CHPDNet: a chromatic high-resolution polarization decoding network for snapshot polarization imaging

To address the issue of spatial resolution loss in existing polarization imaging systems, this paper proposes a snapshot chromatic high-resolution polarization imaging method based on the deep learning technology. Using the images captured by pupil-divided polarization imaging scheme as the input, chromatic high-resolution polarized images can be simultaneously retrieved by the proposed chromatic high-resolution polarization decoding network (CHPDNet). Experimental results show that the intensity ( S 0 ) image was reconstructed with high precision, exhibiting a relative error below 1.2%. Meanwhile, the degree of polarization (DoP) maps were also accurately recovered, with errors below 12% for the G and B channels and below 16% for the R channel, along with structurally complete AoP images. The proposed method may offer a new approach to the design of chromatic simultaneous polarization imaging methods.