Infrared remote sensing imaging via asymmetric compressed sensing

Abstract

Compressed sensing (CS) theory provides a new acquisition idea for sparse signals and sparsely-expressed signals. CS-based hardware design has been widely concerned. And related products have been tentatively tried in many fields. The design of remote sensing imaging based on CS mainly includes single pixel multiple exposure imaging and block focal plane coding multi — pixel single exposure imaging. In this paper, a CS asymmetric processing model, which is different from traditional image reconstruction, is proposed. And it is applied to CS hardware design for infrared (IR) remote sensing imaging. This model fully considers the global information of the image, which combines the multiple neighborhood values of the observed results in the CS process, and also combines the multiple measurement matrix blocks to form a new measurement matrix. At the same time, a sparse dictionary construction method suitable for asymmetric patterns is proposed, which can effectively compensate for the local differences caused by image segmentation. The experimental results show that the proposed method is superior to the conventional block focal plane coding compression reconstruction both in reconstruction time and in reconstruction quality.