News >>
Data-driven polarimetric imaging: a review
This study reviews the recent advances in data-driven polarimetric imaging technologies based on a wide range of practical applications. The widespread international research and activity in polarimetric imaging techniques demonstrate their broad applications and interest. Polarization information is increasingly incorporated into convolutional neural networks (CNN) as a supplemental feature of objects to improve performance in computer vision task applications. Polarimetric imaging and deep learning can extract abundant information to address various challenges. Therefore, this article briefly reviews recent developments in data-driven polarimetric imaging, including polarimetric descattering, 3D imaging, reflection removal, target detection, and biomedical imaging. Furthermore, we synthetically analyze the input, datasets, and loss functions and list the existing datasets and loss functions with an evaluation of their advantages and disadvantages. We also highlight the significance of data-driven polarimetric imaging in future research and development.
This review provides an overview of recent efforts to summarize data-driven polarimetric imaging based on seven classifications and discusses them comprehensively from three perspectives. Based on the application fields, the classifications consist of polarimetric descattering, 3D shape reconstruction, reflection removal, restoration, enhancement of polarization information, target detection, biomedical imaging and pathological diagnosis, and semantic segmentation. Subsequently, we synthetically analyze the input, datasets, and loss function, which are crucial in data-driven polarimetric imaging, listing the existing datasets and loss functions with an evaluation of their advantages and disadvantages. In conclusion, deep-learning-based polarimetric imaging introduces polarization information into the convolutional neural network to achieve better performance than traditional intensity imaging, bringing physical interpretability to CNN through physical models. Through research on existing data-driven polarimetric imaging, the study of the corresponding fields can be improved to a higher level, enabling them to enhance high-level visual tasks.
(From:https://www.oejournal.org/article/)
