Locally-adapted convolution-based super-resolution of irregularly-sampled ocean remote sensing data

摘要

Super-resolution is a classical problem in image processing, with numerousapplications to remote sensing image enhancement. Here, we address thesuper-resolution of irregularly-sampled remote sensing images. Using an optimalinterpolation as the low-resolution reconstruction, we explore locally-adaptedmultimodal convolutional models and investigate different dictionary-baseddecompositions, namely based on principal component analysis (PCA), sparsepriors and non-negativity constraints. We consider an application to thereconstruction of sea surface height (SSH) fields from two information sources,along-track altimeter data and sea surface temperature (SST) data. The reportedexperiments demonstrate the relevance of the proposed model, especiallylocally-adapted parametrizations with non-negativity constraints, to outperformoptimally-interpolated reconstructions.