A new method for image down-scaling using geostatistical interpolation or smoothing based on the Hopfi eld Neural Network (HNN) and zero semivariance value is introduced. The method utilises the smoothing effect of the semivariogram matching process to produce the smoothened sub-pixel multispectral (MS) image with smaller RMSEs in comparison with the bilinear interpolation. In fact, the zero semivariograms increase the spatial correlation between the adjacent sub-pixels of the superresolution image. Containing higher spatial correlation, the resulting super-resolution MS image has smaller RMSEs compared with the original coarse image.

A new method for image down-scaling using geostatistical interpolation or smoothing based on the Hopfi eld Neural Network (HNN) and zero semivariance value is introduced. The method utilises the smoothing effect of the semivariogram matching process to produce the smoothened sub-pixel multispectral (MS) image with smaller RMSEs in comparison with the bilinear interpolation. In fact, the zero semivariograms increase the spatial correlation between the adjacent sub-pixels of the superresolution image. Containing higher spatial correlation, the resulting super-resolution MS image has smaller RMSEs compared with the original coarse image.