Jueves 22 de diciembre 9:30hs, Salón Marrón (705) – Facultad de Ingeniería, J. Herrera y Reissig 565
Tenemos el agrado de invitarlos a la defensa de la tesis de doctorado de Javier Preciozzi titulada “Two Restoration Problems in Satellite Imaging”
Directores de Tesis : Andrés Almansa y Pablo Muse
Tribunal : Mauricio Delbracio, Fernando Paganini, Pablo Sprechmann, Andrés Almansa, Pablo Musé, François Malgouyres y François Champagnat
This thesis is about satellite image restoration using a variational approach. In the first part, SMOS image restoration is analyzed. These images have the particularity of having being obtained indirectly, from the cross-correlation of microwave signals. Although the sensed wavelength band is exclusively designed for scientific purposes, and all emissions in this band are forbidden, there were (and there still are) strong radio frequencies interferences that degrade enormously the acquired data. Without a good restoration process, all these data remain useless.
In the second part, the problem is related to optical images from very high resolution satellites. Because of the great amount of data, these images have to be compressed prior transmission to Earth. This compression is based on a wavelet transform, and most of its compression power comes from the truncation of the obtained coefficients. The problem with this process is that, because of the noise that is always present on any acquisition system, some artifacts may appear on the decompressed image. Although these artifacts are not very significant to the naked eye, they generate enormous problems for post-processing (for instance, on the computation of Digital Elevation Maps).
In this thesis, we propose solutions for both problems under the same theoretical framework, developing algorithms based on variational formulations. Although each of the problems has its particularities (basically because of the image formation model), the way in which the solutions are proposed is quite general, and can be easily extended to other contexts.