File Download
There are no files associated with this item.
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.1109/TMM.2023.3252271
- Scopus: eid_2-s2.0-85149469495
- WOS: WOS:001133324200038
- Find via
Supplementary
- Citations:
- Appears in Collections:
Article: A Decoupled Multi-Task Network for Shadow Removal
Title | A Decoupled Multi-Task Network for Shadow Removal |
---|---|
Authors | |
Keywords | decomposition Estimation feature decoupling Feature extraction illumination compensation Image decomposition Image restoration Lighting multi-task Multitasking Shadow removal Task analysis |
Issue Date | 3-Mar-2023 |
Publisher | Institute of Electrical and Electronics Engineers |
Citation | IEEE Transactions on Multimedia, 2023 How to Cite? |
Abstract | Shadow removal, which aims to restore the illumination in shadow regions, is challenging due to the diversity of shadows in terms of location, intensity, shape, and size. Different from most multi-task methods, which design elaborate multi-branch or multi-stage structures for better shadow removal, we introduce feature decomposition to learn better feature representations. Specifically, we propose a single-stage and decoupled multi-task network (DMTN) to explicitly learn the decomposed features for shadow removal, shadow matte estimation, and shadow image reconstruction. First, we propose several coarse-to-fine semi-convolution (SMC) modules to capture features sufficient for joint learning of these three tasks. Second, we design a theoretically supported feature decoupling layer to explicitly decouple the learned features into shadow image features and shadow matte features via weight reassignment. Last, these features are converted to a target shadow-free image, affiliated shadow matte, and shadow image, supervised by multi-task joint loss functions. With multi-task collaboration, DMTN effectively recovers the illumination in shadow areas while ensuring the fidelity of non-shadow areas. Experimental results show that DMTN competes favorably with state-of-the-art multi-branch/multi-stage shadow removal methods, while maintaining the simplicity of single-stage methods. We have released our code to encourage future exploration in powerful feature representation for shadow removal <uri>https://github.com/nachifur/DMTN</uri> |
Persistent Identifier | http://hdl.handle.net/10722/338113 |
ISSN | 2023 Impact Factor: 8.4 2023 SCImago Journal Rankings: 2.260 |
ISI Accession Number ID |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Liu, Jiawei | - |
dc.contributor.author | Wang, Qiang | - |
dc.contributor.author | Fan, Huijie | - |
dc.contributor.author | Li, Wentao | - |
dc.contributor.author | Qu, Liangqiong | - |
dc.contributor.author | Tang, Yandong | - |
dc.date.accessioned | 2024-03-11T10:26:21Z | - |
dc.date.available | 2024-03-11T10:26:21Z | - |
dc.date.issued | 2023-03-03 | - |
dc.identifier.citation | IEEE Transactions on Multimedia, 2023 | - |
dc.identifier.issn | 1520-9210 | - |
dc.identifier.uri | http://hdl.handle.net/10722/338113 | - |
dc.description.abstract | <p>Shadow removal, which aims to restore the illumination in shadow regions, is challenging due to the diversity of shadows in terms of location, intensity, shape, and size. Different from most multi-task methods, which design elaborate multi-branch or multi-stage structures for better shadow removal, we introduce feature decomposition to learn better feature representations. Specifically, we propose a single-stage and decoupled multi-task network (DMTN) to explicitly learn the decomposed features for shadow removal, shadow matte estimation, and shadow image reconstruction. First, we propose several coarse-to-fine semi-convolution (SMC) modules to capture features sufficient for joint learning of these three tasks. Second, we design a theoretically supported feature decoupling layer to explicitly decouple the learned features into shadow image features and shadow matte features via weight reassignment. Last, these features are converted to a target shadow-free image, affiliated shadow matte, and shadow image, supervised by multi-task joint loss functions. With multi-task collaboration, DMTN effectively recovers the illumination in shadow areas while ensuring the fidelity of non-shadow areas. Experimental results show that DMTN competes favorably with state-of-the-art multi-branch/multi-stage shadow removal methods, while maintaining the simplicity of single-stage methods. We have released our code to encourage future exploration in powerful feature representation for shadow removal <uri>https://github.com/nachifur/DMTN</uri></p> | - |
dc.language | eng | - |
dc.publisher | Institute of Electrical and Electronics Engineers | - |
dc.relation.ispartof | IEEE Transactions on Multimedia | - |
dc.rights | This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. | - |
dc.subject | decomposition | - |
dc.subject | Estimation | - |
dc.subject | feature decoupling | - |
dc.subject | Feature extraction | - |
dc.subject | illumination compensation | - |
dc.subject | Image decomposition | - |
dc.subject | Image restoration | - |
dc.subject | Lighting | - |
dc.subject | multi-task | - |
dc.subject | Multitasking | - |
dc.subject | Shadow removal | - |
dc.subject | Task analysis | - |
dc.title | A Decoupled Multi-Task Network for Shadow Removal | - |
dc.type | Article | - |
dc.identifier.doi | 10.1109/TMM.2023.3252271 | - |
dc.identifier.scopus | eid_2-s2.0-85149469495 | - |
dc.identifier.eissn | 1941-0077 | - |
dc.identifier.isi | WOS:001133324200038 | - |
dc.identifier.issnl | 1520-9210 | - |