Super-resolution land cover mapping with indicator geostatistics

摘要: Abstract Many satellite images have a coarser spatial resolution than the extent of land cover patterns on ground, leading to mixed pixels whose composite spectral response consists responses from multiple classes. Spectral unmixing procedures only determine fractions such classes within coarse pixel without locating them in space. Super-resolution or sub-pixel mapping aims at providing fine map class labels, one that displays realistic structure (without artifact discontinuities) and reproduces fractions. In this paper, existing approaches for super-resolution are placed an inverse problem framework, geostatistical method is proposed generating alternative synthetic maps (target) resolution; these realizations consistent with all information available. More precisely, indicator coKriging used approximate probability belongs particular class, given (if available) sparse set labels some informed pixels. Such Kriging-derived probabilities sequential simulation generate This non-iterative fast procedure yields reproduce: (i) observed fractions, (ii) might be available, (iii) prior structural encapsulated variogram models resolution. A case study provided illustrate methodology using Landsat TM data SE China.