Essential Spectral Pixels for Multivariate Curve Resolution of Chemical Images

Abstract

We propose a methodology to select essential spectral pixels (ESPs) of chemical images. These pixels are on the outer envelope of the principal component scores of the data and can be identified by convex-hull computation. As ESPs carry all the linearly mixed spectral information, large hyperspectral images can be dramatically reduced before multivariate curve resolution (MCR) analysis. We investigated chemical images of different spectroscopies, sizes, and complexities and show that the analysis of full data sets of hundreds of thousands of spectral pixels only require a few tenths of them.

Exploring local spatial features in hyperspectral images

“…The score plot provides a graphical representation of the variation of the morphological descriptors across the wavelet subimages whereas the loading plot accounts for their variation across the spectral channels. As it can be assumed that the most extreme score values are the most informative, as recently pointed out in MCR context, 26 we will focus on the 10 most extreme scores values along PC 1 and PC 2 in Figure 2C. The corresponding points are labeled according to their respective descriptor name and to the wavelet subimage from which such a descriptor has been calculated.…”

Hyperspectral Imaging (HSI) for meat quality evaluation across the supply chain: Current and future trends

“…Besides, MCR (multivariate curve resolution) method is also applied to the processing of HSI data differently. MCR unfolds the 3D HSI data into a two-dimensional data matrix and then performs data reduction based on this simplified data set ( Ghaffari et al, 2019 ). Ghaffari et al (2020) came up with MCR-ALS (multivariate curve resolution alternating least squares) method to compress and reduce HSI data.…”

Short-Wavelength Infrared Hyperspectral Imaging for Biomedical Applications

“…Separating (i.e., resolving) the mixed spectral signals-in pure spectral profiles and relative concentrations-is a challenging field of research, and several methods have been proposed. 148,149 Among other methods, self-modeling mixture analysis approaches are unsupervised techniques aimed at finding-without any previous knowledge of the mixture-the pure/purest variables and their concentration profiles in a series of measured spectra. By definition, a pure variable accounts for one component and has an intensity proportional to its concentration.…”