Abstract: In nature field observation the impedance estimation in the frequency domain will take severely biases while like-impulse exists and makes signal violent fluctuation.In the paper the noise is removed efficiently by characteristic of like-impulse noise in the wavelet domain and iteractive adaptation method.Test results of synthetic and field data show that corrected signal is of high correlation coefficient and less energy loss compared with raw signal, and is high accuracy approximation of raw signal.The like-impulse noise of field data is removed efficiently.