NLM Parameter Optimization for $\varphi$-OTDR Signal

The filtering parameters of non-local means (NLM) that seriously affect the denoising results are usually optimized based on the noise level estimation. But the noise level is unknown in many cases. In this paper, a simple method for optimizing NLM parameters based on autocorrelation function (ACF) is proposed from the perspective of signal estimation, which uses the periodic lag of the autocorrelation function to estimate the vibration signal. This method is verified with a hypothetical clean image corrupted by additive noise, and the NLM parameter optimization results are compared with those obtained by two traditional methods based on peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) respectively. Then it is tested with the practical vibration signal of a phase-sensitive optical time-domain reflectometer (
$\varphi$
-OTDR). The fidelity and spatial resolution of the denoised practical signal are analyzed. This work helps to improve the performance of
$\varphi$
-OTDR and signal recognition.