부산대학교 도서관

The interpretion of GPR data is usually limited by the attenuated signal and noise interference. The signal-to-noise ratio (SNR) of the received GPR data is influenced by the characteristics of GPR systems, the target signatures, the properties of the medium, and by incoherent random noise. The original attributes of the data will be distorted by noise. Usually, noise removal is conducted by employing some regular Fourier-based linear filtering algorithms. Those algorithms not only introduce mixed-mode problems or spurious harmonics but also over smooth the collected GPR data. This paper proposes an approach based on the non-local idea to depress the noise effects shown in GPR data. The non-local approach means that any point can be directly interacted with any other aspect of the whole image, where the weight function can be built according to the similarity of the two points. In doing so, the noise effects can be depressed with keeping the details of images. The non-local means (NL-means) algorithm is implemented to suppress noise effects in the given GPR images. The processed results will be compared with the processed results by employing different noise suppression algorithms. In this paper, the signal-noise ratio (SNR) is used to evaluate the performance of different algorithms.