Denoising of spectra with no user input: a spline-smoothing algorithm



A new method has been developed for denoising a spectrum using cubic-spline smoothing, which requires no user input or judgement, yet outperforms other more conventional algorithms. It relies only on the assumptions that the spectrum is sampled at a resolution much higher than the full-width at half maximum of any peaks, and that the noise is approximately Gaussian distributed, although any other type of noise distribution is acceptable so long as the type of distribution is known. Copyright © 2010 John Wiley & Sons, Ltd.