Remote sensor imagery information contents and their scale effects

Information theoretic perspectives provide important metrics for quantitative remote sensing. Scale is known to be an essential factor in remote sensing image information contents. In this paper, the relationships between information content and scale are analyzed. First, methods for image information measurement are introduced. Then, three re-sampling methods for scaling-up, including the nearest interpolation, bilinear interpolation, and cubic interpolation, are described. Effects of inter-pixel correlations upon information contents are analyzed. Experimental results show that the scale at which the maximum average information contents between adjacent pixels are registered does not accord with the highest resolution. The results provide useful insights for choosing suitable data scales or designing sensors with proper resolution according to the underlying process scales or specific application.