Prof. Yu Lei has Made a Series of Important Progress in the Field of Extreme Condition Imaging

Recently, IEEE CVPR 2021 (2021 International Conference on Computer Vision and Pattern Recognition) released the list of accepted papers. The work, ''Event-based Synthetic Aperture With a Hybrid Network'' from the team of Dr. Yu Lei (Associate Professor in School of Electronics and Information, Wuhan University) has been accepted as the ORAL presentation and will be presented online in June of this year. Prof. Yu Lei is the corresponding author and two of his students, Mr. Zhang Xiang and Mr. Liao Wei, are the co-first authors. This work is done by collaborating with Prof. Yang Wen from the School of Electronics and Information and Prof. Xia Guisong from the School of Computer Science.
Extreme condition imaging, including occluding, high-speed imaging, etc., has great potential for the robot, autonomous driving, military, aerospace, and other fields. In view of the occluded imaging problem, the camera-based synthetic aperture imaging (SAI) is used to capture the occluded scene from multi-view images, and the imaging of the occluded target can be reconstructed by re-focusing and reconstruction. However, due to the low dynamic range and low frame rate of traditional cameras, synthetic aperture imaging systems based on traditional cameras can not deal with dense occlusions and extreme lighting conditions, leading to remarkable imaging degradation.
This work proposes an event-based synthetic aperture imaging system (E-SAI) to address problems of dense occlusion and extreme lighting conditions. A theoretical analysis of E-SAI is also given. Meanwhile,  considering the low signal-to-noise ratio of asynchronous event stream signal, we also propose a hybrid network (spike neural network and convolutional neural network) to reconstruct high-quality images from collected events. Compared with the SAI based on traditional cameras, the proposed E-SAI system largely improves the imaging quality by a large margin (6-10 dB on PSNR), which effectively solves the problem of occluded imaging under extreme conditions.
It is reported that IEEE CVPR is a top academic conference in the field of computer vision and pattern recognition, is one of the most influential international academic conferences in the field of artificial intelligence and has been recognized as a CCF-A conference by the Chinese Computer Society. The acceptance rate of IEEE CVPR is less than 25%, including Poster, Spotlight, and Oral, of which Oral paper acceptance rate is between 3 and 5%.
Prof. Yu Lei has made a series of progress in the field of extreme condition imaging in recent years. In 2020, his students, Ms. Wang Bishan and Mr. He Jingwei worked together to reconstruct extremely blurry images based on event cameras, solving the problem of high-quality imaging in extremely high-speed scenes. The paper "Event Enhanced High-Quality Image Recovery" was received by the ECCV (European International Conference on Computer Vision, Acceptance Rate <30%) and presented online in July of 2020.