Raman spectroscopy is an essential component in spectroscopy and materials science. It’s widely used in the field of chemistry. However, Raman spectrum is hard to precisely observe because of the interference of noises. This will lead to an inaccurate analytic result. Therefore, an effective preprocessing method is indispensable for Raman spectrum analysis. In this thesis, we focus on preprocessing Raman imaging data of a mixture, which is consisted of two substances RhB and AGN. We will calculate an intensity value for each of them from thousands of spectrums. The preprocessing algorithm can be divided into four parts: normalization, noise reduction, background removal and peak detection. First of all, we scale the input data so that all value is within the range of 0 and 1. After that, we use Savitzky–Golay filter to denoise and remove the fluorescence background of the Raman spectrum. Finally, we identify the peak positions and measure its value of the Raman spectrum. As a result, we can quantify the intensity of these two substances and use their intensity ratio to estimate their composition ratio in the mixture.