Weighted Total Variation Iterative Reconstruction for Hyperspectral Pushbroom Compressive Imaging

Compressed sensing is suitable for remote hyperspectral imaging, as it can simplify the architecture of the onboard sensors. To reconstruct hyperspectral image from pushbroom compressive imaging, we present iterative prediction reconstruction architecture based on total variation in this paper. As the conventional total variation prior is not effective at capturing the correlation within spatial-spectral arrays, an improved weighted total variation is proposed. Experimental results run on raw data from AVIRIS confirm the validity of the proposed method.