Denoising Approaches in LIBS Imaging

Optimization of denoising approaches in the context of ultra-fast LIBS imaging

Ultra‑fast Laser‑Induced Breakdown Spectroscopy (LIBS) has emerged as a powerful technique for rapid, multi‑elemental analysis with minimal sample preparation. In particular, μLIBS‑Imaging, which combines spatial resolution with elemental specificity, is gaining traction across applications in industry, geology, forensics, and biomedicine.

Historically, μLIBS systems have operated at sub‑100 Hz laser repetition rates, making high‑resolution mapping time‑consuming. Transitioning to kilohertz (kHz) lasers drastically cuts acquisition time, but this speed boost introduces significant challenges: weaker plasma emission, degraded signal-to-noise ratio (SNR), and cumbersome data volumes.

The new research published in Spectrochimica Acta Part B, addresses these challenges. The researchers evaluate and compare 5 denoising methods: Savitzky–Golay smoothing, Fast Fourier Transform (FFT), wavelet thresholding, Whittaker filtering, and Principal Component Analysis (PCA). The objective is to enhance SNR in fast μLIBS imaging.

For the experimental setup the group use the Cobolt Tor XE 1064 nm, 10 µm spatial resolution, testing a biological sample: an epoxy-embedded rat kidney injected in gold nanoparticles.

Overall the study shows that PCA (Principal Component Analysis)  is by far the most effective method offering a better enhancement than the other methods, providing an important SNR enhancement.