Estimation of cloud motion from satellite pictures

Estimation of lateral and rotational cloud displacement from satellite images is introduced based on various degrees of image abstraction. By using the luminance information, the maximum of the inger product surface of two images serves as a lateral displacement estimator. The peak detectability can be improved by prefiltering with a non-causal inverse filter. If the luminance is fit to a bivariate normal distribution, cloud motion and rotation is estimated from tracing the ellipses of equal density. Finally, the contour of clouds is Fourier transformed. The resulting coefficients reveal information about lateral and rotational displacement as well as size changes.