光子是可探测场景信息的最小单位，实现单光子灵敏度、高时空间分辨率成像可突破现有观测能力极限，为生物医学、空间探测、作战侦察等重大应用领域带来革命性成像技术。. 受到采集系统时空带宽积局限，现有单光子成像技术难以实现单光子灵敏度高时空分辨率的成像。本项目通过光学、信号处理、机器学习等学科交叉，以计算摄像原理为核心，构建单光子时空计算成像系统，突破采集系统时空带宽积局限，具体包括：从波色-爱因斯坦统计出发，通过联合建模单光子响应非线性效应和时空压缩感知理论，建立基于光子时空概率分布的非线性单光子时空压缩感知模型；研制基于"零惯性"时间压缩编码和快照式空间压缩编码的瞬变时空联合压缩编码技术；构建基于内容指导的单光子时空数据库，提出基于变分的长短时记忆神经网络算法，实现高压缩比高精度的单光子时空计算重建。 . 本项目所提方案的主要优势在于：零读出噪声、单光子级成像灵敏度和超高时空带宽积。

Photon is the smallest discrete quantum of light. Capturing light in single-photon sensitivity with high spatiotemporal resolution could break the limit of existing imaging technologies and enable revolutionary imaging technologies with wide-applications in biomedical science, space exploration, and surveillance reconnaissance, etc...Limiting by the spatiotemporal-bandwidth product, current single-photon imaging technologies are unable to realize high spatiotemporal resolution. Based on computational photography theory, as an interdisciplinary subject composed of the optics, signal processing, and machine learning, this project plans to develop a novel single-photon computational imaging system of high spatiotemporal resolution, aiming to break the limit of spatiotemporal-bandwidth product of current imaging system. Specifically, in imaging model, based on the Bose-Einstein statistics, this project plan to propose a novel nonlinear spatiotemporal compressive imaging model investigating both the nonlinear response characteristic of single-photon imaging and the spatiotemporal photon statistics. In imaging technique, this project plans to propose "non-inertness" temporal compressive imaging technique, snapshot spatial compressive imaging technique, and develop a novel transient spatiotemporal compressive imaging technology through jointly combining them. In reconstruction, this project plans to construct a content-aware single-photon spatiotemporal database, a long short-term memory network based spatiotemporal reconstruction sub-model, and constructing a variational neural network model based on the single-photon spatiotemporal statistics, aiming to realize high-compression-ratio high-fidelity computational single-photon spatiotemporal reconstruction...Compared with conventional imaging method, this project aims to provide a novel single-photon spatiotemporal imaging technique which is superior in no-readout-noise, single-photon sensitivity and ultra-high spatiotemporal-bandwidth product.