Cloud Motion Estimation in Satellite Image Sequences by Tracking Skeleton Critical Points Using Lucas-Kanade Method

In this work we propose a new approach to estimate motion of clouds from satellite images using tools from digital geometry in combination with tools from computer vision field. The idea is to represent clouds by their binary skeleton and then to track the critical points of the pruned skeleton using optical flow estimation approach, particularly using the Lucas-Kanade method which generates a sparse motion field of a set of feature points. The critical points of the skeleton can easily tell which pixel is suitable to be tracked by the optical flow algorithm. Our method is motivated by the fact that critical points of a binary skeleton can carry information about the global structure of an object. In the specific context of meteorological imagery, the information about clouds shape and their topology is embedded in the critical points of the skeleton which makes them more suitable for determining the good features to track by an optical flow algorithm.