physica status solidi (RRL) - Rapid Research Letters

Multispectral imaging plays an essential role in applications from biomedical imaging to remote sensing since it records rich spectral and spatial information. Nevertheless, conventional multispectral imaging systems generally suffer from complicated optical systems and compromise between spectral, spatial, and temporal resolution. Multispectral scattering imaging has recently drawn considerable attention due to its simple setup and unique methodology to solve the aforementioned bottlenecks. Herein, multispectral scattering imaging based on metasurface diffuser and deep learning is proposed. Prior knowledge of the metasurface diffuser's wavefront will avoid the time‐consuming characterization required in conventional scattering media, and a well‐trained deep learning algorithm can realize real‐time image reconstruction. With these merits, this method may push the development of metasurface applications and multispectral imaging.