Conventional Brillouin optical time-domain analyzer (BOTDA) requires the process of frequency scanning, which is very time-consuming. The data acquisition speed of the conventional BOTDA is too slow to realize dynamic sensing. The use of the optical frequency comb-based BOTDA can avoid the frequency scanning and speed up the measurement. But, to achieve dynamic measurement, it is also necessary to reduce the time for signal averaging to further improve the measurement speed. However, few averaging times lead to poor signal-to-noise ratio (SNR), which will limit the sensing performance of the system. Block-matching and 3D filtering (BM3D) algorithm is proposed to achieve the denoising of sensing signals in the sweep-free BOTDA using intensity-modulated direct-detection optical orthogonal frequency division multiplexing. The impact of the BM3D parameters on the denoising performance is analyzed, which is used to realize parameter optimization. To demonstrate the denoising performance of BM3D under different noise levels, white Gaussian noise with different standard deviation (STD) is added to the original sensing signal. When the STD of the added noise is 0.07 mV, the SNR after denoising using optimized BM3D is improved by 9.7 dB for the Brillouin gain spectrum (BGS) and 10.6 dB for the Brillouin phase spectrum (BPS), respectively, and the frequency uncertainty is reduced to be below 1 MHz. For a wide range of noise levels, BM3D has shown excellent denoising performance, and the average SNR after denoising is improved to be 16.8 and 15.6 dB for BGS and BPS, respectively, and the corresponding frequency uncertainty is reduced to be below 1.2 MHz.