CIF: Small: Sequential and Compound Estimation for Computational Imaging Systems

成像技术在生物、化学、材料科学、医学和其他领域的广泛应用是由人类与生俱来的视觉信息处理能力推动的。然而，成像通常受到成像过程对样品造成的损害的限制，例如，在扫描光束显微镜中，由于入射离子或电子束而产生的溅射。用传统的方法，在获得足够的信息以形成有用的图像之前，样品可能会被销毁。这个项目开发了减少显微镜系统中必要剂量的方法。这可能使得在成像之前就对分子或生物物种成像成为可能，或者提高成像的速度，从而揭示新的动态现象。该项目还包括通过说明性写作、普通听众演讲、跨学科培训和编写教材向广大受众传播成果。当测量信号是由于检测到的粒子数量较少时，泊松模型经常被利用。该项目的一个主要前提是将此类探测计数的泊松建模与弱电流束流中相互作用粒子数量的泊松建模相结合。该项目为由此产生的复合估计问题开发了理论和方法。该项目探索了使用时间分辨传感来减少由于复合过程中的源散粒噪声而引起的不确定性。此外，时间分辨传感能够将数据收集过程分解为一系列破坏性较小的步骤，该项目将开发顺序估计和自适应数据收集的理论和方法。分析将受到概念验证实验的启发和启发。通过某些无量子力学相互作用的测量，可以达到在不损坏样本的情况下了解样本的最终极限。该项目将开发顺序无相互作用测量的理论，以确定它们何时可能优于传统测量。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The widespread use of imaging technologies across biology, chemistry, material science, medicine, and other fields is motivated by the great innate human ability for visual information processing. Imaging is often limited, however, by the damage to a sample caused by the imaging process, such as sputtering due to the incident ion or electron beam in scanned beam microscopy. With conventional methods, a sample may be destroyed before enough information has been gained to form a useful image. This project develops methods for reducing the necessary dose in microscopy systems. This may make it possible to image molecular or biological species not before imaged, or to increase the speed of imaging so that new dynamic phenomena are revealed. The project also includes dissemination of results to broad audiences through expository writing, general-audience talks, cross-disciplinary training, and production of teaching materials. Poisson models are often exploited when measured signals are due to small numbers of detected particles. A major premise of this project is to combine the Poisson modeling of such detection counts with Poisson modeling of the numbers of interacting particles in low-current beams. The project develops theory and methods for the resulting compound estimation problems. The project explores the use of time-resolved sensing to reduce the uncertainty due to the source shot noise in compound processes. Furthermore, time-resolved sensing enables the decomposition of a data collection process into a sequence of less-damaging steps, and the project will develop theory and methods for sequential estimation and adaptive data collection. Analysis will inspire and be informed by proof-of-concept experiments. The ultimate limit of learning about a sample without damaging it is achieved with certain quantum-mechanical interaction-free measurements. The project will develop theory for sequential interaction-free measurements to establish when they may be superior to conventional measurements.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.