Deep SBP+: breaking through the space-bandwidth product limit based on a physical-driven cycle constraint framework

Abstract

To obtain an image with both high spatial resolution and a large field of view (FoV), we designed a deep space-bandwidth product (SBP)-expanded framework (Deep SBP+). Combining a single-captured low-spatial-resolution image with a large FoV and a few captured high-spatial-resolution images in sub-FoVs, an image with both high spatial resolution and a large FoV can be reconstructed via Deep SBP+. The physical model-driven Deep SBP+ reconstructs the convolution kernel as well as up-samples the low-spatial resolution image in a large FoV without relying on any external datasets. Compared to conventional methods relying on spatial and spectral scanning with complicated operations and systems, the proposed Deep SBP+ can reconstruct high-spatial-resolution and large-FoV images with much simpler operations and systems as well as faster speed. Since the designed Deep SBP+ breaks through the trade-off of high spatial resolution and large FoV, it is a promising tool for photography and microscopy.