Super Resolution Mapping with Multiple Point Geostatistics

摘要: Super-resolution is the process of providing fine scale land cover maps from coarse-scale satellite sensor information. Such a procedure calls for prior model depicting spatial structures types. When available, an analog underlying scene (a training image) may be used such model. The snesim (single normal equation simulation) algorithm allows extracting relevant pattern information image and uses that to downscale coarse fraction data into simulated scene. First, observed downscaled using block indicator Kriging (BIK), with variograms computed image. resulting fractions at any given location are interpreted as probability having specific location. That then merged lifted latter made conditional available previously class along simulation path. Land types drawn posterior distribution. A servo-system keeps track number classes inside each assures exact reproduction data. By repeating new path visiting grid, one can generate several super resolution explore space uncertainty cover. proposed mapping i) exactly reproduce fraction, ii) inject structural carried by image, iii) condition ground observations. case study provided illustrate methodology Landsat TM SE China.