Clutter and Random Noise Elimination Based on Eigen Images and Curvelet Transform

Abstract

A GPR radargram of a buried object shows reflections from the target as well as from several unintentional objects or clutters. Furthermore, the signal is corrupted by the background noise, ground bounce and the direct waves of the antennas. The requirements to effectively extract the target signature are clutter must be eliminated and there should be no extra noise effects. Although the clutters in the data cannot be completely eliminated, background removal techniques significantly reduce their impact. Typically, background can be removed using mean subtraction, but the results are just marginally adequate. An eigen image based background removal is used in this paper. The noise effects are reduced using curvelet transform (CT) based denoising. The output significantly reduces noise and clutter, enhancing the efficiency of the detection and classification stages of a GPR system