Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
In this internship, I explore different optimization algorithms for lensless imaging. Lensless imaging is a new imaging technique that replaces the lens of a camera with a diffuser mask. This allows for simpler and cheaper camera hardware. However, the reconstruction of the image from the sensor measurement is a complex ill-posed problem. The goal of this internship is to improve current reconstruction algorithms for lensless imaging. I first implemented unrolled optimization as a low computational cost way of increasing reconstruction quality. I also explored the possibility of using a plug-and-play algorithm for lensless imaging. Then, I introduced a pre-/post- denoising scheme mixing unrolled optimization and more traditional deep learning which improved current state-of-the-art result on the DiffuserCam Lensless Mirflickr Dataset. Finally, I introduce a framework for learning the mask pattern of the lensless camera jointly with the reconstruction algorithm, allowing for further improvements.
François Maréchal, Julia Granacher
Jürg Alexander Schiffmann, Soheyl Massoudi, Cyril Picard