An Error-Bound-Regularized Sparse Coding for Spatiotemporal Reflectance Fusion

This paper attempts to demonstrate that addressing the dictionary perturbations and the individual representation of the coupled images can generally result in positive effects with respect to sparse-representation-based spatiotemporal reflectance fusion (SPTM). We propose to adapt the dictionary perturbations with an error-bound-regularized method and formulate the dictionary perturbations to be a sparse elastic net regression problem. Moreover, we also utilize semi-coupled dictionary learning (SCDL) to address the differences between the high-spatial-resolution and low-spatial-resolution images, and we propose the error-bound-regularized SCDL (EBSCDL) model by also imposing an error bound regularization. Two data sets of Landsat Enhanced Thematic Mapper Plus data and Moderate Resolution Imaging Spectroradiometer acquisitions were used to validate the proposed models. The spatial and temporal adaptive reflectance fusion model and the original SPTM were also implemented and compared. The experimental results consistently show the positive effect of the proposed methods for SPTM, with smaller differences in scatter plot distribution and higher peak-signal-to-noise ratio and structural similarity index measures.