Automatic segmentation of oceanic eddies on AVHRR thermal infrared sea surface images

Presents some new developments of a technique for automatic detection of oceanic eddies on satellites infrared sea surface images. This technique relies on an algorithm using multiscale analysis of isotherm curvature and characterization of eddies through the phase portraits method. The main interest of the method is that isotherm curvature can be related to ocean dynamic, particularly to detect and describe such phenomena as mesoscale eddies with spiral patterns. The aim of this work is to obtain a much efficient method both in term of calculation time and stability in the characterization step. This algorithm has been then tested on several thermal infrared images of Bay of Biscay issued from AVHRR radiometer. The results on some typical images are given and statistics on the whole set are briefly described and discussed.