyeongseok Jang Soo Young Cho Seunghyun Lee Kwangchul Son |
| author_facet | Dongwoo Lee Kyeongseok Jang Soo Young Cho Seunghyun Lee Kwangchul Son |
| author_sort | Dongwoo Lee |
| collection | DOAJ |
| description | Existing CNN-based super resolution methods have low emphasis on high-frequency features, resulting in poor performance for contours and textures. To solve this problem, this paper proposes single image super resolution using an attention mechanism that emphasizes high-frequency features and a feature extraction process with different depths. In order to emphasize the high-frequency features of the channel and space, it is composed of CSBlock that combines channel attention and spatial attention. Attention block using 10 CSBlocks was used for high-frequency feature extraction. In order to extract various features with different degrees of feature emphasis from insufficient low-resolution features, features were extracted from structures connected with different numbers of attention blocks. The extracted features were expanded through sub-pixel convolution to create super resolution images, and learning was performed through <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mi>L</mi><mn>1</mn></msub></mrow></semantics></math></inline-formula> loss. Compared to the existing deep learning method, it showed improved results in several high-frequency features such as small object outlines and line patterns. In PSNR and SSIM, it showed about 11% to 26% improvement over the existing Bicubic interpolation and about 1 to 2% improvement over VDSR and EDSR. |
| first_indexed | 2024-03-11T06:59:22Z |
| format | Article |
| id | doaj.art-3e8abf68a45841c78a4eedc0783f8d31 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-11T06:59:22Z |
| publishDate | 2023-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-3e8abf68a45841c78a4eedc0783f8d312023-11-17T09:21:15ZengMDPI AGApplied Sciences2076-34172023-03-01136340810.3390/app13063408A Study on the Super Resolution Combining Spatial Attention and Channel AttentionDongwoo Lee0Kyeongseok Jang1Soo Young Cho2Seunghyun Lee3Kwangchul Son4Department of Plasma Bio Display, Kwangwoon University, 20 Gwangun-ro, Nowon-gu, Seoul 01897, Republic of KoreaDepartment of Plasma Bio Display, Kwangwoon University, 20 Gwangun-ro, Nowon-gu, Seoul 01897, Republic of KoreaDepartment of Information Contents, Kwangwoon University, 20 Gwangun-ro, Nowon-gu, Seoul 01897, Republic of KoreaIngenium College, Kwangwoon University, 20 Gwangun-ro, Nowon-gu, Seoul 01897, Republic of KoreaDepartment of Information Contents, Kwangwoon University, 20 Gwangun-ro, Nowon-gu, Seoul 01897, Republic of KoreaExisting CNN-based super resolution methods have low emphasis on high-frequency features, resulting in poor performance for contours and textures. To solve this problem, this paper proposes single image super resolution using an attention mechanism that emphasizes high-frequency features and a feature extraction process with different depths. In order to emphasize the high-frequency features of the channel and space, it is composed of CSBlock that combines channel attention and spatial attention. Attention block using 10 CSBlocks was used for high-frequency feature extraction. In order to extract various features with different degrees of feature emphasis from insufficient low-resolution features, features were extracted from structures connected with different numbers of attention blocks. The extracted features were expanded through sub-pixel convolution to create super resolution images, and learning was performed through <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mi>L</mi><mn>1</mn></msub></mrow></semantics></math></inline-formula> loss. Compared to the existing deep learning method, it showed improved results in several high-frequency features such as small object outlines and line patterns. In PSNR and SSIM, it showed about 11% to 26% improvement over the existing Bicubic interpolation and about 1 to 2% improvement over VDSR and EDSR.https://www.mdpi.com/2076-3417/13/6/3408super resolutionchannel attentionspatial attentionparallel structure |
| spellingShingle | Dongwoo Lee Kyeongseok Jang Soo Young Cho Seunghyun Lee Kwangchul Son A Study on the Super Resolution Combining Spatial Attention and Channel Attention Applied Sciences super resolution channel attention spatial attention parallel structure |
| title | A Study on the Super Resolution Combining Spatial Attention and Channel Attention |
| title_full | A Study on the Super Resolution Combining Spatial Attention and Channel Attention |
| title_fullStr | A Study on the Super Resolution Combining Spatial Attention and Channel Attention |
| title_full_unstemmed | A Study on the Super Resolution Combining Spatial Attention and Channel Attention |
| title_short | A Study on the Super Resolution Combining Spatial Attention and Channel Attention |
| title_sort | study on the super resolution combining spatial attention and channel attention |
| topic | super resolution channel attention spatial attention parallel structure |
| url | https://www.mdpi.com/2076-3417/13/6/3408 |
| work_keys_str_mv | AT dongwoolee astudyonthesuperresolutioncombiningspatialattentionandchannelattention AT kyeongseokjang astudyonthesuperresolutioncombiningspatialattentionandchannelattention AT sooyoungcho astudyonthesuperresolutioncombiningspatialattentionandchannelattention AT seunghyunlee astudyonthesuperresolutioncombiningspatialattentionandchannelattention AT kwangchulson astudyonthesuperresolutioncombiningspatialattentionandchannelattention AT dongwoolee studyonthesuperresolutioncombiningspatialattentionandchannelattention AT kyeongseokjang studyonthesuperresolutioncombiningspatialattentionandchannelattention AT sooyoungcho studyonthesuperresolutioncombiningspatialattentionandchannelattention AT seunghyunlee studyonthesuperresolutioncombiningspatialattentionandchannelattention AT kwangchulson studyonthesuperresolutioncombiningspatialattentionandchannelattention |