Fig. 1. Denoised results for the synthetic ground penetrating radar (GPR) data with random and coherent noise using traditional curvelet transform: (a) Original GPR data; (b) data with random noise, PSNR = 17 dB; (c) data with coherent noise, PSNR = 15.51 dB; (d) the δ value vs. PSNR value curves 含噪声模拟合成探地雷达数据的曲波变换阈值去噪结果　(a)原始合成数据; (b)含随机噪声数据, PSNR = 17 dB; (c)含相关噪声数据, PSNR = 15.51 dB; (d)去噪数据PSNR值随阈值控制系数δ的变化

Fig. 2. Denoised results for the synthetic GPR data (Fig. 1. (b) and Fig. 1. (c)) with δ = 0.08 using traditional curvelet transform: (a) Result for random noise, PSNR = 20.23 dB; (b) result for coherent noise, PSNR = 16.75 dB. δ = 0.08时含噪声数据(图1(b)和图1(c))曲波变换去噪结果　(a)随机噪声数据去噪结果, PSNR = 20.23 dB, 提高3.23 dB; (b)相关噪声数据去噪结果, PSNR = 16.75 dB, 提高1.24 dB

Fig. 3. Denoised results for the synthetic GPR data (Fig. 1. (b) and Fig. 1. (c)) using curvelet transform with statistical self-adaption: (a) Result for random noise, PSNR = 25.3 dB; (b) result for coherent noise, PSNR = 21.92 dB. 含噪声数据(图1(b)和图1(c))的统计量自适应阈值曲波变换去噪结果　(a)随机噪声数据去噪结果, PSNR = 25.3 dB, 提高8.3 dB; (b)相关噪声数据去噪结果, PSNR = 21.92 dB, 提高6.41 dB

Fig. 4. Denoised results for the synthetic GPR data of complex model with random and coherent noise using curvelet transform with statistical self-adaption: (a) Original GPR data; (b) data with random noise, PSNR = 16.04 dB; (c) data with coherent noise, PSNR = 15.7 dB; (d) result for random noise, PSNR = 23.97 dB; (e) result for coherent noise, PSNR = 21.05 dB. 双矩形目标模型含噪声数据统计量自适应阈值曲波变换去噪结果　(a)原始合成数据; (b)含随机噪声数据, PSNR = 16.04 dB; (c)含相关噪声数据, PSNR = 15.7 dB; (d)随机噪声数据去噪结果, PSNR = 23.97 dB; (e)相关噪声数据去噪结果, PSNR = 21.05 dB

Fig. 5. Denoised results for the synthetic GPR data (Fig. 4.) in the 50th receiver using curvelet transform with statistical self-adaption: (a) Result for random noise; (b) result for coherent noise. 双矩形目标模型含噪声数据第50道统计量自适应阈值曲波变换去噪结果　(a)随机噪声数据去噪结果; (b)相关噪声数据去噪结果

Fig. 6. Denoised results for field GPR data using curvelet transform with L2 standard deviation and statistical self-adaption respectively: (a) Original field GPR data; (b) result using curvelet transform with L2 standard deviation; (c) result using curvelet transform with statistical self-adaption.实测探地雷达时间剖面传统曲波变换去噪和本文去噪算法处理结果　(a)实测时间剖面; (b)采用L2标准方差估计阈值曲波去噪结果; (c)统计量自适应阈值曲波去噪结果

Fig. 7. Denoised results for field GPR data (Fig. 6) in the 200th receiver using curvelet transform with L2 standard deviation and statistical self-adaption respectively. 实测探地雷达时间剖面第200道传统曲波变换去噪和本文去噪算法处理结果