Aiming at the problem of weak signal and reduced measurement accuracy caused by noise interference in Brillouin Optical Time Domain Reflectometer (BOTDR), a wavelet threshold denoising method based on Brillouin gain spectrum two-dimensional image is proposed to improve the measurement accuracy of BOTDR system.Analyzed the principle of microwave frequency sweep BOTDR measurement and the characteristics of two-dimensional Brillouin gain spectrum image, designed a two-dimensional image wavelet threshold denoising algorithm flow, built a microwave scanning frequency BOTDR experimental device for denoising performance verification, and determined the optimal parameter matching scheme for two-dimensional wavelet function, degrade level, threshold regulation, and threshold function based on noise reduction performance comparison.The experimental results show that using the two-dimensional image wavelet threshold denoising algorithm, under the condition of accumulating an average of 500 times, the Brillouin frequency shift fluctuation range of the 10km sensing fiber end temperature range before and after denoising is reduced from ±4.61MHz to ±2.59MHz, and the root mean square error is reduced from 2.24MHz to 0.99MHz；Before and after noise reduction, the Brillouin frequency shift fluctuation range of the 50km sensing fiber end temperature range decreased from ±41.25MHz to ±2.83MHz, and the root mean square error decreased from 33.59MHz to 1.69MHz.Research has shown that using this noise reduction scheme can effectively improve the accuracy of BOTDR measurement.