Image processing is proposed to enhance the performance of Brillouin distributed fibre sensors. The technique exploits the two-dimensional nature of the measurements, so that each frequency-position pair is assimilated to a pixel of a noisy image. Based on the level of redundancy existing in the two-dimensional information, the method offers unmatched denoising capabilities when compared to classic unidimensional denoising methods, even if those ones are consecutively used in distance and frequency domains. With no modification of the basic configuration, up to ~14 dB SNR improvement is experimentally demonstrated with unobservable loss of spatial resolution. A figure-of-merit of 115’000 is verified.