Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks
There are many studies that seek to enhance a low resolution image to a high resolution image in the area of super-resolution. As deep learning technologies have recently shown impressive results on the image interpolation and restoration field, recent studies are focusing on convolutional neural ne...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-01-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/11/3/1092 |
| _version_ | 1797407541457584128 |
|---|---|
| author | Seonjae Kim Dongsan Jun Byung-Gyu Kim Hunjoo Lee Eunjun Rhee |
| author_facet | Seonjae Kim Dongsan Jun Byung-Gyu Kim Hunjoo Lee Eunjun Rhee |
| author_sort | Seonjae Kim |
| collection | DOAJ |
| description | There are many studies that seek to enhance a low resolution image to a high resolution image in the area of super-resolution. As deep learning technologies have recently shown impressive results on the image interpolation and restoration field, recent studies are focusing on convolutional neural network (CNN)-based super-resolution schemes to surpass the conventional pixel-wise interpolation methods. In this paper, we propose two lightweight neural networks with a hybrid residual and dense connection structure to improve the super-resolution performance. In order to design the proposed networks, we extracted training images from the DIVerse 2K (DIV2K) image dataset and investigated the trade-off between the quality enhancement performance and network complexity under the proposed methods. The experimental results show that the proposed methods can significantly reduce both the inference speed and the memory required to store parameters and intermediate feature maps, while maintaining similar image quality compared to the previous methods. |
| first_indexed | 2024-03-09T03:43:01Z |
| format | Article |
| id | doaj.art-fe7760e5a38a4b0f9e8c906db01b283e |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T03:43:01Z |
| publishDate | 2021-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-fe7760e5a38a4b0f9e8c906db01b283e2023-12-03T14:38:02ZengMDPI AGApplied Sciences2076-34172021-01-01113109210.3390/app11031092Single Image Super-Resolution Method Using CNN-Based Lightweight Neural NetworksSeonjae Kim0Dongsan Jun1Byung-Gyu Kim2Hunjoo Lee3Eunjun Rhee4Department of Convergence IT Engineering, Kyungnam University, Changwon 51767, KoreaDepartment of Information and Communication Engineering, Kyungnam University, Changwon 51767, KoreaDepartment of IT Engineering, Sookmyung Women’s University, Seoul 04310, KoreaIntelligent Convergence Research Lab., Electronics and Telecommunications Research Institute (ETRI), Daejeon 34129, KoreaIntelligent Convergence Research Lab., Electronics and Telecommunications Research Institute (ETRI), Daejeon 34129, KoreaThere are many studies that seek to enhance a low resolution image to a high resolution image in the area of super-resolution. As deep learning technologies have recently shown impressive results on the image interpolation and restoration field, recent studies are focusing on convolutional neural network (CNN)-based super-resolution schemes to surpass the conventional pixel-wise interpolation methods. In this paper, we propose two lightweight neural networks with a hybrid residual and dense connection structure to improve the super-resolution performance. In order to design the proposed networks, we extracted training images from the DIVerse 2K (DIV2K) image dataset and investigated the trade-off between the quality enhancement performance and network complexity under the proposed methods. The experimental results show that the proposed methods can significantly reduce both the inference speed and the memory required to store parameters and intermediate feature maps, while maintaining similar image quality compared to the previous methods.https://www.mdpi.com/2076-3417/11/3/1092deep learningconvolutional neural networkslightweight neural networksingle image super-resolutionimage enhancementimage restoration |
| spellingShingle | Seonjae Kim Dongsan Jun Byung-Gyu Kim Hunjoo Lee Eunjun Rhee Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks Applied Sciences deep learning convolutional neural networks lightweight neural network single image super-resolution image enhancement image restoration |
| title | Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks |
| title_full | Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks |
| title_fullStr | Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks |
| title_full_unstemmed | Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks |
| title_short | Single Image Super-Resolution Method Using CNN-Based Lightweight Neural Networks |
| title_sort | single image super resolution method using cnn based lightweight neural networks |
| topic | deep learning convolutional neural networks lightweight neural network single image super-resolution image enhancement image restoration |
| url | https://www.mdpi.com/2076-3417/11/3/1092 |
| work_keys_str_mv | AT seonjaekim singleimagesuperresolutionmethodusingcnnbasedlightweightneuralnetworks AT dongsanjun singleimagesuperresolutionmethodusingcnnbasedlightweightneuralnetworks AT byunggyukim singleimagesuperresolutionmethodusingcnnbasedlightweightneuralnetworks AT hunjoolee singleimagesuperresolutionmethodusingcnnbasedlightweightneuralnetworks AT eunjunrhee singleimagesuperresolutionmethodusingcnnbasedlightweightneuralnetworks |