Pre-processing techniques an features extraction for ocean meso-scale structures detection in SST Images

This paper addresses the issue of pre-processing techniques and features extraction for meso-scale structures detection in Sea Surface Temperature (SST) Images. Ocean meso-scale structures are areas in the ocean where the gradient of temperature is high. At the first approximation, the temperature evolution can be modelled as the sum of a seasonal component and perturbations. The perturbation term caters for the effect of the meso-scale structure and the noise. The Empirical Mode Decomposition (EMD) has been recently developed by N.E. Huang et al EMD. By looking at the local trends in the signal, this approach decomposes the information in intrinsic mode functions (IMF) and is of particular interest for extracting harmonic components. The paper presents the fundamentals of the EMD in one dimension and two dimensions (IEMD). An introduction to the EMD in three dimensions (IEMD-CM) is also presented. Simulations are run on a time-spatial series of images obtained from METEOSAT second de generation satellite of the oceans surrounding Southern Africa.