Improving Image Compression With Adjacent Attention and Refinement Block
Recently, learned image compression algorithms have shown incredible performance compared to classic hand-crafted image codecs. Despite its considerable achievements, the fundamental disadvantage is not optimized for retaining local redundancies, particularly non-repetitive patterns, which have a de...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9845406/ |
| _version_ | 1828004628044709888 |
|---|---|
| author | Afsana Ahsan Jeny Md Baharul Islam Masum Shah Junayed Debashish Das |
| author_facet | Afsana Ahsan Jeny Md Baharul Islam Masum Shah Junayed Debashish Das |
| author_sort | Afsana Ahsan Jeny |
| collection | DOAJ |
| description | Recently, learned image compression algorithms have shown incredible performance compared to classic hand-crafted image codecs. Despite its considerable achievements, the fundamental disadvantage is not optimized for retaining local redundancies, particularly non-repetitive patterns, which have a detrimental influence on the reconstruction quality. This paper introduces the autoencoder-style network-based efficient image compression method, which contains three novel blocks, i.e., adjacent attention block, Gaussian merge block, and decoded image refinement block, to improve the overall image compression performance. The adjacent attention block allocates the additional bits required to capture spatial correlations (both vertical and horizontal) and effectively remove worthless information. The Gaussian merge block assists the rate-distortion optimization performance, while the decoded image refinement block improves the defects in low-resolution reconstructed images. A comprehensive ablation study analyzes and evaluates the qualitative and quantitative capabilities of the proposed model. Experimental results on two publicly available datasets reveal that our method outperforms the state-of-the-art methods on the KODAK dataset (by around 4dB and 5dB) and CLIC dataset (by about 4dB and 3dB) in terms of PSNR and MS-SSIM. |
| first_indexed | 2024-04-10T07:17:55Z |
| format | Article |
| id | doaj.art-db6eb3b10bdc40ef8e5f29474c197d23 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-10T07:17:55Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-db6eb3b10bdc40ef8e5f29474c197d232023-02-25T00:02:57ZengIEEEIEEE Access2169-35362023-01-0111176131762510.1109/ACCESS.2022.31952959845406Improving Image Compression With Adjacent Attention and Refinement BlockAfsana Ahsan Jeny0https://orcid.org/0000-0002-8524-9600Md Baharul Islam1https://orcid.org/0000-0002-9928-5776Masum Shah Junayed2https://orcid.org/0000-0003-3592-4601Debashish Das3https://orcid.org/0000-0002-4237-5711Department of Computer Science and Engineering, Daffodil International University, Dhaka, BangladeshDepartment of Computer Science and Engineering, Daffodil International University, Dhaka, BangladeshDepartment of Computer Engineering, Bahçeşehir University, İstanbul, TurkeyFaculty of Computing, Engineering and the Built Environment, Birmingham City University, Birmingham, U.KRecently, learned image compression algorithms have shown incredible performance compared to classic hand-crafted image codecs. Despite its considerable achievements, the fundamental disadvantage is not optimized for retaining local redundancies, particularly non-repetitive patterns, which have a detrimental influence on the reconstruction quality. This paper introduces the autoencoder-style network-based efficient image compression method, which contains three novel blocks, i.e., adjacent attention block, Gaussian merge block, and decoded image refinement block, to improve the overall image compression performance. The adjacent attention block allocates the additional bits required to capture spatial correlations (both vertical and horizontal) and effectively remove worthless information. The Gaussian merge block assists the rate-distortion optimization performance, while the decoded image refinement block improves the defects in low-resolution reconstructed images. A comprehensive ablation study analyzes and evaluates the qualitative and quantitative capabilities of the proposed model. Experimental results on two publicly available datasets reveal that our method outperforms the state-of-the-art methods on the KODAK dataset (by around 4dB and 5dB) and CLIC dataset (by about 4dB and 3dB) in terms of PSNR and MS-SSIM.https://ieeexplore.ieee.org/document/9845406/Image compressionattention mechanismsGaussian merge blockrefinement blockautoencoder |
| spellingShingle | Afsana Ahsan Jeny Md Baharul Islam Masum Shah Junayed Debashish Das Improving Image Compression With Adjacent Attention and Refinement Block IEEE Access Image compression attention mechanisms Gaussian merge block refinement block autoencoder |
| title | Improving Image Compression With Adjacent Attention and Refinement Block |
| title_full | Improving Image Compression With Adjacent Attention and Refinement Block |
| title_fullStr | Improving Image Compression With Adjacent Attention and Refinement Block |
| title_full_unstemmed | Improving Image Compression With Adjacent Attention and Refinement Block |
| title_short | Improving Image Compression With Adjacent Attention and Refinement Block |
| title_sort | improving image compression with adjacent attention and refinement block |
| topic | Image compression attention mechanisms Gaussian merge block refinement block autoencoder |
| url | https://ieeexplore.ieee.org/document/9845406/ |
| work_keys_str_mv | AT afsanaahsanjeny improvingimagecompressionwithadjacentattentionandrefinementblock AT mdbaharulislam improvingimagecompressionwithadjacentattentionandrefinementblock AT masumshahjunayed improvingimagecompressionwithadjacentattentionandrefinementblock AT debashishdas improvingimagecompressionwithadjacentattentionandrefinementblock |