Shape and area correlation in time-varying imagery

In this work we present a global scheme for motion detection and characterization in radar images. It involves 2 steps; (1) each image is rapidly scanned using shape extraction and recognition to extract important clues regarding areas of major changes; (2) using the latter clues, a local analysis based on correlation then is applied to selected areas of the image. This technique was applied to a sequence of real radar cloud images where circulation was suspected; the results showed clear automatic detection of this circulation. The algorithms used in this study and their formulation are fast enough to envision real time tracking of clouds.