Snapshot Compressive Imaging

Snapshot compressive imaging (SCI) aims to reconstruct high-dimensional data, such as videos or hyperspectral images, from a single, compressed 2D measurement, significantly improving acquisition speed and efficiency. Current research emphasizes developing efficient and accurate reconstruction algorithms, focusing on deep learning architectures like transformers, convolutional neural networks, and diffusion models, often within unfolding frameworks or incorporating novel techniques like quantization and hardware-aware training. These advancements are driving progress towards real-time SCI applications on mobile devices and improving the quality and speed of hyperspectral and video imaging in various fields.