Backscattering change detection in SAR images using wavelet techniques

Abstract

Backscattering change detection in synthetic aperture radar (SAR) is a challenging task, due to the presence of speckle noise. A 3-dimensional (2 spatial and 1 time dimension) wavelet shrinkage algorithm is developed to remove the noise. Compared to the often used 2-dimensional SAR wavelet shrinkage algorithm, this 3-dimensional algorithm exploits the correlation between SAR images, and therefore, a better signal to noise ratio is obtained. Furthermore, an autoregressive (AR) model is applied to analyze the temporal changes of radar backscattering. The change detection can be performed according to the AR model coefficients. Radarsat and ERS-2 SAR images are used to test this algorithm. The results show a significant improvement in the backscattering change detection.