CIF: Small: Sampling and Inference Methods for Spatiotemporal Single-Photon Imaging

材料、器件和制造技术的最新进步导致了一种新兴的固态传感器，它可以探测空间和时间中的单个光子。由于它们的单光子灵敏度、超快的速度和快速提高的空间分辨率，这些新型的单光子传感器(SPS)在基于荧光的生物成像、飞行时间3D计算机视觉、高速摄像和天文学等领域具有巨大的应用潜力。该项目的目标是基于新兴的SPS方案为光学成像建立严格的信号处理基础，类似于照相胶片中的卤化银颗粒，SPS的每个像素都具有二进制响应，仅揭示局部光强的一位随机信息。通过像素阵列和高时间采样率，SPS生成大量时空比特，对原始视觉信息进行采样和编码。研究人员研究信号采样和推理中的模型、理论和算法，以应对与SPS相关的挑战。该项目的具体目标包括：(1)建立SP在获取给定时空带宽的光强场时的采样定理；(2)确定性能界限和成像性能与关键设备指标之间的精确权衡；(3)设计自适应传感方案以提高成像性能；以及(4)开发能够有效地对由SP生成的海量比特流进行高效“解码”的离线和在线成像算法。

Recent advances in materials, devices and fabrication technologies have led to an emerging class of solid-state sensors that can detect individual photons in space and time. Thanks to their single-photon sensitivity, ultra-fast speed, and rapidly increasing spatial resolutions, these new single photon sensors (SPS) have great potentials in a wide range of applications, including fluorescence-based bio-imaging, time-of-flight 3D computer vision, high-speed videography, and astronomy. The goal of this project is to build rigorous signal processing foundations for optical imaging based on the emerging SPS scheme.Analogous to silver-halide grains in photographic film, each pixel of the SPS has a binary response, revealing only one-bit and stochastic information of the local light intensity. With an array of pixels and high temporal sampling rates, the SPS generates a massive spatiotemporal volume of bits that sample and encode the original visual information. The investigator studies models, theory, and algorithms in signal sampling and inference to address the challenges associated with the SPS. Specific objectives of this project involve: (1) establishing sampling theorems for the SPS in acquiring light intensity fields of given spatiotemporal bandwidths; (2) identifying performance bounds and the precise tradeoffs between imaging performance and key device metrics; (3) designing adaptive sensing schemes to improve imaging performances; and (4) developing both offline and online image formation algorithms that can efficiently "decode" the massive bitstreams generated by the SPS.