In order to improve the denoising effect of the current time frequency (TF) method on ground penetrating radar signals, the authors accomplished the denoising in the TF domain by using the spectral decomposition reassignment algorithm. The application of a window function in traditional TF transforms to dividing the signal causes a blurred TF distribution, i.e., TF energies of the signal are distributed around the true instantaneous frequencies, which makes it easier to damage useful information of the signal during denosing if the threshold is inappropriate. The spectral decomposition reassignment algorithm is capable of making energy of the signal more concentrated around the true instantaneous frequencies by relocating TF result of the signal, thus obtaining higher TF resolution and reducing the opportunity of damaging useful information of the signal during denosing. Processing results of both synthetic and real data suggest that the denoising effect of the reassignment method is better than that of traditional TF including Gabor transform methods.