Optimizing Image Denoising for Long-Range Brillouin Distributed Fiber Sensing

Linear and nonlinear 2-D image processing approaches are analyzed with the aim of removing the noise from data acquired by distributed optical fiber sensors based on Brillouin optical time-domain analysis (BOTDA). The impact of the filter parameters on the denoised data is analyzed, especially for the nonlinear image denoising method called non-local means (NLM). In particular, an optimization procedure to find the optimal parameters of the NLM method for BOTDA data denoising is proposed. The described optimization procedure has enabled, to the best of our knowledge, the first experimental demonstration of a conventional BOTDA scheme (i.e., with no modifications in the layout) capable of measuring along a 100-km sensing range over a 200-km fiber loop, using a spatial resolution of 2 m, a frequency sampling step of 1 MHz, and reaching a frequency uncertainty of 0.77 MHz with 2'000 averaged time-domain traces. The experimental sensing performance here achieved has been evaluated with a figure of merit (FoM) of 225'000. This is the highest FoM reached without hardware sophistication in BOTDA sensing.