According to the similarity of information in Brillouin optical time-domain analysis (BOTDA) sensing image in spatial domain, Kuwahara filtering method for sensing image denoising is proposed. The central element value can be restored according to the correlation among adjacent pixels in filtering window. The results show that the window size of Kuwahara filter should approach to the theoretical value of spatial resolution. The proposed method can make an average signal noise ratio (SNR) improvement of 6.7 dB and Brillouin frequency shift (BFS) error decrease of 0.58 MHz for BOTDA sensing images in different SNRs, without distorting spatial resolution. The method improves the performance of the sensor without increasing the BOTDA sensor hardware, so it has the potential application to high-resolution, long distance sensing and the other types of distributed optical fiber sensors.