Physics-Informed Imaging: Lensless Cameras & Low-Light Videography

This lecture by the instructor covers the use of physics-informed machine learning for lensless imaging and low-light videography. It explores the challenges of imaging inverse problems, the classic and deep methods for solving them, and the application of physics-based noise GANs. The lecture delves into the importance of noise models in capturing high-quality images in low-light conditions, showcasing the development of a noise model for submillilux videography. It also discusses the training of denoisers using synthetic noisy videos and the simulation of supervised denoising. The presentation concludes with the demonstration of photorealistic videography in submillilux and the creation of an open-source dataset for further research.