A statistical model for describing the texture of cloud cover images from satellite data

Abstract

A statistical model is proposed which describes the texture of the images of 25 types of clouds from the MODIS satellite data with the spatial resolution of 250 m. A technique is presented which compiles the sets of image segments with typical textures for different cloud types. To describe the texture, the following statistical methods are applied: Gray-Level Co-occurrences Matrix, Gray-Level Difference Vector, and Sum and Difference Histogram. The correlation analysis is used to form the sets of informative textural features for the images of different cloud types. Two-parameter distribution laws and the estimates of their parameters are presented which describe fluctuations in the efficient set of textural features of different cloud types.