Abstract: This paper proposed a synchronous processing method using random noise suppression and high-frequency compensation for GPR data in curvelet domain. First， the GPR data were transformed into curvelet domain in order to use its multi-angle and multi-scale sparsity. The random noise thus could be suppressed by the adaptive threshold function changed with angle and scale. Second， according to the propagation laws of electromagnetic waves in elastic medium and the curvelet multi-scale feature， the time-varying compensation factor could be calculated so as to inverse weighting the corresponding scale and angle for compensation of the high-frequency absorption. Finally， inverse curvelet transformation was conducted thus the GPR data after random noise suppression and high-frequency compensation could be obtained. This method is totally a data-driven method thus can overcome the influence from artificial factors in traditional method.

Key words: GPR data, curvelet transform, random noise, high-frequency attenuation, de-noising method, absorption-compensation