Non-local Filtering of TanDEM-X Interferograms for Improving DEM Quality

Kurzfassung

TanDEM-X (TerraSAR-X add-on for Digital Elevation Measurements) is an advanced space interferometer, composed of two TerraSAR-X satellites ying in close formation. The primary goal of TanDEM-X mission is the generation of a global Digital Elevation Model (DEM) of an unprecedented accuracy, coverage and quality. One of the main issues facing Interferometric SAR (InSAR) products is the interferometric phase noise coming from diffrent sources. The typical post-acquisition work-around for the noise problem is altering of interferograms. A plethora of alters have been suggested for the noise reduction and phase recovery. These alters are based on different principles and attained different degrees of success. One of the recently suggested alters seems to have a good potential in tackling the phase noise problem, the Non-Local InSAR (NL-InSAR) alter. This smart alter surpasses the performance of conventional alters by intelligently selecting the pixels to be averaged in order to reduce the noise and preserve structures. In this thesis, an implementation of the NL-InSAR alter is realized, with special attention paid to the heavy computational cost synonymous with nonlocal means concept. Various algorithmic and programming techniques are pursued to reduce the computational burden and bring the alter to an operational state suitable for large InSAR data-sets. To bring the NL-InSAR alter to a new level of performance, a simple improvement to the alter algorithm has been suggested. The performance of the modifed NL-InSAR alter is compared against the original NL-InSAR alter and to other conventional alters, realistic simulations have been done and the results are presented. The last part of the thesis is about evaluation of the NL-InSAR alter as a suggested alternative to the current Boxcar alter used in the standard TanDEM-X DEM generation. A systematic evaluation of NL-InSAR DEMs was done using few TanDEM-X data sets, and compared against the standard TanDEM-X DEMs.