A Multiple Linear Regression–Based Algorithm to Correct for Cosmic Rays in Raman Images

Abstract

Raman imaging is a powerful technique for simultaneously obtaining chemical and spatial information on diverse materials. One of the most common detectors used on Raman equipment is the charge coupled detector (CCD) due its high sensitivity. However, CCDs are also sensitive to cosmic rays, that generate very narrow and intense signals: cosmic ray spikes. Since these peaks can be very intense and numerous, it is important to eliminate them before any data analysis. Some methods to do this use comparison of neighboring pixels to identify spikes, but when using the line-scanning acquisition mode, it is common that these spikes appear in two or more pixels close together. Thus, in this work, a new algorithm has been developed to correct for cosmic ray spikes in Raman images, based on multiple linear regression (MLR). This algorithm takes less than 1 min in images with more than 70,000 spectra and removes all spikes, even those at low intensity.