A novel approach to programmable imaging using MOEMS

Vasileios T. Nasis

doi:10.17918/etd-2842

New advancements in science are frequently sparked by the invention of new instruments. Possibly the most important scientific instrument of the past fifty years is the digital computer. Among the computers many uses and impacts, digital imaging has revolutionized images and photography, merging computer processing and optical images. In this thesis, we merge an additional reconfigurable micro-mechanical domain into the digital imaging system, introducing a novel imaging method called Programmable Imaging. With our imaging method, we selectively sample the object plane, by utilizing state-of-the-art Micro-Optical-Electrical-Mechanical Systems (MOEMS) of mirror arrays. The main concept is to use an array of tiny mirrors that have the ability to tilt in different directions. Each mirror acts as an "eye" which images a scene. The individual images from each mirror are then reassembled, such that all of the information is placed into a single image. By exact control of the mirrors, the object plane can be sampled in a desired fashion, such that post-processing effects, such as image distortion and digital zoom, that are currently performed in software can now be performed in real time in hardware as the image gets captured. It is important to note that even for different sampling or imaging functions, no hardware components or settings are changed in the system. In this work, we present our programmable imaging system prototype. The MOEMS chipset used in our prototype is the Lucent LambdaRouter mirror array. This device contains 256 individually-controlled micro-mirrors, which can be tilted on both the x and y axes ±8o. We describe the theoretical model of our system, including a system model, capacity model, and diffraction results. We experimentally prototype our programmable imaging system using both a single mirror, followed by multiple mirrors. With the single mirror imaging, we explore examples related to single projection systems and give details of our required mirror calibration. Using this technique, we show mosaic images, as well as images in which a single pixel was extracted for every mirror tilt. Using this single pixel approach, the greatest capabilities of our programmable imaging are realized. When using multiple mirrors to image an object, new features of our system are demonstrated. In this case, the object plane can be viewed from different perspectives. From these multi-perspective images, virtual 3-D images can be created. In addition, stereo depth estimation can be performed to calculate the distance between the object and the image plane. This depth measurement is significant, as the depth information is taken with only one image from only one camera.

A novel approach to programmable imaging using MOEMS

Files and links (1)

Abstract

Metrics

Details

A novel approach to programmable imaging using MOEMS

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media