Sparse representations based clutter removal in GPR images

In GPR system, the reflected signal is composed of three components; clutter, target signal and system noise. As system noise has less importance compared to the other components, clutter reduction methods aim to decompose the reflected signal as target signal and clutter. In this paper, target signal and clutter are modeled sparsely with appropriate dictionaries via morphological component analysis. Resulting sparse coefficients and corresponding dictionaries are used to reconstruct clutter and target components. The proposed method is applied to experimental B-scan data and it is shown that the results have higher performance compared to the widely used Singular Value Decomposition (SVD), Principal Component Analysis (PCA) and Independent Component Analysis (ICA) based clutter reduction methods.