RI: Small: Lensless Cameras --- Enabling Novel Imaging Capabilities with Programmable Masks and Computational Imaging

这个项目从根本上开发了非常薄和灵活的新型相机。今天，成像和摄影的主要技术是使用透镜将光线聚焦到平面传感器上。然而，这种设计需要一定的最小距离--通常是几毫米--来聚焦来自场景的光线以产生高分辨率图像。此外，对于非平面表面的成像，透镜从根本上是不兼容的。通过从成像系统中移除镜头，该项目设计了厚度在100微米量级的相机。这使得在移动摄影、内窥镜检查、显微镜和物联网中对相机厚度有严格要求的新型应用成为可能。此外，通过利用柔性电子技术的进步，该项目正在建造新的相机设计，用于在柔性和曲面上成像。无镜头成像的研究也被纳入到大匹兹堡地区初中和高中的推广活动中，在那里，关于相机制造的工作坊向学生介绍光学和工程的基本概念，同时激励他们追求STEM职业生涯。这项研究侧重于薄型因素和曲面和柔性表面的成像设计。这是通过在传感器附近放置一个精心设计的面具来实现的。通过分析基于掩模成像的基本能力和局限性，设计了高度优化的掩模和相关的重建算法，以解析高分辨率的场景。具体地说，掩模经过优化，可提供高光吞吐量，同时提供可逆的场景到传感器传输功能。还探索了可编程掩模设计，以实现更丰富的传递函数空间，这些传递函数不仅能够获取图像和视频，而且能够获取光场。这项研究使得表征薄型成像的基本极限以及实现这些极限的相关成像架构成为可能。

This project develops fundamentally new types of cameras that are extremely thin and flexible. Today, the primary technology for imaging and photography is by using a lens to focus light onto a planar sensor. However, this design requires a certain minimum distance - often, several millimeters - to focus the light from the scene to produce a high-resolution image. In addition, lenses are fundamentally incompatible for imaging on non-planar surfaces. By removing lenses from the imaging system, this project designs cameras that have thicknesses in the order of 100s of micrometers. This enables novel applications in mobile photography, endoscopy, microscopy, and the internet-of-things where there are stringent requirements on the thickness of the camera. Further, by exploiting advances in flexible electronics, the project is building novel camera designs for imaging on flexible and curved surfaces. The research in lens-less imaging is also incorporated into outreach activities in middle and high schools in the Greater Pittsburgh area, where workshops on camera building introduce students to basics concepts in optics and engineering while inspiring them to pursue a STEM career.This research focuses on designs for imaging at thin form factors and on curved and flexible surfaces. This is achieved by placing a carefully-designed mask in close proximity to the sensor. By analyzing the fundamental capabilities and limitations of mask-based imaging, highly optimized masks and associated reconstruction algorithms are designed to resolve scenes at high resolutions. Specifically, masks are optimized to deliver high light-throughput while providing invertible scene-to-sensor transfer functions. Programmable mask designs are also explored for enabling a richer space of transfer functions that are capable of acquiring not just images and videos, but also light fields. The research makes it possible to characterize the fundamental limits of thin form-factor imaging and the associated imaging architectures that achieve them.