Globally optimal robust DSM fusion.

This work presents the mathematical formulation of a novel globally optimal robust digital surface model (DSM) fusion method, that can be used to combine several 2.5D DSMs extracted from airborne or spaceborne stereo images. The main novelty is the definition of a convex energy functional with a β-Lipschitz continuous gradient that allows a trivial solution of the posed minimization problem, where the robustness is achieved by incorporating the Huber norm into the energy functional. All according mathematical proofs are derived within this work. The experiments are based on two different minimization schemes and are applied on airborne optical, on spaceborne optical and on spaceborne synthetic aperture radar (SAR) images. The resulting fused 2.5D DSMs are rich in detail and are of higher quality than results of other local fusion methods.