Denoising of Raman Spectroscopy Signals

• Online Access: http://hdl.handle.net/2047/d10008949

The detection system (Charge-Coupled Device, CCD) in Raman spectroscopy measurements are affected by spurious signals and noise, mainly produced by cosmic rays, shot noise and thermal noise. Generally, due to the nature of the noise signals, the spectrum estimation is divided in two sequential stages. The first stage removes the impulsive noise caused by cosmic rays. The second stage attempts to remove the rest of the noise, it is assumed that the statistics of the noise follows a Poisson process. In this work, the algorithm for removing the impulsive noise is based on a system which uses both a median filter and classic pattern recognition techniques. For the second stage is considered the Wavelet transform like alternative to denoise the spectra and it is compared with the classical smoothing method of Savitzky-Golay. The implemented algorithms are tested with synthetic and real spectra, real spectra are from Raman Imaging of biological materials which were provided by the research group led by professor Max Diem at Northeastern University research group led by professor Max Diem at Northeastern University. The algorithms are useful for all software tools that analyze Raman spectroscopy data.