A new publication from Opto-Electronic Science; DOI 10.29026/oes.2024.230042 discusses data-driven polarimetric imaging.

Polarization is one of the fundamental properties of light waves, manifesting the vector transverse wave nature of light. When light waves interact with objects or media, the wave vector can exhibit intensity variations corresponding to the inherent properties of the material, thereby characterizing the intrinsic features of the target. This unique characteristic, serving as an additional information dimension, has been widely employed in various fields of optical imaging, such as diffusive scattering imaging, remote sensing imaging, biomedical imaging, etc., expanding the application scope of existing optical imaging techniques.

Deep learning, as an advanced machine learning paradigm based on neural networks, achieves exceptional performance in fields such as computer vision by leveraging the non-linear data representation capabilities of convolutional kernel modules for learning and extracting abstract high-dimensional features. Combining data-driven deep learning with polarization imaging based on physical modeling, it effectively promotes the extraction and representation capabilities of feature information in higher-level visual tasks, holding vast potential applications. Therefore, this paper, starting from the essence and trends of data-driven polarization imaging technology, briefly reviews the development in various application domains and explores the paradigm of polarization information extraction and representation based on deep learning in different application fields.

The research group of Prof. Xiaopeng Shao from Xidian University review "Data-driven polarimetric imaging: a review". This review addresses issues in existing polarization imaging applications and provides an overview of the research progress in data-driven polarization imaging, focusing on trends, applications, information utilization, and discussions on future development directions.

Data-driven polarimetric imaging is a novel approach aimed at compensating for the defects and difficulties of a single-information interpretation model. With the exploration of data-driven polarimetric imaging, the application fields and utilization of polarization information have gradually increased. Regarding the input and use of polarization information, the utilization of polarization information has shifted from directly acquired polarization data to preprocessed polarization features. Furthermore, physical modes are crucial during network training, existing physical models are playing an increasingly integral role in guiding and planning the design and training of neural networks. Based on the polarimetric information fused into the network, other physical properties of light have also been introduced in network training, expanding the application domain from the realm of image processing to semantic tasks.

Existing data-driven polarization imaging technologies have gradually found applications in polarization information reconstruction and enhancement, target detection, biomedical imaging, pathological diagnosis, semantic segmentation, diffusive scattering media, 3D reconstruction, reflection removal, and other fields. Due to the high-order non-linear representation capabilities of convolutional neural networks, on the one hand, they can extract higher-dimensional information from complex imaging media and scenes, significantly improving the interpretation and reconstruction of physical properties in challenging environments. This optimization enhances imaging results in low signal-to-noise ratio environments, such as natural settings, scattering media, noise, ambient light interference, low dynamics, and biological tissues. On the other hand, the introduction of polarization information provides additional supplementary information, expanding the application scope of existing intensity-based deep learning algorithms. This expansion includes target segmentation, authentication, camouflage target identification, medical diagnosis, and further extends to new application domains like 3D reconstruction and physical information transformation. This development provides powerful tools and methods for improving imaging quality, enhancing target interpretation accuracy, and driving the progress of emerging application areas.

In the realm of data-driven polarization imaging, as a novel interdisciplinary research area, the complementary strengths of data-driven approaches and physical models effectively enhance existing information interpretation and imaging outcomes, thereby expanding into unknown application domains. In future research, deeper integration with existing physical models can be considered to optimize network training results, improve the interpretability of neural networks, and provide a research foundation for generating more comprehensive synthetic datasets. Various deep learning architectures such as semi-supervised learning, unsupervised learning, transfer learning, multi-task learning, and federated learning play a significant role in reducing dependence on datasets, further improving existing imaging outcomes, and holding significant implications for expanding into new application areas. Simultaneously, leveraging advanced manufacturing technologies like metasurfaces and meta-lenses as novel optoelectronic devices allows for the precise manipulation of light in specific polarization states. This capability enables the customized acquisition, separation, and interpretation of polarized light signals, potentially enhancing the sensitivity and accuracy of acquiring polarization datasets. Therefore, further research into data-driven polarization imaging technology holds substantial exploration space and development potential.

# # # # # #
Prof. Xiaopeng Shao's research team focuses primarily on fundamental theories of computational imaging, the acquisition and interpretation of imaging light field information, and related research and cutting-edge innovations in image information processing, compression, and transmission. The research spans multiple areas, including new-generation optoelectronic imaging, optoelectronic image processing and analysis, and the development and testing of optoelectronic instruments.

The group comprises 15 outstanding faculty members, including leading talents in science and technology innovation for middle-aged and young individuals in Shaanxi Province, and directors of the Chinese Optical Society. Additionally, the team consists of over 80 graduate students at both the Ph.D. and master's levels. The group have undertaken more than 60 research projects at the national and provincial levels, and their work has been published in over 200 papers in high-quality journals such as Nature Communications, Photonics Research, Optics Letters, and Optics Express, accumulating more than 1100 citations. The group has also been granted over 60 patents and received 16 awards at the provincial and ministerial levels, including the Award of Aerospace Science and Technology Progress, Award of the Ministry of Industry and Information Technology Science and Technology Progress.

# # # # # #
Opto-Electronic Science (OES) is a peer-reviewed, open access, interdisciplinary and international journal published by The Institute of Optics and Electronics, Chinese Academy of Sciences as a sister journal of Opto-Electronic Advances (OEA, IF=9.682). OES is dedicated to providing a professional platform to promote academic exchange and accelerate innovation. OES publishes articles, reviews, and letters of the fundamental breakthroughs in basic science of optics and optoelectronics.
# # # # # #

More information: https://www.oejournal.org/oes
Editorial Board: https://www.oejournal.org/oes/editorialboard/list
OES is available on OE journals (https://www.oejournal.org/oes/archive)
Submission of OES may be made using ScholarOne (https://mc03.manuscriptcentral.com/oes)
CN 51-1800/O4
ISSN 2097-0382
Contact Us: oes@ioe.ac.cn
Twitter: @OptoElectronAdv (https://twitter.com/OptoElectronAdv?lang=en)
WeChat: OE_Journal

# # # # # #

Yang K, Liu F, Liang SY et al. Data-driven polarimetric imaging: a review. Opto-Electron Sci 3, 230042 (2024). doi: 10.29026/oes.2024.230042