Compressive spectral image reconstruction using deep prior and low-rank tensor representation

Compressive spectral imaging (CSI) has emerged as an alternative image acquisition technology, which reduces the number of measurements at cost requiring a recovery process. In general, reconstruction methods are based on hand-crafted priors used regularizers in optimization algorithms or recent deep neural networks employed generator to learn non-linear mapping from low-dimensional compressed space. However, these data-driven need many images obtain good performance. this work, framework for CSI without training data is presented. The proposed method fact that structure some and appropriated sufficient impose underlying CSI. We analyzed low-dimension via Tucker representation, modeled first net layer. scheme obtained by minimizing $\ell_2$-norm distance between compressive predicted measurements, desired recovered formed just before forward operator. Simulated experimental results verify effectiveness method.