New transformer and binarized networks advance hyperspectral image reconstruction
ByPulseAugur Editorial·[6 sources]·
Researchers have developed several novel deep learning architectures for hyperspectral image reconstruction, aiming to improve efficiency and performance. These methods, including sparse transformers and binarized networks, address limitations of traditional CNNs in capturing long-range dependencies and spectral similarities. Several approaches leverage transformer architectures with specialized attention mechanisms to better model spectral correlations and spatial sparsity, often outperforming existing state-of-the-art methods in both accuracy and computational cost. Some also focus on resource-limited deployment through techniques like binarization.
AI
IMPACT
These advancements in efficient hyperspectral image reconstruction could enable new applications in fields like remote sensing, medical imaging, and autonomous systems.
RANK_REASON
Multiple arXiv papers introducing new models and techniques for hyperspectral image reconstruction.
arXiv:2203.04845v4 Announce Type: replace Abstract: Many algorithms have been developed to solve the inverse problem of coded aperture snapshot spectral imaging (CASSI), i.e., recovering the 3D hyperspectral images (HSIs) from a 2D compressive measurement. In recent years, learni…
arXiv cs.CV
TIER_1English(EN)·Yuanhao Cai, Jing Lin, Zudi Lin, Haoqian Wang, Yulun Zhang, Hanspeter Pfister, Radu Timofte, Luc Van Gool·
arXiv:2204.07908v2 Announce Type: replace Abstract: Existing leading methods for spectral reconstruction (SR) focus on designing deeper or wider convolutional neural networks (CNNs) to learn the end-to-end mapping from the RGB image to its hyperspectral image (HSI). These CNN-bas…
arXiv:2305.10299v4 Announce Type: replace Abstract: Existing deep learning models for hyperspectral image (HSI) reconstruction achieve good performance but require powerful hardwares with enormous memory and computational resources. Consequently, these methods can hardly be deplo…
arXiv cs.CV
TIER_1English(EN)·Yuanhao Cai, Jing Lin, Xiaowan Hu, Haoqian Wang, Xin Yuan, Yulun Zhang, Radu Timofte, Luc Van Gool·
arXiv:2111.07910v3 Announce Type: replace-cross Abstract: Hyperspectral image (HSI) reconstruction aims to recover the 3D spatial-spectral signal from a 2D measurement in the coded aperture snapshot spectral imaging (CASSI) system. The HSI representations are highly similar and c…
arXiv cs.CV
TIER_1English(EN)·Xiaowan Hu, Yuanhao Cai, Jing Lin, Haoqian Wang, Xin Yuan, Yulun Zhang, Radu Timofte, Luc Van Gool·
arXiv:2203.02149v3 Announce Type: replace-cross Abstract: The rapid development of deep learning provides a better solution for the end-to-end reconstruction of hyperspectral image (HSI). However, existing learning-based methods have two major defects. Firstly, networks with self…
arXiv cs.CV
TIER_1English(EN)·Yuanhao Cai, Jing Lin, Haoqian Wang, Xin Yuan, Henghui Ding, Yulun Zhang, Radu Timofte, Luc Van Gool·
arXiv:2205.10102v4 Announce Type: replace-cross Abstract: In coded aperture snapshot spectral compressive imaging (CASSI) systems, hyperspectral image (HSI) reconstruction methods are employed to recover the spatial-spectral signal from a compressed measurement. Among these algor…