To improve the signal-to-noise ratio of a Brillouin optical time-domain analysis (BOTDA) system, reduce the cumulative average number, and improve the real-time performance while ensuring measurement accuracy, a BOTDA scattering spectrum image denoising algorithm based on Armijo line search is proposed. The method uses the anisotropy of the partial differential equation from the perspective of energy diffusion to ensure that the noise-reduced image has good edge-holding characteristics and improves the measurement accuracy of the sensing system based on local features. The Armijo retrospective search method is used to adaptively select the steepest descending step size, and 256 cumulative average BOTDA experimental data are denoised. The best noise reduction effect can be achieved in just two iterations, which effectively reduces data acquisition time, thereby improving the real-time performance of the system.