We aim to improve the imaging quality of metasurface-based image sensors by using a stack of freeform metasurfaces made of silicon nitride nanopillars and computational post-processing. Multiple metasurfaces will be used to engineer the incident wavefront, which will be further processed using computational algorithms. We will primarily focus on using this computational imaging approach to capture high quality images for human perception that have high resolution, high signal-to-noise ratio, and are free of chromatic and geometric aberrations. We will use a composite metasurface optical system that creates an extended depth of focus, which allows capturing the information from the scene in full color in addition to correcting for geometric aberrations. We will build an end-to-end image processing model that allows us to optimize both the metasurface optical system itself, in addition to the parameters used for the computational reconstruction algorithm.
