High-spectral resolution imaging systems provide critical information enabling a better identification and characterization of the objects in a scene of interest. Nevertheless, multiple factors may impair spectral resolution, as in the case of modern snapshot spectral imagers that associate each “hyperpixel” with a specific spectral band.
In this work, we propose a novel post-acquisition computational technique aiming to enhance the spectral dimensionality of imaging systems by exploiting the mathematical frameworks of Sparse Representations and Dictionary Learning. We formulate our spectral coupled dictionary learning technique within the context of the Alternating Direction Method of Multipliers, optimizing each variable via closed-form expressions.
The particular algorithmic framework can be considered in a wide range of remote sensing applications for Earth Observation. For instance, acquired imagery from low spectral resolution satellites, e.g., MODIS, could be enhanced using images acquired over the same region from higher resolution spectrometers aboard newer platforms, e.g., the EO-1 Hyperion, as shown in the figure.
SCDL System Block Diagram: The system takes as input a hypercube acquired with a limited number of spectral bands and produces an estimate of an extended spatio-spectral hypercube. During the training phase, multiple high and low-spectral resolution “hyper-pixels” are extracted from training hypercubes. Given these “hyper-pixels” pairs, a coupled sparse dictionary learning scheme is employed for learning two sparsifying dictionaries, corresponding to the two resolution cases. During runtime, low resolution “hyper-pixels” are mapped to the low resolution dictionary and the identified sparse coding coefficients are subsequently combined with the high resolution dictionary for producing the final estimates. An indicative example of reconstruction for the proposed and two state-of-the-art methods is shown below.
Konstantina Fotiadou, Grigorios Tsagkatakis, and Panagiotis Tsakalides, “Spectral Super-Resolution via Coupled Sparse Dictionary Learning”, submitted to IEEE Transactions on Computational Imaging, 2016.