Chapter 13 - Evaluation of speckle filtering methods using polarimetric Sentinel-1A data

Abstract

All polarimetric synthetic aperture radar images are inherently affected by speckle noise. It degrades the quality of images and is an obstacle to further image analysis and terrain classification. This work examined the effectiveness of various filtering methods dedicated to speckle suppression in dual-polarization Sentinel 1A synthetic aperture radar images. The speckle suppression outcomes of VV and VH polarization are presented using six built-in filters of Sentinels Application Platform software: Boxcar, Median, Frost, Gamma map, Lee, and Refined Lee. Several indices, such as the standard deviation to mean ratio (SD/M), relative standard deviation (RSD), speckle suppression index (SSI), and equivalent number of looks (ENL), were used to evaluate the speckle suppression abilities of filters. A median filter with a window size of 11×11 gave the best performance for VV and VH polarization images. This gave lower values of SD/M, RSD, and SSI and a higher value of ENL while revealing the best performance specifically for an image at VH polarization among all of the filters.