Sparsity-seeking fusion of digital elevation models

Nowadays, different Digital Elevation Models (DEMs) are typically available for the same geographic area, which are provided by different sensors and/or processing techniques. Each DEM contains measurement errors due to the acquisition technology, the atmospheric conditions, the processing chain and the characteristics of the observed terrain. DEM fusion aims at overcoming the limitations of individual DEMs by merging them in an intelligent way that exploits the available complementary information and discards the redundancies. In this paper we present a generic algorithmic approach for fusing two arbitrary DEMs of the same geographical area, using the framework of sparse representations. A signal is called sparse when most or all of its information can be captured by a linear combination of a few elementary signals, also known as atoms. The atoms are chosen from a dictionary, i.e. an over-complete basis set, from which only a small subset is required to effectively represent each actual signal. We provide experimental results on real elevation datasets from different earth observation satellites to validate the proposed approach. Our evaluation shows that the proposed algorithm, along with carefully chosen fusion weights, yields consistently better DEMs.