Meteorological remote sensing images may contain a huge amount of information, but the matter of extracting and turning them into usable forms is generally quite complex. A wide range of operations is still performed by human experts, as active intelligence and human skills are required. Automatic procedures usually take charge of low- and middle-level tasks, relieving human experts from part of the work. An example of a partly automated task is the tracking of meteorological structures, which is usually performed using techniques based on local correlation between pixel intensity values. We test a technique that is based on the global shape of the observed phenomenon instead. As opposed to traditional techniques based on local correlation, this technique uses the whole shape of the observed object to infer local correspondences. Two modal shape descriptions have been used as the basis for this shape-based technique, and both descriptions are useful for analyzing, classifying, and tracking cloud structures in satellite images. Results using Meteosat and GOES datasets are proposed, discussing the advantages and disadvantages of shape analysis techniques in different meteorological applications. A comparison of interframe matching is provided between the methods discussed here and the traditional block-matching method.