A novel demodulation method of the channeled modulated polarization imaging pictures by hybrid feature modulated autoencoders

Abstract

Channeled modulated polarization imaging technology offers advantages owing to its simple structure and low cost. However, the loss of high-frequency information due to channel crosstalk and the filter demodulation method has consistently hindered the mature application of this technology. We analyzed the data structure of pictures detected using this technology and proposed a demodulation method using hybrid feature modulated autoencoders. Training the network with a substantial number of images, it effectively addresses the issue of high-frequency information loss and demonstrates proficient demodulation capabilities for both simulated and real detected pictures.