SEMINAIRE DU 4 JUIN 2015 – 16H @ MAP5 – SALLE DU CONSEIL
Joint phase denoising and unwrapping for 3D reconstruction in
multi-channel SAR interferometry: combining patch-based non-local
filtering and total variation regularization
SAR (Synthetic Aperture Radar) systems are powerful imaging systems allowing height retrieval thanks to interferometric phase measurements (InSAR). While a single interferometric phase implies a phase unwrapping step, multi-channel InSAR systems reduce ambiguities for height retrieval but still provide very noisy measurements.
In this work we investigate how the combination of patch-based approaches and total variation minimization (TV) improve 3D InSAR reconstruction.
After a brief introduction to SAR imagery and InSAR acquisition, the adaptation of non-local methods to complex vectorial data will be presented. Then a non local likelihood term is introduced in the TV regularization framework, and optimized thanks to a graph-cut minimization approach. Results on synthetic and real InSAR data will be discussed.
This work is a joint work with C. Deledalle (IMB), L. Denis (Laboratoire Hubert Curien), and G. Ferraioli (Parthenope University, Naples, Italy).