Optical flow-assisted multi-level fusion network for Light Field image angular reconstruction
文献类型：期刊论文
| 作者 | Liu, Deyang1,3,4; Mao, Yifan4; Huang, Yan2; Cao, Liqun4; Wang, Yuanzhi4; Fang, Yuming3 |
| 刊名 | SIGNAL PROCESSING-IMAGE COMMUNICATION
 |
| 出版日期 | 2023-11-01 |
| 卷号 | 119页码:11 |
| 关键词 | Light field image Angular reconstruction Optical flow Multi-level fusion network |
| ISSN号 | 0923-5965 |
| DOI | 10.1016/j.image.2023.117031 |
| 通讯作者 | Fang, Yuming(fa0001ng@e.ntu.edu.sg) |
| 英文摘要 | Light Field (LF) imaging can record both the intensities and directions of light rays in a single exposure, which has received extensive attentions. However, the limited angular resolution becomes the primary bottleneck for the wide-spread applications of LF imaging. To this end, this paper proposes a novel optical flow-assisted multi-level fusion network for LF angular reconstruction. In our method, we propose to infer the multi-angular optical flows to explore long-range dependency of LF sub-aperture images (SAIs) for high-quality angular reconstruction. By aligning the SAIs in multi-angular directions, the geometric consistency of reconstructed dense LF can be preserved. Moreover, a multi-level fusion framework for LF angular reconstruction is introduced, which consists of two stages, namely texture-optical flow feature fusion and parallax structure-information fusion. The former firstly extracts the texture and optical flow features from the reconstructed coarse LF and then fuses these two features by using the proposed texture-optical flow fusion-block. The latter further blends the LF parallax structure information with the fused texture and optical flow features using the proposed parallax structure-information fusion network. Comprehensive experiments on both real-world and synthetic LF scenes demonstrate the superiority of the proposed method for reconstructing high-quality dense LF. Moreover, practical application on depth estimation also validates that our method can recover more texture details, particularly for some occlusion regions. |
| WOS关键词 | SUPERRESOLUTION |
| 资助项目 | National Natural Science Foundation of China[62171002] ; National Natural Science Foundation of China[62132006] ; Shenzhen Municipal Science and Technology Innovation Council[2021Szvup051] ; STCSM[SKLSFO2021-05] ; Anhui Provincial Key Laboratory of Network and Information Security[AHNIS2023002] ; University Discipline Top Talent Program of Anhui[gxbjZD2022034] ; Anhui Outstanding Youth Fund by Colleges and Universities[2022AH030106] |
| WOS研究方向 | Engineering |
| 语种 | 英语 |
| WOS记录号 | WOS:001069612700001 |
| 出版者 | ELSEVIER |
| 资助机构 | National Natural Science Foundation of China ; Shenzhen Municipal Science and Technology Innovation Council ; STCSM ; Anhui Provincial Key Laboratory of Network and Information Security ; University Discipline Top Talent Program of Anhui ; Anhui Outstanding Youth Fund by Colleges and Universities |
| 源URL | [http://ir.ia.ac.cn/handle/173211/53128]  |
| 专题 | 多模态人工智能系统全国重点实验室 |
| 通讯作者 | Fang, Yuming |
| 作者单位 | 1.Anhui Normal Univ, Natl Anhui Prov Key Lab Network & Informat Secur, Wuhu 240002, Peoples R China 2.Chinese Acad Sci CASIA, Natl Lab Pattern Recognit NLPR, Inst Automat, Beijing 100190, Peoples R China 3.Jiangxi Univ Finance & Econ, Sch Informat Management, Nanchang 330000, Jiangxi, Peoples R China 4.Anqing Normal Univ, Sch Comp & Informat, Anqing 246000, Peoples R China |
| 推荐引用方式 GB/T 7714 | Liu, Deyang,Mao, Yifan,Huang, Yan,et al. Optical flow-assisted multi-level fusion network for Light Field image angular reconstruction[J]. SIGNAL PROCESSING-IMAGE COMMUNICATION,2023,119:11. |
| APA | Liu, Deyang,Mao, Yifan,Huang, Yan,Cao, Liqun,Wang, Yuanzhi,&Fang, Yuming.(2023).Optical flow-assisted multi-level fusion network for Light Field image angular reconstruction.SIGNAL PROCESSING-IMAGE COMMUNICATION,119,11. |
| MLA | Liu, Deyang,et al."Optical flow-assisted multi-level fusion network for Light Field image angular reconstruction".SIGNAL PROCESSING-IMAGE COMMUNICATION 119(2023):11. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。