The spatial resolution of hyperspectral imaging systems is constrained by a spatial-spectral resolution tradeoff and current technique limitations. However, spatial resolution is a critical feature for applications that require high spatial resolution and utilization of details. We present a method of restoring high-resolution (HR) images from a set of low-resolution (LR) hyperspectral data cubes with subpixel shifts across different bands. A new observation model is introduced to demonstrate LR hyperspectral images at different bands and an HR image that covers all these bands. A regularized super-resolution (SR) algorithm is then implemented to solve the problem. Experiments of the proposed algorithm and existing SR algorithms are performed and the results are evaluated. The results demonstrate the feasibility of the proposed SR method. Moreover, the image fusion results also demonstrate that the restored image is suitable for enhancing the spatial resolution of entire hyperspectral data cubes.
As the LR images from the plenoptic camera is greatly constrained by the number of micro-lenses, we can apply the multi-frame super resolution methods to enhance the spatial resolution. Multi-frame super resolution reconstruction is a technology which obtains a high resolution image from several low resolution images of the same scene. Among various super resolution methods, the regularized methods are widely used since they have advantages for solving the ill posed problems. In this paper, some regularized super resolution methods are applied to enhance the spatial resolution of the light field image. The reconstruction results of synthetic low resolution images confirm that all the regularized super resolution algorithm can suppress the Gaussian noise and preserve the edge information. The real data experiment results also confirm the effectiveness of the applied algorithms.