In order to ensure the safe operation of oil and gas production systems, real-time online defect detection can play an important role. Among them, the first-hand magnetic eddy current signals can effectively identify the defects existing in the oil and gas pipeline, thereby avoiding serious casualties and economic losses. Aiming at the noise interference problem in signals, this research proposes a comprehensive adaptive noise reduction method based on ensemble empirical mode decomposition (EEMD) method and an energy-based adaptive thresholding method. The detailed steps are as follows: Firstly, a noisy signal is randomly selected in the defect signal database, and then EEMD is carried out to obtain a series of intrinsic mode functions (IMFs). Secondly, the distances measure method and the probability density function are used to identify the high noise IMFs and the low noise IMFs. Thirdly, an energy-based adaptive thresholding method is used to remove the noise of the high noise IMFs. Finally, the signal is reconstructed by combining the low noise IMFs with the high noise IMFs after noise reduction. The result of the proposed noise reduction method is compared with the results of other conventional methods. It is superior to other noise reduction methods in terms of signal-to-noise ratio, mean square error and percent root mean square difference. Therefore, the proposed noise reduction method is efficient and lays a foundation for pattern recognition of pipeline corrosion defects.