When using remote sensing for land cover mapping

super-resolution reconstruction is widely used.Prior models containing the features of land cover maps can constrain the uncertainty of reconstruction.These prior models can be used properly by multiple-point statistics (MPS) by extracting the intrinsic features from them

and copying these features to the simulated regions.However

because traditional MPS methods based on linear dimensionality reduction are not suitable to deal with nonlinear data

isometric mapping (ISOMAP) is introduced in MPS to reduce the dimensionality reduction of nonlinear data and then these lower-dimensional data are classified.Current data event and the average of every classified class are compared so that a pattern can be extracted from the class that is closest to the current data event.Besides

the low-resolution original image is viewed as soft data for generating super-resolution land cover maps.Tests show that the super-resolution reconstructions of land cover maps have the similar structural features with those of reference images.