Detection of vortices and saddle points in SST data

We extend the Horn-Schunck model of flow field computation to incorporate incompressibility for tracking fluid motion. This is expressed as a weak form of zero-divergence constraint in the variational problem and implemented with a multigrid approach for efficient computation. The resulting feature displacement velocity field provides the basis for higher level abstraction and representation of the data for data mining. A robust and efficient algorithm, based on the Jordan curve index, for detecting vortices and saddle points in feature displacement fields derived from sequences of satellite-derived SST fields is presented.