Search

Numerous high-computational-complexity algorithms have been proposed for random noise removal in ground-penetrating radar (GPR) images, which are not suitable for real-time applications. Further, most of them ignore the non-stationarity of random noise in GPR images, resulting in a limited scope for their applications. To address this issue, a self-guided filter is proposed in this study for removing non-stationary random noise in GPR images. The proposed method consists of two steps: prefiltering and final filtering. In prefiltering, the original noisy GPR image is used as the guidance and input image for guided filtering, which employs the proposed time-varying regularization parameter for obtaining the prefiltering result. In final filtering, the prefiltering result and the original noisy GPR image are used as the guidance and input image, respectively, for the proposed guided filtering combined with edge information. The edge information is obtained by using the Sobel edge detection operator to extract the prefiltering result. The experimental results demonstrate that the proposed method effectively removes the non-stationary random noise in GPR images, and it is suitable for real-time applications. • A method for obtaining time-varying regularization parameters is proposed. • Improved guided filtering combined with edge information is proposed. • A real-time method is proposed to remove non-stationary random noise in GPR images. • Simulated and real results validate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]