Experimental and Data Analytical Approaches to Automating Multivariate Curve Resolution in the Analysis of Hyperspectral Images

Abstract Analyses of hyperspectral images with multivariate curve resolution (MCR) can be enhanced with the use of automated data preprocessing and improved MCR methods described and demonstrated in this chapter. These new approaches serve to greatly reduce or eliminate the need for user input and to increase the success, sensitivity, and accuracy of hyperspectral image analyses. We have pioneered the use of a dark spectral region, which is a general approach that can be introduced into any hyperspectral imaging system, to automatically remove offsets, structured noise, and uninformative pixels from the raw spectral images without the requirement of user input. The dark spectral region can also be used to minimize mixing between MCR-estimated spectral components to further improve the accuracy of the final pure spectra and their corresponding concentration images. The success of these improved preprocessing and MCR analysis methods is demonstrated with both realistically simulated images and experimentally measured spectral images.