Speckle noise reduction in PolSAR images with binary partition tree
Ferran Pérez Gamonal · UPCommons institutional repository (Universitat Politècnica de Catalunya) · 2017
In some remote sensing applications such as PolSAR (Polarimetric Synthetic Aperture Radar), the use of Binary Partition Trees (BPTs) for Speckle Noise filtering schemes is currently gaining interest. In this thesis, a new approach using this representation is investigated: branch filtering. This approach consists in searching for each leaf its ancestors and selecting the one that best represents it, that is, the one that yields the lower error. A potentiality assessment is done to evaluate the margin of improvement that new techniques based on this approach may provide and describe the basic specifications of the algorithms based on it. After that, different new techniques are developed, analysed and compared against the State-of-the-Art. We point out the main strengths and weaknesses of each technique. Our main goal is to understand the behaviour of the filtered data along the BPT branch and interpret how this information can be used in the future for speckle noise reduction in PolSAR images. Finally some conclusions are drawn and some possible future lines of work are exposed and commented.