Speckle filtering of Polarimetric SAR data

Abstract

In the field of Remote Sensing the main device, used to obtain the surface images, are the so-called Synthetic Aperture Radar. This systems are devices able to catch high-resolution images, which keep peculiar informations about the observed surface. Through the use of a Radar, mounted on board of a spaceborne or airborne vehicle, large overflow areas are electromagnetically radiating. The electromagnetic answer of the illuminated surface under discussion, is then analyzed in such a way to extract the wished informations. This kind of image acquisition presents an intrinsic trouble generated by the set of electromagnetic waves, which are interacting each other, on the path from the target to the receiver system. The trouble is well-known as Speckle and it will be the main topic of this thesis project.Over the last 30 years, several algorithms able to significantly reduce the trouble effect have been implemented. However, trouble reduction, is done to the detriment of the preserved information. On this basis, an equal important research is to evaluate in detail, as more as possible, the speckle filtering performance and moreover which informations are preserved and which are degraded. For this reason, a comparison between the filtered images and the untroubled images may be useful, but as it has been said above, signal and trouble are inseparable, therefore an untroubled version of the acquired image is not achievable.To work around the problem, has been generated a synthetic image where the speckle contribute is absent, using some representative sample extracted from a real SAR image, that in this case is the well-known SAR image over the San Francisco Bay (CA). Thus, based on it, speckle contribute has been added on the image. Furthermore, to make the simulation more realistic, it has been added texture, which may represent high density forest or urban area, as well as target point, which may represent naval ships at open sea, or more generally, small dimension object anywhere.Subsequently, two types of parameters have been implemented for the evaluation of the information preserved. First, polarimetric preservation parameters, which express a measurement about intensity of speckle contribute for each channel, entropy, anisotropy, mean angle alpha and the polarimetric signature. Second, spatial preservation parameters, which measure edge preservation, target point preservation and the Equivalent Number of Look of the filtered image. Next, a collection of test images has been stored with Monte Carlo Method and several filters through the platform PolSARpro have been applied. Each sample image has been evaluate in term of the parameters above presented. Finally, each filter has been applied to the real image in such a way to have the opportunity to highlight and to compares the conclusion obtained about the parameters and their respective filtered image.