Goal of this project is the study and development of a computational method for speckle reduction of SAR imagery. The main novelty of the proposed approach is performing Bayesian estimation with PDF models of undecimated wavelet transform coefficients of reflectivity and of signal-dependent noise that may vary within each wavelet plane, according to a preliminary partition of the latter into statistically homogeneous segments. Different wavelet filters, different Bayesian estimators and different PDF models, as well as non-separable multiresolution analysis, like the contourlet transform, shall be considered in the cost-performance trade-off. The filtering procedure shall be optimized on COSMO-Skymed data by devising objective quality measurements of denoised images.
|Development of advanced segmentation-based multiresolution methods for speckle reduction and texture restoration in high-resolution SAR imagery|
|Tipology||ASI “COSMO-SkyMed Scientific Projects”: Contract # I/043/09/0 – “COSMO-SkyMed: Announcement of Opportunity”, ID-2293|
|Duration||Feb. 2010 - Feb. 2012|
|Budget||121,011.58 € (49.188,78 €)|
|Research area||Multimedia & Digital Signal Processing|
|Unit Coordinator||Università degli Studi di Firenze – Luciano Alparone|