Mask and Algorithm Co-Design of a Lensless Camera in Multiple Reconstruction Settings
Mask and Algorithm Co-Design of a Lensless Camera in Multiple Reconstruction Settings
ESE 498 Capstone Design Project
Client Electrical & Systems Engineering Department
Advisors Dorothy Wang1dorothyw@wustl.edu Matthew Lew2mdlew@wustl.edu Jason Trobaugh3 jasont@wustl.edu
Engineers Jasmine Cheng4 jasmine@wustl.edu Mae Martel5 m.mae@wustl.edu
Navigation
Home: Contains the sponsors and team information, abstract, introduction, and deliverables.
Software: Describes and contains the methodology behind the algorithmic component of our project.
Hardware: Describes the hardware component of image generation and the reconstruction process.
Future Works: Delves into further hardware advancements to achieve optimal PSF sparsity, and enhanced image reconstruction algorithmic techniques.
Abstract
Under the advisory of Drs. Dorothy Wang, Matthew Lew, and Jason Trobaugh, this capstone project explores the hardware configuration responsible for encoding information and the algorithmic framework designed to decode information within a highly adaptable lensless imaging system. The hardware setup utilizes a camera sensor and LED light source from Thorlabs, along with a customized cardboard aperture. Images captured by the sensor, initially uninterpretable to humans, undergo processing and reconstruction via a deconvolution algorithm implemented in Python, utilizing optimization techniques like gradient descent (GD). By synergizing hardware and software components, the lensless imaging system achieves object reconstruction. The final objective of the project is to assess the impact of Point Spread Function (PSF) and object sparsity on the system’s reconstruction efficacy.
Introduction
In traditional optical imaging, bulky systems comprising multiple lenses project clear, detailed images onto sensors. However, these setups are impractical for confined spaces. Lensless imaging sidesteps this issue by using masks, like diffusers, to encode object data onto sensors directly. This approach enables diverse applications, from neural circuitry monitoring to robotics and 3D display content generation. In an educational context, lensless cameras offer a hands-on demonstration of imaging systems and signal processing, fitting well in undergraduate classrooms due to their portability, accessibility, and tolerance towards erasures. This capstone project aims to construct a benchtop lensless imaging system and investigate the influence of object and point spread function sparsity on reconstruction accuracy.
Explore our project deliverables: the ESE 498 Capstone Final Report, the poster presented during WashU's ESE Day on 26 APR 2024, and our GitHub repository.
Footnotes
Lecturer, Department of Electrical and Systems Engineering ↩︎
Associate Professor, Department of Electrical and Systems Engineering ↩︎
Professor in Practice, Department of Electrical and Systems Engineering ↩︎
