Abstract

We proposed a super-resolution human-silhouette imaging system using end-to-end optimization of coded illumination and a reconstruction network, increasing resolution without altering camera hardware. We experimentally verified the performance and analyzed it using pseudo-color-based visualization.