Ghost imaging under low-rank constraint.

Abstract

Rather than the commonly used sparsity constraint, a new assumption taking advantage of regularity between rows or columns of a two-dimensional image is introduced to ghost imaging using compressive sensing, namely, low-rank constraint. Both simulation and experiment suggest explicit improvement on image quality of ghost imaging under low-rank constraint over ghost imaging under sparsity constraint (GISC), especially for under-sampling cases, and being in particular advantageous on image smoothness assessed by equivalent numbers of looks. Robustness of low-rank parameter setting is demonstrated. Low-rank constraint is also shown to be a powerful reference-less image quality enhancement tool for images restored by GISC.