Respiratory sounds contain crucial information concerning the physiologies and pathologies of pulmonary system. Computerized respiratory sound analysis can provide an objective evaluation of the respiratory function. Usually, the real environments of respiratory sound measurement are noisy. The less clarity of respiratory sounds could cause the utilization of computerized respiratory sound analysis technology in normal circumstances unfeasible. The objective of this paper was to evaluate the performance of least mean square (LMS) adaptive filter, the dual sensor spectral subtraction algorithm (DSSS), and independent component analysis (ICA) in eliminating the environmental noises from respiratory sounds. The performance analysis of these three methods was quantified by relative error of power spectral density between the clean lung sound and the denoised lung sound. Our experimental results indicate that the DSSS algorithm has a relatively better performance in removing the environmental noises at low SNR levels (5 dB and 10dB).