Coded Exposure

Coded exposure techniques aim to capture high-dynamic-range (HDR) images and high-speed videos using computationally efficient methods that encode multiple exposures or frames into a single sensor reading. Current research focuses on improving the reconstruction of sharp images and videos from these coded exposures, often employing neural networks like NeRFs and Swin Transformers, along with novel algorithms for handling motion blur and noise. These advancements are significant for creating compact, low-power imaging systems for various applications, including virtual reality, microscopy, and autonomous driving, by reducing the need for high-bandwidth sensors and complex processing.